Programmable resistors are the key building blocks in analog deep learning, just like transistors are the core elements for digital processors. By repeating arrays of programmable resistors in complex layers, researchers can create a network of analog artificial "neurons" and "synapses" that execute computations just like a digital neural network. This network can then be trained to achieve complex AI tasks like image recognition and natural language processing. Nanosecond protonic programmable resistors for analog deep learning, Science (2022)
Automated discovery of fundamental variables hidden in experimental data Nature Computational Science volume 2, pages 433-442 (2022) Abstract All physical laws are described as mathematical relationships between state variables. These variables give a complete and non-redundant description of the relevant system. However, despite the prevalence of computing power and artificial intelligence, the process of identifying the hidden state variables themselves has resisted automation. Most data-driven methods for modelling physical phenomena still rely on the assumption that the relevant state variables are already known. A longstanding question is whether it is possible to identify state variables from only high-dimensional observational data. Here we propose a principle for determining how many state variables an observed system is likely to have, and what these variables might be. We demonstrate the effectiveness of this approach using video recordings of a variety of physical dynamical systems, ranging from elastic double pendulums to fire flames. Without any prior knowledge of the underlying physics, our algorithm discovers the intrinsic dimension of the observed dynamics and identifies candidate sets of state variables. Cite this article Chen, B., Huang, K., Raghupathi, S. et al. Automated discovery of fundamental variables hidden in experimental data. Nat Comput Sci 2, 433-442 (2022).
Thanks for clearing that up. But that is strictly an inside joke among math persons. It would have been equally "funny" to us non-math folks if she had said, "...born knowing long division." 🤪
@@Johnboy33545 Well, I'm glad you're not making crass generalizations or anything. But, hey, if antipsychotic medications make you see the humor in obscure things, good for you! 😜
@jim Sure, but then when new types of lights are added, they’ll be bamboozled and in-need of more data to co-analyze We’d already put context together to know
There are many people who would have trouble consistently identifying a picture of a traffic light. That is why captcha gives people several chances at different identification tasks instead of rejecting you the first time you miss one.
They learn what you teach them. Another big problem is the idiots telling the doubting supercomputer to „shut up and calculate“. The analogy is sufficiently comparable to what children are told in school and end up incapable to recognize the context of a problem and just do calculations.
How exactly does a computer "struggle"? What experience indicates to you than a computer is "struggling"? What exactly do you mean by "struggle or struggling"? Is it not the case that computers are more similar to men (human beings) than men are to computers those that are art and computers? May it not be that computers reflect indicate something about the nature of the associative apparatus of men (human beings)?
Good vid Sabine. Lots of tangents to explore. The issue of memories not being localized in the brain is something I'm fascinated by. It makes me wonder if we mistakenly think of memories as "recordings" of "data" when they're probably something else entirely. The unreliability of human memory and its ability to severely distort experience suggests we don't "record" experiences, but somehow real time experiences get severely filtered and then represented or imprinted in a manner that reuses parts of many other memories. This could accountant for the lack of localization of individual memories, their unreliability and the seemingly limitless "storage space" in the brain.
Woah the heck you doing here Robbie . Go watch shining for 69th time and tell me what jack and Dany's red clothes have to do with river of blood. Off now
Memory is not a "recording". It is an organic part of the mind and memories vary to serve the mind's purposes. That is why only 10% of memories older than 30 years are true. New research suggests that memory's main purpose is future planning and they are altered according the the success or failure of plans thus created. Every memory is altered every time it is accessed.
@@mikemondano3624 If we take an example of a learnt skill, recall will actually solidify the memory or set of memories. I think different types of memories are stored and recalled differently because they serve different purposes. The steps involved in making a spear will have high accuracy because you would get worse at performing the skill, not better. But storing memories that have some emotional significance might change depending on your current life, who you know, etc, because some factor other than "truth" or "fact" might come into play. Evolution is our guide here. It's all about _purpose, advantage, usefulness._ There's also the _subconscious_ vs _consciousness_ aspect to consider. Memories are being accessed and interpreted, and used all without our awareness _constantly._ Do the studies on memory focus on conscious memory necessarily because they depend on questionnaires, for example? If so, we'll only be getting a small part of the memory picture from those studies.
I certainly remember new things if I can fit them into an already existing framework/picture in my mind like an additional piece of a puzzle that makes additional sense.
I am following neuroscience news and since so little is known about brains, the goal to create an artificial brain-like computer seemingly recedes away as we chase it. Studies have shown that specific brain cells don't just work like special analog gates, for instance the recent interneuron study from the Max Planck institute showed neurons have interconnections that signal to inhibit activity of other neurons and that are active at different times than when neurons usually seem to be which means the interconnectivity is far more complex than just messages being sent forward to the next neuron to do things. I read that the ability to reform connections, neuroplasticity, together with this seems to imply that each neuronal connection, of which there's an order of magnitude more than there are neurons, is in a sense a specialised and reprogrammable computer itself, working in concert with every other nerve cell around. It gets complicated.
Indeed the most complicated Neural Network is extremely small compared to even say a mouse brain counting NN nodes VS Neural connections. These ANN don't have the other capabilities you have mentioned either.
There are also simulated spiking neural networks which require inhibitory interneurons. The networks we simulate are small compared to our brains though. Learning in these networks indeed always work via changing the connections between neurons. The deep neural networks that are famous now do not spike, they get constant values which could be said to represent firing rate. These do not require 'specialized' inhibiting neurons.
"each neuronal connection, of which there's an order of magnitude more than there are neurons" - On average, a neuron has around 7,000 synapses. Estimated to be around 100-140 trillion of them in total. Of the brain’s 86 billon neurons, 69 billion (77.5%) are in the cerebellum and are responsible for motor control and various bodily functions. The cerebellum is not involved in creating our intelligence or consciousness, which brings the maximum number that could be down a fair bit. Much of the remainder is involved in processing that takes place subconsciously, such as processing visual and auditory inputs. The counscious area/s of our brain just receives the processed and filtered output. We don't yet know how much of our brain is involved in creating our conscious experience but it seems to be concentrated in the area of the parietal cortex, the occipital cortex and part of the temporal cortex and probably involves less than 12% of the brain's neurons. This is the area right at the back of the brain. This is why the guy who lost most of his brain is still here with us. He may have some behavioural difficulties and problems with memory but he's still the same guy. Had he lost just this rear part of the brain, he'd be gone for good - regardless of how well the rest of his brain functioned, there'd be no-one home. In 2024, British company, Graphcore, will commision their supercomputer, the Good Machine, built from their own IPUs (Intelligence Processing Unit). This supercomputer will support up to 500 trillion parameters (a parameter is the equivalent of a synapse). That's around 4X the number of synapses in the human brain. It will have 10 exa-flops of AI floating point compute and 4 Petabytes of memory with a bandwidth of over 10 Petabytes/second. The memory is distributed throughout each processor and most tasks will not require the use of any external memory. The goal of the founders is to produce a superhuman inteligence. Early investors in the company include DeepMind founder, Demis Hassabis and the founders of OpenAI. The Good machine will be used to run DeepMind AIs at some point. - Just to bring you up to date a bit. Just what the latest generations of supercomputers, combined with the latest language models will be able to achieve is not yet clear. Will they be able to surpass human intelligence? No-one knows. The Good machine may have the capacity but the human brain is incredibly complex. Even it's physical structure is. For example, it's made up of interconnected neural networks. Each of these networks consists of about 100 neurons with a total of 700,000 synapses. There are aprox. 300 million of these networks in the human brain and they are connected together hierarchically. Good luck trying to replicate that! We still have little idea how conscioussness emerges too and so trying to predict when we might be able create one artificially is a pretty futile exercise. Still, it will be interesting to see what the coming generations of supercomputers and AI models bring. Demis Hassabis thinks it's possible to have intelligence without consciousness. We might not have long to wait to find out.
Such an awesome video! Love the 80's computer in the thumbnail! Fun fact: the Cantonese word for computer, 電腦 (deen noh), literally means "electric brain". I think that symbolizes the similarities of the brain and the computer in such a beautiful way.
Recently found your channel! I’m enjoying all your videos. I’m finally understanding a few basic scientific concepts, learning a few new ones, and laughing every now and then. Thanks so much for the content!
It appears complex because the brain is refusing to understand and explore itself. It is the result of a program (algorithm) encoded in our genes that is executed slowly over a period about an year. It is also in an autopilot mode and can learn and explore and modify itself. If we can decode our genes we shall surely find it very modular and structured- I guess!!
Yet it is our ego that thinks we are so much better than computers, yet we are equally unable to work outside what our "software and hardware" allows :D
The history of human self-understanding is dominated by rationalizations (i.e., mental models). The better models explain things well at the level of the behaviors they are considering but none explain things well at lower or higher levels. We simply have no clue as to how the brain works bottom-up or top-down. Whether complexity is the primary obstacle is unknown.
I just love Sabine's unapologetic humor. It's so low brow and so dead-pan that I can't help myself but to laugh. The way she delivers it is like she's saying "look, I know this isn't funny, but I wanted to do this bit and you're not going anywhere... sucker".
@11:57 As a biologist, I learned the opposite. Babies _do_ have to learn these things; they do _not_ come pre-wired. See the Wikipedia article about "Object permanence" ... Anyone still remember their nose disappearing or getting detached when certain adults where present?
Yes it would really need some explanation how the object permanence is hard wired if actual science expect a baby to develop a first intuition of the concept object at the age of 6-7 months.
You may have heard of the diathesis-stress model, which we can generalize naturally to a predisposition-stimuli description of development. Not that we know much conclusively here, but I gravitate strongly to (pretty banal) models like these: both our neuroanatomy and social-physical upbringing are far from random and both matter. We’re neither fully preprogrammed, nor completely tabulae rasa. I don’t think anything in the “object permanence” article disproves this; Piaget theories were very coarse, and there is certainly room for nuance, even with more modern naïve/intuitive physics experiments.
I was thinking that, too. The entire point of the "peek-a-boo" game is to teach object permanence, and the reason children find it delightful is because they're surprised by it.
Don't babies stare when you perform magic tricks? Of course what is preprogrammed is probably more structural codes, certain methods of conceptualizing data etc. Because of this, human languages for example all fit into a narrow band of structural possibilities which human children find easy to learn. That's where we become really efficient.
I am a grunt, that uses spanners and hammers to make a living. I really like your videos because most of them talk of thing's I could never understand, but your delivery makes it almost possible.
Do you suppose that your occupation is what you*are*? Would you seek to exist if you ceased to perform whatever function it is that you perform? If you are not no more than whatever occupation you pursue or functions you carry out, what exactly are you? - What it mean to be a man (human being)?
Thanks Sabine. Just a couple of points regarding computers and abstract reasoning. Anthropomorphizing computers (while convenient and sometimes hilarious) can make it hard to think critically about these questions which belong to the domain of philosophy. "With suitable software, they are capable of abstract reasoning, just like we are." No - in fact computers are not capable of reasoning of any kind. The inputs and outputs of electrical systems are assigned semantic content by us. The information, while being processed by the computer, has no intrinsic meaning to the machine itself and is merely making its way from point A to point B. What's more, human minds have already done the reasoning from premises to conclusions in order to produce the very inputs computers are fed. The reason a computer can identify a picture of a cat is because we have fed it thousands of images and a human mind has already specified which of those images are images of cats. Take a look at John Searle's "Chinese Room" and "Wordstar" thought experiments for more information on this. "If you ask your computer software if Pi is a rational number, it will hopefully say 'no.'" Except that you can't "ask" a computer to consider whether Pi is a rational number. You can have it perform some calculations to verify that it doesn't appear to be a repeating decimal or other tests of number rationality, but at the end of the day you have to assign the semantic meaning to its outputs and make the actual determination that Pi is irrational.
"No - in fact computers are not capable of reasoning of any kind." If superhuman poker software, superhuman chess engines, incredibly strong image classification and labeling (including, for example, superhuman radiology diagnostics), machine translation, text and code generation, automated mathematics software like Metamath and Wolfram Alpha, etc. are not "reasoning" according to your definition of that word, you simply have the wrong definition. I invite you to consider how a computing procedure might genuinely said to be "reasoning" on its inputs: it doesn't matter that someone coded the reasoning procedure (and for ML systems, this isn't even really the case), nor does it matter that the outputs deterministically came from the inputs. "The information, while being processed by the computer, has no intrinsic meaning to the machine itself and is merely making its way from point A to point B" So what? Why does the computer need to be aware of what its behavior means to us in order for the information processing to amount to genuine reasoning? Why does "understanding meaning" in the very anthropomorphic way that most people (including Searle) use that phrase actually matter? An AND Gate implements exactly the same sort of thing we do when we think about logical conjunction; no part of using "and" as a step in human reasoning actually requires that the step be taken with any sort of self-awareness. "True reasoning" and what you might call "mere information processing" are one and the same. "The reason a computer can identify a picture of a cat is because we have fed it thousands of images and a human mind has already specified which of those images are images of cats." If supervised ML isn't kosher to you, take a look at how well unsupervised ML can find semantic meaning and implement analogy-based reasoning in systems like, say, word2vec. But supervised ML is more impressive than you make it sound: we create an architecture with millions or trillions of parameters, feed data, and the system *reverse-engineers the very reasoning procedure* that recognizes cats itself. If all a cat-classifying algorithm could do were compare images to saved training examples, you'd be right in calling it unimpressive. But modern supervised ML is impressive because our models apply well to genuinely new input. It gets a robust bearing for "catlike features" and often spots details that humans don't. If this doesn't look something like "reasoning" to you, you again have the wrong definition. A lot of the giant language models pretrained on human text have a surprising capacity to draw multi-step connections, generalize, summarize, and problem-solve. Keep in mind: this is with little-to-no explicit labeling. And the field is super new--expect models in 5 years to do even more stuff, even more reliably. How much more "reasoning"-y could a system be before people swallow their pride and agree that human reasoning is not so different?
Eh, this clearly is not what she meant, she could've worded it better but obviously no honest interpretation of what she said would lead you to conclude she thought the computers themselves were assigning the meaning to those abstract symbols. Rather, the point is that computers can deal with abstractions, so the fact humans can deal with abstractions is not unique or interesting. Yes, humans assign meaning to computers, but this is a separate point than the point she was trying to make. Modern day computers are not generally intelligent, so they cannot operate with independence, they cannot dress themselves, feed themselves, apply for jobs, drive to that job in a car, perform that job, pay their taxes, clean their rooms, etc. While a computer might be able to do 1 of these tasks, no modern AI can do all of them, so they cannot operate as an independent agent in society and require humans to constantly tend to them and tell them everything to do and cannot do things for their own sake. But this is just a factor of modern AI not being generally intelligent. The fact modern computers lack general intelligence, even with the most sophisticated AI, is true, but was clearly not the point she was getting at and is a separate topic.
Love it! I'll point out that even digital computers are analog at the level below the digits. That is, voltage in the circuits is analog, and we just arrange to apply a lot or a little, in order to make it easier to distinguish with other circuits, but every circuit has a range of voltages over which it's "zero" and a range over which it is "one". Things like FLASH memory might have as much as eight different levels in one cell, storing 3 bits as one of 8 voltages. A lot of the things like object permanence do have to be learned, though. That's why there's a peek-a-boo game. Baby brains aren't really fully understanding basic physics until about 3 years old. Also, people can't prove their own Godel string. Think about actually doing that. You'd have to hold in your head several times over the entire formal description of your head, and manipulate it without making any mistakes. You might argue that someone could *theoretically* do that, but then you're not talking about the humans that are actually walking around doing computations today, but talking about people with much better brains than we actually have, which would entail a different godel string. Also, love the dress and key, which I didn't notice until about the third reference. 🙂
But the same is true about neurons; the final signal is essentially digital. The neurotransmitter is an essential repeater just to see that the signal does stay within the limits of 0 and 1. How do we add? Adding integers is easy because we too use digital rules (basic counting). When we want to add more messy numbers, we slow down because of the limitations of the algorithm.
@@janami-dharmam Correct. If you look at the highest level down to the lowest level, most things flop back and forth between discreet and continuous a half dozen times. It's just a question of what level of detail you're looking at and how you're interpreting it.
Don't agree with that argument about the impossibility of a person calculating their own Gödel statement. It would require that you knew the "axioms" underlying your brain - probably the same for all brains - then following a standard procedure to calculate the appropriate self-referential statement, and because of Godel's theorem you've no idea if it is true or not. But because you've studied second order logic, you know it is true. This contradiction proves that your brain wasn't based on those axioms; but computers are based on a set of simple rules (axioms) so brains aren't computers. Thats basically Penrose's argument.
"A lot of the things like object permanence do have to be learned, though. " the fact that it develops at a certain consistent point is evidence that it is not learned. Another example of something that isn't learned is contact action. Given an apparent causal relation, babies will assume a hidden contact force.
@@MassDefibrillator I'm not sure it would be learned if it wasn't actually shown to the babies. And I can't imagine any even vaguely moral way of testing that. But you have a fair point.
Gödel's incompleteness theorem boils to be the following: Let S be a consistent and a complex enough theory without being too complex. Then S can simulate itself using propositions relating to numbers. Let's call the simulation S'. Gödel's incompleteness says, that there is a proposition P such that there doesn't exists a proof from S to P and there doesn't exists a proof from S to (not P). The simulation S' exists in the sense that for every proposition Q permitted by S, we can formulate a number theoretic proposition denoted Q' that is interpreted to mean that the simulation S' proves' the simulated version of Q. So we have a way to turn a proposition Q to a proposition Q' which is a proposition about the the Gödel number of Q. This procedure has the following 2 properties: 1) If S proves Q, then S proves Q'. (read If S proves Q, then S proves that the simulation S' proves' the simulation of Q). 2) If S proves (not Q), then S proves (not Q'). A proposition P is created using the simulation S'. The proposition P has this funny property that P is equivalent with the statement (not P'). S cannot prove P, because otherwise S would prove P' by property 1) and (not P') by the equivalent formulation of P and this cannot be by the consistency of S. Hence S does not prove P. Assume S proves (not P), which is equivalent with P'. Thus S proves P'. Since S proves (not P), by the property 2), S proves (not P') and by the consistency of S, this cannot be. So S does not prove (not P). Hence S is an incomplete theory. You brought up that a computer has proven Gödel's incompleteness. Penrose's whole point is that the computer has to use some kind of a foundation to make deductions/calculations. The computer itself is still some kind of a system. Even though with a suitable foundation the computer can prove the incompleteness theorem, the incompleteness theorem still applies to the computer since its calculations are governed by some system S. Hence by Gödel's theorem there will be statements that the computer cannot prove one way or the other and we humans can create such a proposition. In this sense human can transcend a formal system. This leads to a notion that the universe as whole cannot be reducible to a purely computational system. Penrose's understanding is that the only place non-computability is introduced to quantum physics is when there's the collapse of the wave function. Hence if quantum physics is the system explaining consciousness and consciousness has this property of transcending computationability, then the non-computational aspect of the quantum physics have to be a part in the explanation of conscious experience.
Isabelle! I never thought I'd see that in a popular science video. It takes me back to the 90s and a former life in Functional Programming and Theorem Proving.
2:11 🧠 5:08 Analog or Digital 6:55 How does a computer do what it does? How does a brain do what it does? Brain - Adapts Computer - Specific Purpose *Making computers similar to brains* 8:50 Neuralmorphic Computers 10:53 Memcomputers 12:45 Energy 13:52 Self-Repair 15:13 Parallel Processing 16:10 Infinite Time For Pi, record 62.8 Million Digits 17:39 Human Thought cannot be alogorithimed 18:58 19:28 Brain Vs Computer 20:15 Social Learning
A computer is -as I understand it a form of adding or calculating machine, socalled because it performs to duties of what an original human computer did- perform routine calculations only rather more quickly that men(human beings) although seemingly there is an Indian being of the passive sex or woman that can do square roots in her head faster than a machine computer or calculator, but how, god knows; she says she just 'sees ' the answer; do you know 'how' you see? How* do you do *knowing*, and What* is understanding? - a computer can neither 'know' nor understand, but if you could work out how a computer works that would tell you all you need to know about the chap that made it or rather designed it, which brings one within an inch of understanding the loon that invented the mister god fantasy arrived at the queer idea, given that men (human beings appear to be little more than organic computers with organic software an hard ware, but the idea that mister Evolution designed and programmed men is as daft as the mister god fantasy-but the two ideas are broadly similar barring the followers of the religion scientism often comes within a whisker anthropomorphism and the mister evolution fantasy and the child Dawkins get almost tearful if anyone suggests that they doo not buy the mister evolution fantasy, but what's to choose between mister evolution and mister god? - They as broad as they are long. O one view computers tell as much about their designer as men do their designer. The fact that they are obviously designed does not tell you a damn thing about the nature and identity of the designer- similiter a computer or computers, although I don't know, if the computer tells you a good deal about its designer if not the identity thereof, why not apply the same to men? I suppose one might wangle an apology for mister god out of the same general reasoning, but why "Mister" god, rather than god a thing? - which is far more rational and has the advantage of avoiding girly anthropomorphism, which really is very queer indeed.
Great video, as always! One thing where you made a mistake though is saying that deep down "on physical level" computers are digital. Exactly the opposite is true: all electronic devices are analogue under all the discrete abstractions. And as components get smaller and more complicated, we keep tripping on new ways physical nature messes up with our shoehorning it into discrete world. Translating continuous values of voltage into discrete-value and discrete-time logic is a pretty complicated matter, and it inevitably causes loss of potential efficiency in exchange for reproducibility and predictability. There is a famous case when an AI researcher used machine learning to train an FPGA to detect certain sound inputs (FPGA is basically a bunch of "binary" logic gates that you can program and connect as desired). The resulting network was significantly smaller and more efficient than anything a human engineer could make, it operated in continuos time (no clock) and utilised full range of voltages. But the downside were, no one had any idea how it worked (even though it was just a couple dozen gates), the same network won't work if you remove one of the gates that was unconnected to anything (it was still affected and affecting EM field, after all), and it wouldn't work on any other FPGA.
Computability is dual to non computability -- the duality of Godel's theorem! Analog is dual to digital -- binary or on is dual to off. Problem, reaction, solution -- the Hegelian dialectic. Solving problems is based upon the Hegelian dialectic. Thesis is dual to anti-thesis creates the converging thesis or synthesis -- the time independent Hegelian dialectic. Action is dual to reaction -- Sir Isaac Newton (the duality of force), forces are dual. Concepts are dual to percepts -- the mind duality of Immanuel Kant. Analytic (Mathematics) is dual to synthetic (Physics) -- Immanuel Kant. The brain is a duality machine! Duality creates reality. "Always two there are" -- Yoda. The left hemisphere is dual to the right hemisphere -- brain dominance, watch the spinning dancer:- www.medicaldaily.com/right-your-eyes-science-behind-famous-spinning-dancer-optical-illusion-336122 Master (Lordship) is dual to slave (bondsman) -- the Hegelian dialectic. Absolute truth is dual to relative truth -- Hume's fork.
Voltage is not "continuous" except in the macroscopic world. Time is not continuous but quantized with the smallest unit being the Planck interval. Everything is digital.
@@mikemondano3624 The British empire was built using the Hegelian dialectic so both sides in politics use it, left is dual to right. TV adverts use the Hegelian dialectic all the time, they create a problem say dirty teeth, there ia a reaction of disgust and horror and then a solution is provided, a new better tooth paste. If you watch the spinning dancer video that I provifded you can see the Hegelian dialectic in action as you mind is forced to either choose clockwise or anti-clockwise rotation -- this is hemisphere dominance (master & slave). Hegel is therefore correct and you have duality. "Philosophy is dead" -- Stephen Hawking. If Hegel is correct then Stephen Hawking is incorrect. Ignore Hegel at your peril. The Necker cube is a good visual example of duality. Forwards is dual to backwards, which way is the train moving?:- th-cam.com/video/kmO_OaYWFD8/w-d-xo.html Attraction is dual to repulsion, push is dual to pull -- forces are dual. If forces are dual then energy must be dual. Energy = force * distance -- simple physics. Energy is dual to mass -- Einstein. Dark energy is dual to dark matter. Energy is duality, duality is energy -- the 5th law of thermodynamics. Potential energy is dual to kinetic energy -- gravitational energy is dual. Electro is dual to magnetic -- Maxwell's equations, electro-magnetic energy is dual. Everything in physics is made from energy or duality -- your mind is creating duality right now, perceptions are being converted into concceptions (thinking). "May the force (duality) be with you" -- Jedi teaching. "The force (duality) is strong in this one" -- Jedi teaching. Yoda is correct as he understands the metaphysics of Hegel.
@@mikemondano3624 Correct is dual to incrorrect, right is dual to wrong. The one is defined in terms of the other. Antinomy (duality) is two truths that contradict each other -- Immanuel Kant.
and there is a reason for that ---------------------++++++++++++ sabine tongue scope ++++++++++++++++++++++++-------- ---------------Science ---- non questions ----pseudo science --- mumble jumble------ plus utra
Your combination of factual information and capricious, humorous comments is brilliant. I really think that adding jokes to the mix, makes it easier to retain the 'sachliche' information you are sharing. Thank you for increasing our synaptic flow! :-)
This statement at the beginning… 😅 I had to literally stop the video after first second and calm down to overcome my uncontrolled laughter. Even Monty Python couldn't make me laugh that quickly.
Dr. Hossenfelder, I love your sense of humor. I discovered your channel very recently, and I must say I love the way you present your content, that little humorous touch just makes it a million times better 😌
About 30 years ago, I started typing notes into my computer like work notes, social notes, how to notes, people notes, writing notes, movie notes, book notes… It recently dawned on me that everything stored on my USB memory drive was an extension of my mind and personality. All those things I wrote down over my adult life that meant something to me or still has meaning to me. Sort of a mind meld with a computer that holds an exact copy of meaningful memories and ideas. So glad I learned to do backups and backups to the backups as a retired computer programmer (and failed screenwriter).
Yeah, I have a drug problem too, stay strong we will make it to ai and then everyone will have the ability to live forever on a usb for a monthly fee, you can maybe make it as a museum director if you hang onto that usb until then?
I'm really impressed by how well you presented theory of what computers can and can't do. I was about to challenge your assertions about pi and all knowledge being digitzable, but you already knew! I'd add that Turing completeness is the level of computational ability that we're currently sure humans and computers can both do.
Computability is dual to non computability -- the duality of Godel's theorem! Analog is dual to digital -- binary or on is dual to off. Problem, reaction, solution -- the Hegelian dialectic. Solving problems is based upon the Hegelian dialectic. Thesis is dual to anti-thesis creates the converging thesis or synthesis -- the time independent Hegelian dialectic. Action is dual to reaction -- Sir Isaac Newton (the duality of force), forces are dual. Concepts are dual to percepts -- the mind duality of Immanuel Kant. Analytic (Mathematics) is dual to synthetic (Physics) -- Immanuel Kant. The brain is a duality machine! Duality creates reality. "Always two there are" -- Yoda. The left hemisphere is dual to the right hemisphere -- brain dominance, watch the spinning dancer:- www.medicaldaily.com/right-your-eyes-science-behind-famous-spinning-dancer-optical-illusion-336122 Master (Lordship) is dual to slave (bondsman) -- the Hegelian dialectic. Absolute truth is dual to relative truth -- Hume's fork.
Self evidently computers can do what they are designed or programmed to do, which is perform calculations, which not all human beings can do and certainly the writer can't do, but you can't expect computer to understand anything that they are not program to understand they do not have ideas or concepts and cannot hold images in the heads - having no heads, and thus cannot do imagination which for them is an advantage , but for human beings a weakness, but you might equally say that human beings can only do what they are programmed to do; both are species of machine, and it goes without saying that machines cannot choose when it comes to the reactions of their functions and that is equally true of computers end men (human beings)To describe men (human beings) as "conscious" is laughable; occasionally capable of a species or degree of consciousness but conscious, never.They cannot possibly have any experience of consciousness because they are dreamers; dreamers cannot experience consciousness.
Congrats for this excellent and very instructive video. I really enjoyed your definition of a "gallon" of water ("lots of water"). I was also surprised to learn that physicists would take pi as equal to 1, for I thougt it would be closer to 10 on a logarithmic scale, but I was mistaken (pi is less than Sqrt[10] = 3.1622...).
pi = 1? I was also a bit shocked at yet another thing physicists do that is weird. Like the sum of all positive integer being -1/12. I guess, as Sabine would point out, it must be true if the calculations come out agreeing with the observations.
I agree with Roger Penrose about the fact that the human brain use quantum mechanics to operate in detail. I'm convinced of this because all creative functions require knowledge of the short-term future on many small but meaningful details; computation alone is incapable of any truly creative action. And I believe, although I can't prove it, that both self-conscience and creativity require knowledge of short-term future on a battery of elements - matrixes, if you like. If so, sentient computers are out of reach of Mankind until we fix our knowledge of the quantum world; as we are currently limited by a Science that is only merely an operational "description" of our world. Thank you for the great video. Regards, Anthony
I dunno. Things like Dalle-1 & 2 etc that are able to emulate specific art styles of brilliant painters and combine all different elements together, to me is an example of creative function.
You are just speaking in voodoo magic word salad and don't know the definitions of the words you are even using. Quantum mechanics does not have anything to do with human ability to know the future in the short term. The whole point of a neural network is to be able to build a statistical model so that it can predict what inputs would produce some outputs. This is true for all neural networks, whether biological or artificial. Humans are more complex than anything we have created in software, but even though we are more complex, the principle is the same. We can predict the future because we've observed the past enough to form models capable of doing so. Predicting the future does not require some sort of magic voodoo, it just requires observing the past and forming models based upon it.
Very good and careful explanations. Personally I tend to agree with Daniel Dennett on consciousness, and so I am not forced to try to "introduce" uncertainty into the brain.
Seeing the light: Researchers develop new AI system using light to learn associatively Source: University of Oxford Summary: Researchers have developed an on-chip optical processor capable of detecting similarities in datasets up to 1,000 times faster than conventional machine learning algorithms running on electronic processors. The new research published in Optica took its inspiration from Nobel Prize laureate Ivan Pavlov's discovery of classical conditioning. In his experiments, Pavlov found that by providing another stimulus during feeding, such as the sound of a bell or metronome, his dogs began to link the two experiences and would salivate at the sound alone. The repeated associations of two unrelated events paired together could produce a learned response -- a conditional reflex. Co-first author Dr James Tan You Sian, who did this work as part of his DPhil in the Department of Materials, University of Oxford said: 'Pavlovian associative learning is regarded as a basic form of learning that shapes the behaviour of humans and animals -- but adoption in AI systems is largely unheard of. Our research on Pavlovian learning in tandem with optical parallel processing demonstrates the exciting potential for a variety of AI tasks.' Journal Reference: 1. James Y. S. Tan, Zengguang Cheng, Johannes Feldmann, Xuan Li, Nathan Youngblood, Utku E. Ali, C. David Wright, Wolfram H. P. Pernice, Harish Bhaskaran. Monadic Pavlovian associative learning in a backpropagation-free photonic network. Optica, 2022; 9 (7): 792 DOI: 10.1364/OPTICA.455864
As far as I can tell Dennett has said absolutely nothing about how phenomenal consciousness allegedly emerges from (or "is identical to") the brain--which is to say why neural activity should be accompanied by a first-person perspective. So there's not much to agree with, really. (Yes, I read _Consciousness Explained_ and found it comically mistitled.)
the efficiency thing really got driven home to me when the warehouse i work at switched to laser guided robots for most of the work. they have lasers all over them a GPS unit ect ect and they are quite slow, and cant deal with anything unexpected. throw a stick in front of it and it just stops and sounds a alarm. it cant go around it. Me on the other hand, in a very basic sense do the same job with 2 cameras and 2 microphones, and deal with almost anything. All powered by 2 cheese burgers and a few cups of coffee.
I think you gave a great overview, but this is a topic that could be explored for hours. Let me use one simple example. There are many reasons that we build digital logic circuits the way we do. One of the biggest is scalability. The amount of circuitry that we put into one of these devices is massive. And yet, we test and prove each unit correct in testing. We want each and every digital integrated circuit to work within specifications that are compatible with all the others of its type. An 8051 is an 8051 (and as a joke I used to tell young engineers if you need more than an 8051, 22V8 and a 555 to make something it was probably too complex). The same is not true with people. We expect them to work differently from one another. We have this emergent phenomenon called consciousness (assuming it is emergent). As far as I can tell, we greatly struggle to explain emergent things in our universe. Why does a pile of the right chemicals not self organize into life? Why does it? I think this notion of computation and thinking as the same thing is probably incorrect. Where did Einstein (yeah that guy) get his (maybe) original notions from? You can't analyze yourself into a vision. You can vision things and then analyze them. All of this leads me to believe the answers are way more complicated than we think they are.
@NewTube Channel I agree that human memory and processing are not like a digital computer. I have seen no science (not belief) about the whole body storage idea. It would imply that amputees or quadriplegics would lose significant memory, processing, or language skills. This is not obvious to me.
What’s most amazing about neurons is that they probably evolved from cells that were just really good at communicating with other cells. All of our thoughts exist because of a strange adaptation that came along somewhere between sponges and animals with bilateral symmetry. A completely random event perhaps, who can say for sure. That’s the biggest difference between something we build and something that just occurs naturally. When humans want to build a circuit we have a goal in mind and each part has a purpose (even if it’s not that well made), but in nature there is no goal. Things that live will continue to live by passing on their traits, and if they accumulate errors in their genetic coding perhaps that increases or decreases the odds of that. Animals and all other life on earth have arrived at their current forms by chance, not even the goal of survival is consciously considered in what forms emerge but rather plays a role passively. The way a nervous system is laid out is so alien to how we would design something that it would probably be easier to understand something built by aliens. Why? Because at least it too was designed.
Great analysis, thank you. Might I add another suggestion that brains are give purpose by emotions and that emotions are driven by needs and dopamine rewards. In other words we are programmed by our physical and social environment while computers are given purpose by us. A person without emotions ( such an anomaly has happened) cannot survive without guidance.
Are not what you call emotions merely functions that react mechanically automatically without any participation of any part of your common presence that is not only that function?
Computability is dual to non computability -- the duality of Godel's theorem! Analog is dual to digital -- binary or on is dual to off. Problem, reaction, solution -- the Hegelian dialectic. Solving problems is based upon the Hegelian dialectic. Thesis is dual to anti-thesis creates the converging thesis or synthesis -- the time independent Hegelian dialectic. Action is dual to reaction -- Sir Isaac Newton (the duality of force), forces are dual. Concepts are dual to percepts -- the mind duality of Immanuel Kant. Analytic (Mathematics) is dual to synthetic (Physics) -- Immanuel Kant. The brain is a duality machine! Duality creates reality. "Always two there are" -- Yoda. The left hemisphere is dual to the right hemisphere -- brain dominance, watch the spinning dancer:- www.medicaldaily.com/right-your-eyes-science-behind-famous-spinning-dancer-optical-illusion-336122 Master (Lordship) is dual to slave (bondsman) -- the Hegelian dialectic. Absolute truth is dual to relative truth -- Hume's fork.
@@vhawk1951kl Peter, emotions can be thought of as a calculation, but it is based on complex factors such as our needs, social context, current mood and baked in survival biases we inherit from evolution. What makes this different from a computer is that it is too complex to get a consistent result. This is analogous to the weather. In principle and as a statistical large picture, we can predict the weather a few days out, but not on a micro climate level.
Your videos are always informative and the straight-faced jokes are part of what makes them highly watchable and enjoyable. Thanks so much for the time and effort you put into making them.
Sabine, this is a great video. It makes me wonder, is there an upper limit on how much a human can learn? It seems like every generation builds so much new technology on previous generations' work, but it often requires knowing everything that came before. So I'm curious if we'll ever hit some limit.
Unless the size of your skull is infinite, then the amount of information you can store in it is definitionally finite. While information may be intangible, storing that information and processing it definitely takes physical resources, and physical resources always, always have limitations. Arguably, physics sets an upper limit to the amount of processing power. A computer the size of a planet is affected by how fast you can move information, which is restricted to the speed of light. A computer the size of the universe would take many billions of years to access and process information.
I suspect that every person who watches Sabine’s videos knows there is a finite limit somewhere. But I’m curious if we will hit some other limit first, or maybe the brain doesn’t even work that way. Maybe, as we learn things, we automatically discard things too. I don’t know, but I’m curious what that limit might be if it actually exists in any practical terms.
Computability is dual to non computability -- the duality of Godel's theorem! Analog is dual to digital -- binary or on is dual to off. Problem, reaction, solution -- the Hegelian dialectic. Solving problems is based upon the Hegelian dialectic. Thesis is dual to anti-thesis creates the converging thesis or synthesis -- the time independent Hegelian dialectic. Action is dual to reaction -- Sir Isaac Newton (the duality of force), forces are dual. Concepts are dual to percepts -- the mind duality of Immanuel Kant. Analytic (Mathematics) is dual to synthetic (Physics) -- Immanuel Kant. The brain is a duality machine! Duality creates reality. "Always two there are" -- Yoda. The left hemisphere is dual to the right hemisphere -- brain dominance, watch the spinning dancer:- www.medicaldaily.com/right-your-eyes-science-behind-famous-spinning-dancer-optical-illusion-336122 Master (Lordship) is dual to slave (bondsman) -- the Hegelian dialectic. Absolute truth is dual to relative truth -- Hume's fork.
I really enjoyed this one Sabine, thank you. I make computer games, and trying to make a computer program do unexpected positive things is a very interesting area of study. I wonder if the first indications that our computers are approaching human capability will be the sudden appearance of backwards baseball caps and refusals to perform....
Computability is dual to non computability -- the duality of Godel's theorem! Analog is dual to digital -- binary or on is dual to off. Problem, reaction, solution -- the Hegelian dialectic. Solving problems is based upon the Hegelian dialectic. Thesis is dual to anti-thesis creates the converging thesis or synthesis -- the time independent Hegelian dialectic. Action is dual to reaction -- Sir Isaac Newton (the duality of force), forces are dual. Concepts are dual to percepts -- the mind duality of Immanuel Kant. Analytic (Mathematics) is dual to synthetic (Physics) -- Immanuel Kant. The brain is a duality machine! Duality creates reality. "Always two there are" -- Yoda. The left hemisphere is dual to the right hemisphere -- brain dominance, watch the spinning dancer:- www.medicaldaily.com/right-your-eyes-science-behind-famous-spinning-dancer-optical-illusion-336122 Master (Lordship) is dual to slave (bondsman) -- the Hegelian dialectic. Absolute truth is dual to relative truth -- Hume's fork.
A Computer funnels a huge amount of memory through a small static matrix of deterministic decisions (processor). A Brain is a dynamic and huge self modifying matrix combining memory and computation. What we consider AI is a relatively small previously trained probability matrix doing deterministic decisions we cannot explain from looking at it.
Computability is dual to non computability -- the duality of Godel's theorem! Analog is dual to digital -- binary or on is dual to off. Problem, reaction, solution -- the Hegelian dialectic. Solving problems is based upon the Hegelian dialectic. Thesis is dual to anti-thesis creates the converging thesis or synthesis -- the time independent Hegelian dialectic. Action is dual to reaction -- Sir Isaac Newton (the duality of force), forces are dual. Concepts are dual to percepts -- the mind duality of Immanuel Kant. Analytic (Mathematics) is dual to synthetic (Physics) -- Immanuel Kant. The brain is a duality machine! Duality creates reality. "Always two there are" -- Yoda. The left hemisphere is dual to the right hemisphere -- brain dominance, watch the spinning dancer:- www.medicaldaily.com/right-your-eyes-science-behind-famous-spinning-dancer-optical-illusion-336122 Master (Lordship) is dual to slave (bondsman) -- the Hegelian dialectic. Absolute truth is dual to relative truth -- Hume's fork. Brahman (thesis, creator God) is dual to Shiva (anti-thesis, destroyer God) synthesizes Vishnu (preserver God) -- Hegel, Hinduism.
I have always thought of computers as translating machines. Humans figured out how to build machines that could translate problems into data and instruction that manipulated that data into something that makes sense to the human at the other end of the process. All of that design came from our ideas and needs. The brain is the product of evolution and is, as a result ,capable of things we struggle to understand while it is still not able to help you remember why you are in the kitchen. It's beer, wine and/or cheese... you are welcome. IBM '68-'99 Love your videos.
That's a pretty interesting approach. However, it's still true that the base function of computer is to add 1s and 0s - so it is a calculating machine very fundamentally.
What computers evolved from are devices designed to stepwise automate physical tasks. Things like mechanical systems to automate weaving fabric. However complicated they get in terms of abstractions and appearance, at the core is still a sequential stream of instructions to perform. The computational model that we arrived at is extremely useful for many things, but not so much for simulating a biological brain.
@@Jim-Tuner I think that's machines. Mechanical machines and later electrically running machines were built to replace manual work. Computers were in the beginning really meant for shortening the time of tedious math operations. Software was added late in the game to further specify mechanical operations. That's sort of the marriage of machines and computers.
@@susanne5803 People say that, but there is more to the story than that. Take a look at the 1803 jacquard loom and its control cards. The mechanism of the loom was the very influential in the work of the Babbage on his difference engine. the loom was a very basic computer with a sequential instruction set which a generated a mechanical result. In my opinion, its not possible to seperate the history of automated mechanical operations from the history of computing. There are people who differ and who consider "computers" strictly electronic devices created quite late for math calculations. But those electronic devices were in fact using mechanical structures in their design from far earlier.
@@Jim-Tuner Basil Bouchon used something akin to punch cards already 1725 for looms. He took the punch stripes from organs. Pinned drums were also used for barrel organs and music boxes. I still think of computers and machines as different. In a way computers are a specific type of machine helping with math operations.
For everyone in the comment section saying 'humans aren't computers', the Turing machine WAS LITERALLY MODELLED ON WHAT HUMANS CAN DO, like actually look at Turing original paper.
Great episode, I think once we crack the “feeling” problem , I.e once we work out the brain represents feelings, we can emulate human brain in computers.
I wonder if you understand that when you use the word "we" it indicates, points to, or means, the user of the term - that is*you*sunshine and his immediate interlocutor, and since you have no immediate interlocutor, he can only be referring to yourself or say "I" when you employ the term or word "we". Self evidently you can experience nothing of the experiences of others insofar as they are their own experiences and thus "we" is and can only possibly be, imaginary or fanciful To what exactly are you referring when you speak of the "squealing problem", and why is it "a problem"? What exactly do you seek to convey by feeling? - Sensation or some sort of mechanical reaction in the emotional function?
Presumably you mean once we figure out how to make a computer have a first-person perspective and thus capable of experiencing a feeling or anything else. That is, indeed, something of a toughie.
@@halfacanuck It might help the both of you if the both of you clearly understood that the term "we" indicates the user of the term and his immediate interlocutor and since you are short of immediate interlocutors to the tune of any at all, you are referring to yourselves so were better suited to say "I" rather than the imaginary "we"
@@vhawk1951kl The first definition of "we" in Merriam-Webster's dictionary is "I and the rest of a group that includes me". Thus linkin543210 and I (henceforth "we") are using it correctly, because we mean humankind, to which group we both (presumably) belong. Your sense of "we" as in "you and I" is the second definition which in this case is not the correct one. Your grammatical policing is in this case unwarranted.
Brilliant video, Sabine! Thanks! 😃 Two things I was thinking about... The first are false memories. I never saw something like that happen to a computer! The other is that I lost 75% of my field of view, because of a surgery I did when I was 13 years old. I don't see from my left eye, neither from the right half of my right eye. (I had a tumor between the two most important glands of the human body, I don't remember their names in English right now.) Either way, it's weird... Because I remember effects of 3d vision, like those tricks you do with your fingers... But I don't remember any other difference. And I didn't noticed from the beginning, I just realized that I wasn't seeing with my left eye when I blinked (still in the ICU). Really weird! Anyway, stay safe there with your family! 🖖😊
There is something called bit flip, a bit flip can be caused by cosmic radiation for example, switching a bit that was 0 to 1 or the other way round. That's similar to false memory I believe.
@@IngTomT Computers can definitely have random memory errors, but that isn't really the same as false memories. What happens in your brain is that your brain doesn't actually store memories with a lot of detail, instead you only remember the important bits and your brain tries to fill in the rest in a way that makes sense but isn't exactly the same as what really happened. And then the new modified version of the memory overrides the original, so you keep getting small changes that add up over time and can eventually give you a completely different memory than what actually happened. This also means that counterintuitively the more often you remember something the less accurate it probably is, even if it seems more vivid to you. You don't really get the same kind of effect in a computer for two reasons. First, we do compress data but when we uncompress it it's a relatively deterministic process and we don't really try to fill in missing details the way the brain does. I could see this changing pretty soon though, since stuff like AI upscaling for images is getting pretty good. More importantly though, since the data and the processing are done in different parts of the computer we always have to copy the data over to the CPU before doing anything to it and the original version isn't affected. So you can get things like JPEG artifacts in images, but they don't build on each other over time like in the brain. You'd have to deliberately save the processed data back to the disk for that to happen, which usually isn't what we want to do.
It's not actually cosmic rays. It's because there were radioactive isotopes in the ceramic they used for the chip casing! They are more careful about their materials now, and it's only a problem for electronics operating in hazardous zones.
I think it's hilarious how we can discount our own biology and pretend to know better than billions of years of evolution. Sure, I may forget why I'm in the kitchen sometimes, but I'm also maintaining perfect homeostasis and processing unfathomable amounts of sensory information while being both assaulted and fueled by these super nachos in front of me
Haven't seen the video yet. Going to watch it now, but I'm guessing yeah it's probably a kind of biological computing device. Only it works very differently from our binary computers we use in everyday life, since not only are multiple neurons connected to multiple other neurons. The signals and its effects on each other also vary wildly. This makes it far superior to most modern day computing. I wonder how many years it will take before we have something artificial (either silicon, biological or something else) that can match our brains.
Uhm, but you do know that computers are outperforming us in many tasks already, like image recognition for instance? You’re right, our brains work differently, but we’re on a good way to be outperformed by computing devices in any given task.
@@thisismyalias yeah I know, but those are pretty much single task oriented neural networks which still use binary computing. The real wonders will start happening if we have some kind of synergy between them. If you were able to actually program a brain I'm sure it would vastly outperform anything out there today. I just saw the video and the thing I'm most curious about is whether and how quantum computing has anything to do with us being conscious. I believe if we get far enough there will exist 'artificial' consciousness. We as humans always like to believe we are apart from nature and that our intelligence and, being conscious, is reserved for us humans. I believe this is due to the evolution of our big old ego and a bit of hubris. I believe consciousness is just a gradient shared between animals in the animal kingdom, including ourselves. I even believe that different people experience different kinds and values or a spectrum of consciousness. I mean just look at how easily our consciousness is altered not only by the smallest amount of substances, but even by emotion or time for example. There is a big chance that we will quickly be the inferior intelligence if we were to develop an A.I. that is able to iterate on itself. I hope don't experience it, but looking at the biggest picture possible I'd be okay with humanity being a "boot-loader" for something beyond our capabilities.
@@DeltaNovum I agree with everything you said. Regarding the question if quantum physics might be involved; I strongly believe that to be the case to some degree, although I do not believe in true randomness but some variant of the hidden variables theorem (and if not local, then global). And yes, let’s hope to not be around when AGI comes into existence and treats us the same way we treat everything we deem irrelevant or unworthy in our way forward.
@@thisismyalias indeed ^^. And I carry the same assumptions you do. You probably agree with Sabina about free will as well? I assume our reality to be deterministic and that true free will is not really a thing. In my daily life I just follow my biology and live like I do have free will, but deep down I believe it not to be true. I know many way more intelligent people than me don't agree. But as with many things in the science community I think fear of death and the ego are things that make us biased from time to time :).
@@DeltaNovum no but ai cant think on its own so we might never become inferior although it can also cant develop any physical laws or invent or discover anything regarding science and technology also our human brains are much more versatile as u have seen in the video and another reason is that an ai can be shut down by an emp attack but our brains survives it. u know that no computer can store information permanently becaz space radiation flicks 1s and 0s in storage device so pal i don't think ai can even surpass us in day to day reasoning and common sense and thus we will never become inferior, mark the point ai cant think on its own or invent anything but can make copies of its own which is not lethal, ai is also meant to help humans for specific purposes🤓 have a nice day dude
Why, because him and Hameroff figured there might be a way to store quantum States at room temperature, or the very likely possibility the brain is far more capable due to the microtubules being the basic processing nodes, instead of the higher level neuron "scientis" are focused on. Or the genious assumption that entanglement is very likely a basic function of our brain. Etc etc, Penrose is the closest intelligent scientist we have to Einstein. Sabine would wish she could stand in his shadow. I bet she doesn't even realise the another fundamental difference between computers and brains, brains don't really have a filesystem for looking up information, all information is passed to relevant regions, these regions correlate all incoming info and autonomously report if correlation exceeds certain threshold. That architecture is nowhere to be found in any current AI system, current AI is just at infant level, not even close to human brain capacity not now, not in a 100 years. Since these AI networks basically have no understanding of what they are "interpreting" it is a great danger to use them in safety critical systems, probably some high pricd will be paid, if people use this technology in critical cases. That is something Penrose is trying to tell us, and I have a lot more confidence in Penrose than some numbnut that owns this channel or her "clever" respondents.
In the interest of preserving history: I was in college before hand-held electronic calculators were inexpensive enough for starving students to afford. So I used a slide rule in my engineering and physics classes. Sabine, that's pronounced "slide-rule" not slide-ruler. I'm 77 years old. Your videos help me exercise my brain. Thanks so much for all the work and time........Radio valve
I too remember the slide rule but it was very expensive. My first interaction with a calculator was a mechanical swedish calculator (I forget the make) which could do the basic arithmetic with lots of noise. The electronic calculator I have used during my Ph.D. days had nixie tubes and discrete transistors. I have not seen calculators that had valves but did see a russian computer that used valves.
Computability is dual to non computability -- the duality of Godel's theorem! Analog is dual to digital -- binary or on is dual to off. Problem, reaction, solution -- the Hegelian dialectic. Solving problems is based upon the Hegelian dialectic. Thesis is dual to anti-thesis creates the converging thesis or synthesis -- the time independent Hegelian dialectic. Action is dual to reaction -- Sir Isaac Newton (the duality of force), forces are dual. Concepts are dual to percepts -- the mind duality of Immanuel Kant. Analytic (Mathematics) is dual to synthetic (Physics) -- Immanuel Kant. The brain is a duality machine! Duality creates reality. "Always two there are" -- Yoda. The left hemisphere is dual to the right hemisphere -- brain dominance, watch the spinning dancer:- www.medicaldaily.com/right-your-eyes-science-behind-famous-spinning-dancer-optical-illusion-336122 Master (Lordship) is dual to slave (bondsman) -- the Hegelian dialectic. Absolute truth is dual to relative truth -- Hume's fork.
I speculate it is possible to build individual neurons using programmable Uni-junction transistors without programming them. They are configured with a capacitor across the anode, and the gate, and powered in the micro-amp range with a reverse biased signal diode across the gate, and ground. A multi-meg-ohm resistor across B+, and the anode, and another capacitor across the anode, and ground. Each neuron is connected to other neurons via multi-meg-ohm resistors to their respective gates. Each neuron fires an electrical signal depending on the amount of energy that is stored in the gate capacitor that is electrostatically charged in proportion (synapses) to the voltage variation of output from sensors that are connected to the capacitor gate combination of each neuron . In other words, each neuron fires when a threshold is reached that is determined by previous firings of neurons, or many of them. Each neuron is inconsequential individually, but as a collective act as intelligent deterministic execution of process depending on the extent of sensor input. More sensor input, more intelligent process (decision making pathways that output to various machine functions, ie. motors, lights, sound, camera, etc., analogous to eyes, ears, touch, taste ) output from the multitude of neurons. More neurons, more sensory input, more intelligence. Its kind of a programmable resistor of sorts. I'm probably out of my mind, but I think it would work using millions of these hardware configured neurons. Memory is a function of the neuronal collective that functions in real time as a whole, never as discrete individual units of memory (one & zeros) that a microprocessor accesses for instance.
There is nothing to argue about. From mathematical perspective: 1. Our brain is dynamical system. 2. Any dynamical system can be represented (not just approximated) by two layer recurrent neural network. 3. Neural networks can be arbitrary precise implemented on any Turing machine. 4. Classical computers implement Turing machine. QED The question is how slow it would be. It is a concern since our brains seems to implement Non-deterministic Turing Machine. Though, we are on verge on second computing revolution. Not quantum, it’s first. I am about DNA computers. They are implementable and implement non-deterministic Turing machine therefore brain will be possible to effectively simulate on it. So the question is not whether we are computers or not, we are, any dynamical system can be implemented on computer and we can at least manually do computers job, so we present equivalent models of computations. The question is how to effectively model brain. And we cannot do it effectively on Turing machine, only on Non-Deterministic Turing machine.
I've never wanted to have pre marital relations more in my life. I love it when you talk nerdy to me..... the human brain is capable of multiple perceptions at a time. And if you're in is high enough you can practically see the future.
Man I wish my grandma was a computer. That'd be just so cool. Great video. The point Sabine makes about general purpose vs. specified purpose is a really good one. This is why I'm very skeptical that humans will develop generalized AI. There's really no market for one, and it's probably too complicated to do if we just decided we wanted to do it.
If you watch infants' and children's reacting to events in cartoons, you realise that they're based on quite sophisticated understanding of natural laws, especially physics, even if they could not verbalise them. A species that spent part of its evolution in trees had better understand gravity from an early age. Laughter is a combination of a scream at danger, and a reassuring coo, so it indicates the subject's reaction to an "impossible" situation.
Computability is dual to non computability -- the duality of Godel's theorem! Analog is dual to digital -- binary or on is dual to off. Problem, reaction, solution -- the Hegelian dialectic. Solving problems is based upon the Hegelian dialectic. Thesis is dual to anti-thesis creates the converging thesis or synthesis -- the time independent Hegelian dialectic. Action is dual to reaction -- Sir Isaac Newton (the duality of force), forces are dual. Concepts are dual to percepts -- the mind duality of Immanuel Kant. Analytic (Mathematics) is dual to synthetic (Physics) -- Immanuel Kant. The brain is a duality machine! Duality creates reality. "Always two there are" -- Yoda. The left hemisphere is dual to the right hemisphere -- brain dominance, watch the spinning dancer:- www.medicaldaily.com/right-your-eyes-science-behind-famous-spinning-dancer-optical-illusion-336122 Master (Lordship) is dual to slave (bondsman) -- the Hegelian dialectic. Absolute truth is dual to relative truth -- Hume's fork. Brahman (thesis, creator God) is dual to Shiva (anti-thesis, destroyer God) synthesizes Vishnu (preserver God) -- Hegel, Hinduism.
You're a remarkable woman. Thank you for your efforts. Edit: Your English is better than most native Americans. English can be a confusing disaster at times. I'll wager you're a great Mom.
Hi Sabine, John from Australia here, your English is great, better than a lot of native English speakers that I have known. Great content, three thumbs up! I live near a nuclear reactor, hehehehe.
By analogy with human designed neural networks, brain memory is most likely to be embodied in the pattern of the strengths of interconnections between countless neurons, which may be as simple as how many axons (activating connections) are passing from one neuron to another. These strengths develop as a result of 'learning', by analogy training the network. Perhaps the more often one neuron is stimulated to stimulate another, the stronger the interconnection becomes. Inputs take the form of the stimulation of a selection of targeted neurons, these then fire off to stimulate other neurons they are connected to. The pattern of weights or strengths of these interconnections activate interconnected feedback loops of neurons within the network. The result is a stimulus of a small number of neurons will tip the neural network into a stable state in which a particular pattern of neurons is active. This would form a response or a memory. Underneath it is the pattern of connection strengths which embody memory or response. This is in turn learned by training the network.
My brain is often a source to frustration. It forgets important information, and it can't shut down pain signals from health issues, that will never get better. It's age related. In other words, it's falling apart. Just like the rest of the body. And unlike a computer, the worn out drive can't be cloned to a fresh drive. Don't get old and sick. It's not worth it. Superb video. Your English is better than most native English speakers, by the way.
The faculty of learning is an important aspect of what the organic brain does, and remains an important frontier for computers. I expect that in this century, computers will begin to surpass humans in the ability to learn autonomously and autodidactically.
On that last note of thought, I learn greatly from TH-cam, objective (as objective as any human can be) creators like you, and even AI like the new Chatbot ChatGPT. I don’t need to worry about personal agendas, how much time the person has, and I don’t need to listen to the condescending and intimidating tone which some people express now and then.
... a machine that's very good at adapting to new situations with new problems! Gotta a love a smart and well-communicating person with that kind of humor.
Love the Dr. Who dress! :-)
I was wondering why she had this dress, because I never watched Dr. Who.
So, thanks for mentioning it ;-)
I'd love to see Sabine as the next Doctor! (So funny and no Gobbledygook!)
Programmable resistors are the key building blocks in analog deep learning, just like transistors are the core elements for digital processors. By repeating arrays of programmable resistors in complex layers, researchers can create a network of analog artificial "neurons" and "synapses" that execute computations just like a digital neural network. This network can then be trained to achieve complex AI tasks like image recognition and natural language processing.
Nanosecond protonic programmable resistors for analog deep learning, Science (2022)
Automated discovery of fundamental variables hidden in experimental data
Nature Computational Science volume 2, pages 433-442 (2022)
Abstract
All physical laws are described as mathematical relationships between state variables. These variables give a complete and non-redundant description of the relevant system. However, despite the prevalence of computing power and artificial intelligence, the process of identifying the hidden state variables themselves has resisted automation. Most data-driven methods for modelling physical phenomena still rely on the assumption that the relevant state variables are already known. A longstanding question is whether it is possible to identify state variables from only high-dimensional observational data. Here we propose a principle for determining how many state variables an observed system is likely to have, and what these variables might be. We demonstrate the effectiveness of this approach using video recordings of a variety of physical dynamical systems, ranging from elastic double pendulums to fire flames. Without any prior knowledge of the underlying physics, our algorithm discovers the intrinsic dimension of the observed dynamics and identifies candidate sets of state variables.
Cite this article
Chen, B., Huang, K., Raghupathi, S. et al. Automated discovery of fundamental variables hidden in experimental data. Nat Comput Sci 2, 433-442 (2022).
i note she has the key...
"I was upset to learn, though, that infants aren't born knowing Gödel's theorem"
I laughed a lot at this 😂
Thanks for clearing that up. But that is strictly an inside joke among math persons. It would have been equally "funny" to us non-math folks if she had said, "...born knowing long division." 🤪
@@genepozniak Actually, I'm a theoretical physicist and I know this theorem, but I didn't know the joke
@@mureebe1 That's because it's an inside mathematician joke. Man, scientists are SO cliquey. 🤣
@@genepozniak: I'm neither a mathematician nor a physicist but can see the humor and get the joke. So can most well read and curious people.
@@Johnboy33545 Well, I'm glad you're not making crass generalizations or anything. But, hey, if antipsychotic medications make you see the humor in obscure things, good for you! 😜
Me seeing a supercomputer struggle to identify a picture of a traffic light:
"Look what they need to mimic a fraction of our power"
@jim Sure, but then when new types of lights are added, they’ll be bamboozled and in-need of more data to co-analyze
We’d already put context together to know
DALL-E 2 understands millions of objects, self-taught. And can make great pictures quickly of any of them, and billions of things it has never seen.
There are many people who would have trouble consistently identifying a picture of a traffic light. That is why captcha gives people several chances at different identification tasks instead of rejecting you the first time you miss one.
They learn what you teach them. Another big problem is the idiots telling the doubting supercomputer to „shut up and calculate“.
The analogy is sufficiently comparable to what children are told in school and end up incapable to recognize the context of a problem and just do calculations.
How exactly does a computer "struggle"?
What experience indicates to you than a computer is "struggling"?
What exactly do you mean by "struggle or struggling"?
Is it not the case that computers are more similar to men (human beings) than men are to computers those that are art and computers?
May it not be that computers reflect indicate something about the nature of the associative apparatus of men (human beings)?
Finally a scientific channel that makes complex subjects more accessible and also bring humor to the table.
You got a new subscriber.
Good vid Sabine. Lots of tangents to explore. The issue of memories not being localized in the brain is something I'm fascinated by. It makes me wonder if we mistakenly think of memories as "recordings" of "data" when they're probably something else entirely. The unreliability of human memory and its ability to severely distort experience suggests we don't "record" experiences, but somehow real time experiences get severely filtered and then represented or imprinted in a manner that reuses parts of many other memories. This could accountant for the lack of localization of individual memories, their unreliability and the seemingly limitless "storage space" in the brain.
Woah the heck you doing here Robbie . Go watch shining for 69th time and tell me what jack and Dany's red clothes have to do with river of blood. Off now
Memory is not a "recording". It is an organic part of the mind and memories vary to serve the mind's purposes. That is why only 10% of memories older than 30 years are true. New research suggests that memory's main purpose is future planning and they are altered according the the success or failure of plans thus created. Every memory is altered every time it is accessed.
@@mikemondano3624 If we take an example of a learnt skill, recall will actually solidify the memory or set of memories. I think different types of memories are stored and recalled differently because they serve different purposes. The steps involved in making a spear will have high accuracy because you would get worse at performing the skill, not better. But storing memories that have some emotional significance might change depending on your current life, who you know, etc, because some factor other than "truth" or "fact" might come into play. Evolution is our guide here.
It's all about _purpose, advantage, usefulness._
There's also the _subconscious_ vs _consciousness_ aspect to consider. Memories are being accessed and interpreted, and used all without our awareness _constantly._ Do the studies on memory focus on conscious memory necessarily because they depend on questionnaires, for example? If so, we'll only be getting a small part of the memory picture from those studies.
I certainly remember new things if I can fit them into an already existing framework/picture in my mind like an additional piece of a puzzle that makes additional sense.
I am following neuroscience news and since so little is known about brains, the goal to create an artificial brain-like computer seemingly recedes away as we chase it.
Studies have shown that specific brain cells don't just work like special analog gates, for instance the recent interneuron study from the Max Planck institute showed neurons have interconnections that signal to inhibit activity of other neurons and that are active at different times than when neurons usually seem to be which means the interconnectivity is far more complex than just messages being sent forward to the next neuron to do things.
I read that the ability to reform connections, neuroplasticity, together with this seems to imply that each neuronal connection, of which there's an order of magnitude more than there are neurons, is in a sense a specialised and reprogrammable computer itself, working in concert with every other nerve cell around.
It gets complicated.
Indeed the most complicated Neural Network is extremely small compared to even say a mouse brain counting NN nodes VS Neural connections. These ANN don't have the other capabilities you have mentioned either.
Quick note: There is no "the Max Planck Institute". There are dozens of them, so one should ideally always specify which one they're referring to.
There are also simulated spiking neural networks which require inhibitory interneurons. The networks we simulate are small compared to our brains though. Learning in these networks indeed always work via changing the connections between neurons.
The deep neural networks that are famous now do not spike, they get constant values which could be said to represent firing rate. These do not require 'specialized' inhibiting neurons.
@@jamielondon6436 Thank you, yeah I didn't realise this, the study at hand was from Max Planck Institute for Brain Research in Frankfurt!
"each neuronal connection, of which there's an order of magnitude more than there are neurons"
- On average, a neuron has around 7,000 synapses. Estimated to be around 100-140 trillion of them in total.
Of the brain’s 86 billon neurons, 69 billion (77.5%) are in the cerebellum and are responsible for motor control and various bodily functions. The cerebellum is not involved in creating our intelligence or consciousness, which brings the maximum number that could be down a fair bit.
Much of the remainder is involved in processing that takes place subconsciously, such as processing visual and auditory inputs. The counscious area/s of our brain just receives the processed and filtered output. We don't yet know how much of our brain is involved in creating our conscious experience but it seems to be concentrated in the area of the parietal cortex, the occipital cortex and part of the temporal cortex and probably involves less than 12% of the brain's neurons. This is the area right at the back of the brain. This is why the guy who lost most of his brain is still here with us. He may have some behavioural difficulties and problems with memory but he's still the same guy. Had he lost just this rear part of the brain, he'd be gone for good - regardless of how well the rest of his brain functioned, there'd be no-one home.
In 2024, British company, Graphcore, will commision their supercomputer, the Good Machine, built from their own IPUs (Intelligence Processing Unit).
This supercomputer will support up to 500 trillion parameters (a parameter is the equivalent of a synapse). That's around 4X the number of synapses in the human brain. It will have 10 exa-flops of AI floating point compute and 4 Petabytes of memory with a bandwidth of over 10 Petabytes/second. The memory is distributed throughout each processor and most tasks will not require the use of any external memory.
The goal of the founders is to produce a superhuman inteligence. Early investors in the company include DeepMind founder, Demis Hassabis and the founders of OpenAI. The Good machine will be used to run DeepMind AIs at some point.
- Just to bring you up to date a bit.
Just what the latest generations of supercomputers, combined with the latest language models will be able to achieve is not yet clear.
Will they be able to surpass human intelligence?
No-one knows. The Good machine may have the capacity but the human brain is incredibly complex. Even it's physical structure is. For example, it's made up of interconnected neural networks. Each of these networks consists of about 100 neurons with a total of 700,000 synapses. There are aprox. 300 million of these networks in the human brain and they are connected together hierarchically.
Good luck trying to replicate that!
We still have little idea how conscioussness emerges too and so trying to predict when we might be able create one artificially is a pretty futile exercise.
Still, it will be interesting to see what the coming generations of supercomputers and AI models bring. Demis Hassabis thinks it's possible to have intelligence without consciousness. We might not have long to wait to find out.
I want to say honestly that your newsletter is the first I actively went out of my way to sign up for. And I have no regrets.
Such an awesome video! Love the 80's computer in the thumbnail!
Fun fact: the Cantonese word for computer, 電腦 (deen noh), literally means "electric brain". I think that symbolizes the similarities of the brain and the computer in such a beautiful way.
Recently found your channel! I’m enjoying all your videos. I’m finally understanding a few basic scientific concepts, learning a few new ones, and laughing every now and then. Thanks so much for the content!
The irony of the brain being so complex is it trying to understand it’s own complexity.
It appears complex because the brain is refusing to understand and explore itself. It is the result of a program (algorithm) encoded in our genes that is executed slowly over a period about an year. It is also in an autopilot mode and can learn and explore and modify itself.
If we can decode our genes we shall surely find it very modular and structured- I guess!!
Yet it is our ego that thinks we are so much better than computers, yet we are equally unable to work outside what our "software and hardware" allows :D
@@XXveny
Especially after doing cocaine, crack and marijuana.
@@XXveny our “ego” is backed by flesh and blood, though.
The history of human self-understanding is dominated by rationalizations (i.e., mental models). The better models explain things well at the level of the behaviors they are considering but none explain things well at lower or higher levels. We simply have no clue as to how the brain works bottom-up or top-down. Whether complexity is the primary obstacle is unknown.
Ur show has definitely become one of my most favourite on TH-cam.
Easy to absorb.
Thank you ♥️
I just love Sabine's unapologetic humor. It's so low brow and so dead-pan that I can't help myself but to laugh. The way she delivers it is like she's saying "look, I know this isn't funny, but I wanted to do this bit and you're not going anywhere... sucker".
Absolutely love the Dr Who / Tardis dress with key. Sabine is awesome. So many funny one liners in this episode, thank you so much ❤
@11:57 As a biologist, I learned the opposite. Babies _do_ have to learn these things; they do _not_ come pre-wired. See the Wikipedia article about "Object permanence" ...
Anyone still remember their nose disappearing or getting detached when certain adults where present?
Yes it would really need some explanation how the object permanence is hard wired if actual science expect a baby to develop a first intuition of the concept object at the age of 6-7 months.
You may have heard of the diathesis-stress model, which we can generalize naturally to a predisposition-stimuli description of development. Not that we know much conclusively here, but I gravitate strongly to (pretty banal) models like these: both our neuroanatomy and social-physical upbringing are far from random and both matter. We’re neither fully preprogrammed, nor completely tabulae rasa. I don’t think anything in the “object permanence” article disproves this; Piaget theories were very coarse, and there is certainly room for nuance, even with more modern naïve/intuitive physics experiments.
I was thinking that, too. The entire point of the "peek-a-boo" game is to teach object permanence, and the reason children find it delightful is because they're surprised by it.
I had my 3 yo niece ask me for her nose back after about a half and hour without it.
Don't babies stare when you perform magic tricks?
Of course what is preprogrammed is probably more structural codes, certain methods of conceptualizing data etc. Because of this, human languages for example all fit into a narrow band of structural possibilities which human children find easy to learn. That's where we become really efficient.
Love the way you choose interesting scientific topics. Thank you.
I am a grunt, that uses spanners and hammers to make a living. I really like your videos because most of them talk of thing's I could never understand, but your delivery makes it almost possible.
Do you suppose that your occupation is what you*are*?
Would you seek to exist if you ceased to perform whatever function it is that you perform? If you are not no more than whatever occupation you pursue or functions you carry out, what exactly are you? - What it mean to be a man (human being)?
The one liners in this video are amazing xD These videos are immensely entertaining and even more informative. Thank you Sabine :)
We come for the science, we stay for those nuggets of the driest humour.
3:23 "fuzzy logic" was pretty good, but "admitting" she mortally threatens her own offspring @ 13:43 made me lol
Thanks Sabine. Just a couple of points regarding computers and abstract reasoning. Anthropomorphizing computers (while convenient and sometimes hilarious) can make it hard to think critically about these questions which belong to the domain of philosophy.
"With suitable software, they are capable of abstract reasoning, just like we are."
No - in fact computers are not capable of reasoning of any kind. The inputs and outputs of electrical systems are assigned semantic content by us. The information, while being processed by the computer, has no intrinsic meaning to the machine itself and is merely making its way from point A to point B. What's more, human minds have already done the reasoning from premises to conclusions in order to produce the very inputs computers are fed. The reason a computer can identify a picture of a cat is because we have fed it thousands of images and a human mind has already specified which of those images are images of cats. Take a look at John Searle's "Chinese Room" and "Wordstar" thought experiments for more information on this.
"If you ask your computer software if Pi is a rational number, it will hopefully say 'no.'"
Except that you can't "ask" a computer to consider whether Pi is a rational number. You can have it perform some calculations to verify that it doesn't appear to be a repeating decimal or other tests of number rationality, but at the end of the day you have to assign the semantic meaning to its outputs and make the actual determination that Pi is irrational.
"No - in fact computers are not capable of reasoning of any kind."
If superhuman poker software, superhuman chess engines, incredibly strong image classification and labeling (including, for example, superhuman radiology diagnostics), machine translation, text and code generation, automated mathematics software like Metamath and Wolfram Alpha, etc. are not "reasoning" according to your definition of that word, you simply have the wrong definition. I invite you to consider how a computing procedure might genuinely said to be "reasoning" on its inputs: it doesn't matter that someone coded the reasoning procedure (and for ML systems, this isn't even really the case), nor does it matter that the outputs deterministically came from the inputs.
"The information, while being processed by the computer, has no intrinsic meaning to the machine itself and is merely making its way from point A to point B"
So what? Why does the computer need to be aware of what its behavior means to us in order for the information processing to amount to genuine reasoning? Why does "understanding meaning" in the very anthropomorphic way that most people (including Searle) use that phrase actually matter? An AND Gate implements exactly the same sort of thing we do when we think about logical conjunction; no part of using "and" as a step in human reasoning actually requires that the step be taken with any sort of self-awareness. "True reasoning" and what you might call "mere information processing" are one and the same.
"The reason a computer can identify a picture of a cat is because we have fed it thousands of images and a human mind has already specified which of those images are images of cats."
If supervised ML isn't kosher to you, take a look at how well unsupervised ML can find semantic meaning and implement analogy-based reasoning in systems like, say, word2vec. But supervised ML is more impressive than you make it sound: we create an architecture with millions or trillions of parameters, feed data, and the system *reverse-engineers the very reasoning procedure* that recognizes cats itself. If all a cat-classifying algorithm could do were compare images to saved training examples, you'd be right in calling it unimpressive. But modern supervised ML is impressive because our models apply well to genuinely new input. It gets a robust bearing for "catlike features" and often spots details that humans don't. If this doesn't look something like "reasoning" to you, you again have the wrong definition.
A lot of the giant language models pretrained on human text have a surprising capacity to draw multi-step connections, generalize, summarize, and problem-solve. Keep in mind: this is with little-to-no explicit labeling. And the field is super new--expect models in 5 years to do even more stuff, even more reliably. How much more "reasoning"-y could a system be before people swallow their pride and agree that human reasoning is not so different?
Eh, this clearly is not what she meant, she could've worded it better but obviously no honest interpretation of what she said would lead you to conclude she thought the computers themselves were assigning the meaning to those abstract symbols. Rather, the point is that computers can deal with abstractions, so the fact humans can deal with abstractions is not unique or interesting. Yes, humans assign meaning to computers, but this is a separate point than the point she was trying to make. Modern day computers are not generally intelligent, so they cannot operate with independence, they cannot dress themselves, feed themselves, apply for jobs, drive to that job in a car, perform that job, pay their taxes, clean their rooms, etc. While a computer might be able to do 1 of these tasks, no modern AI can do all of them, so they cannot operate as an independent agent in society and require humans to constantly tend to them and tell them everything to do and cannot do things for their own sake. But this is just a factor of modern AI not being generally intelligent. The fact modern computers lack general intelligence, even with the most sophisticated AI, is true, but was clearly not the point she was getting at and is a separate topic.
Love it! I'll point out that even digital computers are analog at the level below the digits. That is, voltage in the circuits is analog, and we just arrange to apply a lot or a little, in order to make it easier to distinguish with other circuits, but every circuit has a range of voltages over which it's "zero" and a range over which it is "one". Things like FLASH memory might have as much as eight different levels in one cell, storing 3 bits as one of 8 voltages.
A lot of the things like object permanence do have to be learned, though. That's why there's a peek-a-boo game. Baby brains aren't really fully understanding basic physics until about 3 years old.
Also, people can't prove their own Godel string. Think about actually doing that. You'd have to hold in your head several times over the entire formal description of your head, and manipulate it without making any mistakes. You might argue that someone could *theoretically* do that, but then you're not talking about the humans that are actually walking around doing computations today, but talking about people with much better brains than we actually have, which would entail a different godel string.
Also, love the dress and key, which I didn't notice until about the third reference. 🙂
But the same is true about neurons; the final signal is essentially digital. The neurotransmitter is an essential repeater just to see that the signal does stay within the limits of 0 and 1.
How do we add? Adding integers is easy because we too use digital rules (basic counting). When we want to add more messy numbers, we slow down because of the limitations of the algorithm.
@@janami-dharmam Correct. If you look at the highest level down to the lowest level, most things flop back and forth between discreet and continuous a half dozen times. It's just a question of what level of detail you're looking at and how you're interpreting it.
Don't agree with that argument about the impossibility of a person calculating their own Gödel statement. It would require that you knew the "axioms" underlying your brain - probably the same for all brains - then following a standard procedure to calculate the appropriate self-referential statement, and because of Godel's theorem you've no idea if it is true or not. But because you've studied second order logic, you know it is true. This contradiction proves that your brain wasn't based on those axioms; but computers are based on a set of simple rules (axioms) so brains aren't computers. Thats basically Penrose's argument.
"A lot of the things like object permanence do have to be learned, though. " the fact that it develops at a certain consistent point is evidence that it is not learned. Another example of something that isn't learned is contact action. Given an apparent causal relation, babies will assume a hidden contact force.
@@MassDefibrillator I'm not sure it would be learned if it wasn't actually shown to the babies. And I can't imagine any even vaguely moral way of testing that. But you have a fair point.
Gödel's incompleteness theorem boils to be the following:
Let S be a consistent and a complex enough theory without being too complex. Then S can simulate itself using propositions relating to numbers. Let's call the simulation S'. Gödel's incompleteness says, that there is a proposition P such that there doesn't exists a proof from S to P and there doesn't exists a proof from S to (not P).
The simulation S' exists in the sense that for every proposition Q permitted by S, we can formulate a number theoretic proposition denoted Q' that is interpreted to mean that the simulation S' proves' the simulated version of Q. So we have a way to turn a proposition Q to a proposition Q' which is a proposition about the the Gödel number of Q. This procedure has the following 2 properties:
1) If S proves Q, then S proves Q'. (read If S proves Q, then S proves that the simulation S' proves' the simulation of Q).
2) If S proves (not Q), then S proves (not Q').
A proposition P is created using the simulation S'. The proposition P has this funny property that P is equivalent with the statement (not P'). S cannot prove P, because otherwise S would prove P' by property 1) and (not P') by the equivalent formulation of P and this cannot be by the consistency of S. Hence S does not prove P. Assume S proves (not P), which is equivalent with P'. Thus S proves P'. Since S proves (not P), by the property 2), S proves (not P') and by the consistency of S, this cannot be. So S does not prove (not P). Hence S is an incomplete theory.
You brought up that a computer has proven Gödel's incompleteness. Penrose's whole point is that the computer has to use some kind of a foundation to make deductions/calculations. The computer itself is still some kind of a system. Even though with a suitable foundation the computer can prove the incompleteness theorem, the incompleteness theorem still applies to the computer since its calculations are governed by some system S. Hence by Gödel's theorem there will be statements that the computer cannot prove one way or the other and we humans can create such a proposition. In this sense human can transcend a formal system.
This leads to a notion that the universe as whole cannot be reducible to a purely computational system. Penrose's understanding is that the only place non-computability is introduced to quantum physics is when there's the collapse of the wave function. Hence if quantum physics is the system explaining consciousness and consciousness has this property of transcending computationability, then the non-computational aspect of the quantum physics have to be a part in the explanation of conscious experience.
Isabelle! I never thought I'd see that in a popular science video. It takes me back to the 90s and a former life in Functional Programming and Theorem Proving.
2:11 🧠
5:08 Analog or Digital
6:55 How does a computer do what it does? How does a brain do what it does?
Brain - Adapts
Computer - Specific Purpose
*Making computers similar to brains*
8:50 Neuralmorphic Computers
10:53 Memcomputers
12:45 Energy
13:52 Self-Repair
15:13 Parallel Processing
16:10 Infinite Time For Pi, record 62.8 Million Digits
17:39 Human Thought cannot be alogorithimed
18:58
19:28 Brain Vs Computer
20:15 Social Learning
A computer is -as I understand it a form of adding or calculating machine, socalled because it performs to duties of what an original human computer did- perform routine calculations only rather more quickly that men(human beings) although seemingly there is an Indian being of the passive sex or woman that can do square roots in her head faster than a machine computer or calculator, but how, god knows; she says she just 'sees ' the answer; do you know 'how' you see?
How* do you do *knowing*, and What* is understanding? - a computer can neither 'know' nor understand, but if you could work out how a computer works that would tell you all you need to know about the chap that made it or rather designed it, which brings one within an inch of understanding the loon that invented the mister god fantasy arrived at the queer idea, given that men (human beings appear to be little more than organic computers with organic software an hard ware, but the idea that mister Evolution designed and programmed men is as daft as the mister god fantasy-but the two ideas are broadly similar barring the followers of the religion scientism often comes within a whisker anthropomorphism and the mister evolution fantasy and the child Dawkins get almost tearful if anyone suggests that they doo not buy the mister evolution fantasy, but what's to choose between mister evolution and mister god? - They as broad as they are long.
O one view computers tell as much about their designer as men do their designer.
The fact that they are obviously designed does not tell you a damn thing about the nature and identity of the designer- similiter a computer or computers, although I don't know, if the computer tells you a good deal about its designer if not the identity thereof, why not apply the same to men?
I suppose one might wangle an apology for mister god out of the same general reasoning, but why "Mister" god, rather than god a thing? - which is far more rational and has the advantage of avoiding girly anthropomorphism, which really is very queer indeed.
Thank you for another lovely video. I appreciate how you have fun while being thought provoking.
Great video, as always!
One thing where you made a mistake though is saying that deep down "on physical level" computers are digital. Exactly the opposite is true: all electronic devices are analogue under all the discrete abstractions. And as components get smaller and more complicated, we keep tripping on new ways physical nature messes up with our shoehorning it into discrete world. Translating continuous values of voltage into discrete-value and discrete-time logic is a pretty complicated matter, and it inevitably causes loss of potential efficiency in exchange for reproducibility and predictability.
There is a famous case when an AI researcher used machine learning to train an FPGA to detect certain sound inputs (FPGA is basically a bunch of "binary" logic gates that you can program and connect as desired). The resulting network was significantly smaller and more efficient than anything a human engineer could make, it operated in continuos time (no clock) and utilised full range of voltages. But the downside were, no one had any idea how it worked (even though it was just a couple dozen gates), the same network won't work if you remove one of the gates that was unconnected to anything (it was still affected and affecting EM field, after all), and it wouldn't work on any other FPGA.
Computability is dual to non computability -- the duality of Godel's theorem!
Analog is dual to digital -- binary or on is dual to off.
Problem, reaction, solution -- the Hegelian dialectic.
Solving problems is based upon the Hegelian dialectic.
Thesis is dual to anti-thesis creates the converging thesis or synthesis -- the time independent Hegelian dialectic.
Action is dual to reaction -- Sir Isaac Newton (the duality of force), forces are dual.
Concepts are dual to percepts -- the mind duality of Immanuel Kant.
Analytic (Mathematics) is dual to synthetic (Physics) -- Immanuel Kant.
The brain is a duality machine!
Duality creates reality.
"Always two there are" -- Yoda.
The left hemisphere is dual to the right hemisphere -- brain dominance, watch the spinning dancer:-
www.medicaldaily.com/right-your-eyes-science-behind-famous-spinning-dancer-optical-illusion-336122
Master (Lordship) is dual to slave (bondsman) -- the Hegelian dialectic.
Absolute truth is dual to relative truth -- Hume's fork.
Voltage is not "continuous" except in the macroscopic world. Time is not continuous but quantized with the smallest unit being the Planck interval. Everything is digital.
@@hyperduality2838 Almost none of that is correct, though it supports the Marxist view of 100 years ago. Yoda is fictional.
@@mikemondano3624 The British empire was built using the Hegelian dialectic so both sides in politics use it, left is dual to right. TV adverts use the Hegelian dialectic all the time, they create a problem say dirty teeth, there ia a reaction of disgust and horror and then a solution is provided, a new better tooth paste. If you watch the spinning dancer video that I provifded you can see the Hegelian dialectic in action as you mind is forced to either choose clockwise or anti-clockwise rotation -- this is hemisphere dominance (master & slave).
Hegel is therefore correct and you have duality.
"Philosophy is dead" -- Stephen Hawking.
If Hegel is correct then Stephen Hawking is incorrect. Ignore Hegel at your peril.
The Necker cube is a good visual example of duality.
Forwards is dual to backwards, which way is the train moving?:-
th-cam.com/video/kmO_OaYWFD8/w-d-xo.html
Attraction is dual to repulsion, push is dual to pull -- forces are dual.
If forces are dual then energy must be dual.
Energy = force * distance -- simple physics.
Energy is dual to mass -- Einstein.
Dark energy is dual to dark matter.
Energy is duality, duality is energy -- the 5th law of thermodynamics.
Potential energy is dual to kinetic energy -- gravitational energy is dual.
Electro is dual to magnetic -- Maxwell's equations, electro-magnetic energy is dual.
Everything in physics is made from energy or duality -- your mind is creating duality right now, perceptions are being converted into concceptions (thinking).
"May the force (duality) be with you" -- Jedi teaching.
"The force (duality) is strong in this one" -- Jedi teaching.
Yoda is correct as he understands the metaphysics of Hegel.
@@mikemondano3624 Correct is dual to incrorrect, right is dual to wrong.
The one is defined in terms of the other.
Antinomy (duality) is two truths that contradict each other -- Immanuel Kant.
You're always amazing with your simplified explanation of science!
and there is a reason for that
---------------------++++++++++++ sabine tongue scope ++++++++++++++++++++++++--------
---------------Science ---- non questions ----pseudo science --- mumble jumble------ plus utra
Yes even I understood some of it. Adding the video to my playlist for future reference.🌱
Your combination of factual information and capricious, humorous comments is brilliant. I really think that adding jokes to the mix, makes it easier to retain the 'sachliche' information you are sharing. Thank you for increasing our synaptic flow! :-)
This statement at the beginning… 😅 I had to literally stop the video after first second and calm down to overcome my uncontrolled laughter. Even Monty Python couldn't make me laugh that quickly.
Dr. Hossenfelder, I love your sense of humor. I discovered your channel very recently, and I must say I love the way you present your content, that little humorous touch just makes it a million times better 😌
It's much easier to understand her with her accent due to this sense of humour.
Great post Sabine. I love the way your mind works.... and your sense of humor. Big hugs 🤗
About 30 years ago, I started typing notes into my computer like work notes, social notes, how to notes, people notes, writing notes, movie notes, book notes… It recently dawned on me that everything stored on my USB memory drive was an extension of my mind and personality. All those things I wrote down over my adult life that meant something to me or still has meaning to me. Sort of a mind meld with a computer that holds an exact copy of meaningful memories and ideas. So glad I learned to do backups and backups to the backups as a retired computer programmer (and failed screenwriter).
Yeah, I have a drug problem too, stay strong we will make it to ai and then everyone will have the ability to live forever on a usb for a monthly fee, you can maybe make it as a museum director if you hang onto that usb until then?
I'm really impressed by how well you presented theory of what computers can and can't do. I was about to challenge your assertions about pi and all knowledge being digitzable, but you already knew! I'd add that Turing completeness is the level of computational ability that we're currently sure humans and computers can both do.
Computability is dual to non computability -- the duality of Godel's theorem!
Analog is dual to digital -- binary or on is dual to off.
Problem, reaction, solution -- the Hegelian dialectic.
Solving problems is based upon the Hegelian dialectic.
Thesis is dual to anti-thesis creates the converging thesis or synthesis -- the time independent Hegelian dialectic.
Action is dual to reaction -- Sir Isaac Newton (the duality of force), forces are dual.
Concepts are dual to percepts -- the mind duality of Immanuel Kant.
Analytic (Mathematics) is dual to synthetic (Physics) -- Immanuel Kant.
The brain is a duality machine!
Duality creates reality.
"Always two there are" -- Yoda.
The left hemisphere is dual to the right hemisphere -- brain dominance, watch the spinning dancer:-
www.medicaldaily.com/right-your-eyes-science-behind-famous-spinning-dancer-optical-illusion-336122
Master (Lordship) is dual to slave (bondsman) -- the Hegelian dialectic.
Absolute truth is dual to relative truth -- Hume's fork.
Self evidently computers can do what they are designed or programmed to do, which is perform calculations, which not all human beings can do and certainly the writer can't do, but you can't expect computer to understand anything that they are not program to understand they do not have ideas or concepts and cannot hold images in the heads - having no heads, and thus cannot do imagination which for them is an advantage , but for human beings a weakness, but you might equally say that human beings can only do what they are programmed to do; both are species of machine, and it goes without saying that machines cannot choose when it comes to the reactions of their functions and that is equally true of computers end men (human beings)To describe men (human beings) as "conscious" is laughable; occasionally capable of a species or degree of consciousness but conscious, never.They cannot possibly have any experience of consciousness because they are dreamers; dreamers cannot experience consciousness.
Nice distillation of an enormous topic. A thorough treatment of the subject would take days.
This one had some funny lines in it! Love studying complexity and information; always wondered what, if any, extra properties the brain has.
OMG Sabine! Thank you for keeping up with the jokes! They're hilarious for us nerdy types! 🙂
You're awesome.
I would prefer less jokes and being true to the topic.
@@mimo9906 Ther rest of the world could use some levity in their lives, not just someone rattling off facts...
@@mimo9906 you sound great fun
Congrats for this excellent and very instructive video. I really enjoyed your definition of a "gallon" of water ("lots of water"). I was also surprised to learn that physicists would take pi as equal to 1, for I thougt it would be closer to 10 on a logarithmic scale, but I was mistaken (pi is less than Sqrt[10] = 3.1622...).
not literally 1 but "unit". same could be said with velocity of light as a "unit".
Though I suspect the reasoning is more like "ignore the constant factors".
-(( π ))- n ← (π * n for all n)
rather than
π ≈ 1 ≠ 3.14159...
pi = 1? I was also a bit shocked at yet another thing physicists do that is weird. Like the sum of all positive integer being -1/12. I guess, as Sabine would point out, it must be true if the calculations come out agreeing with the observations.
My thought of how to define a gallon of water for those unsure was "about eight pints of beer".
Sabine, you are one of the most unique voices on TH-cam - love your deadpan delivery.
I agree with Roger Penrose about the fact that the human brain use quantum mechanics to operate in detail.
I'm convinced of this because all creative functions require knowledge of the short-term future on many small but meaningful details; computation alone is incapable of any truly creative action. And I believe, although I can't prove it, that both self-conscience and creativity require knowledge of short-term future on a battery of elements - matrixes, if you like. If so, sentient computers are out of reach of Mankind until we fix our knowledge of the quantum world; as we are currently limited by a Science that is only merely an operational "description" of our world.
Thank you for the great video.
Regards,
Anthony
I dunno. Things like Dalle-1 & 2 etc that are able to emulate specific art styles of brilliant painters and combine all different elements together, to me is an example of creative function.
You are just speaking in voodoo magic word salad and don't know the definitions of the words you are even using. Quantum mechanics does not have anything to do with human ability to know the future in the short term. The whole point of a neural network is to be able to build a statistical model so that it can predict what inputs would produce some outputs. This is true for all neural networks, whether biological or artificial. Humans are more complex than anything we have created in software, but even though we are more complex, the principle is the same. We can predict the future because we've observed the past enough to form models capable of doing so. Predicting the future does not require some sort of magic voodoo, it just requires observing the past and forming models based upon it.
Always such a pleasure to listen to her speak on the subject. And beautiful too!
Very good and careful explanations. Personally I tend to agree with Daniel Dennett on consciousness, and so I am not forced to try to "introduce" uncertainty into the brain.
Seeing the light: Researchers develop new AI system using light to learn associatively
Source: University of Oxford
Summary: Researchers have developed an on-chip optical processor capable of detecting similarities in datasets up to 1,000 times faster than conventional machine learning algorithms running on electronic processors.
The new research published in Optica took its inspiration from Nobel Prize laureate Ivan Pavlov's discovery of classical conditioning. In his experiments, Pavlov found that by providing another stimulus during feeding, such as the sound of a bell or metronome, his dogs began to link the two experiences and would salivate at the sound alone. The repeated associations of two unrelated events paired together could produce a learned response -- a conditional reflex.
Co-first author Dr James Tan You Sian, who did this work as part of his DPhil in the Department of Materials, University of Oxford said: 'Pavlovian associative learning is regarded as a basic form of learning that shapes the behaviour of humans and animals -- but adoption in AI systems is largely unheard of. Our research on Pavlovian learning in tandem with optical parallel processing demonstrates the exciting potential for a variety of AI tasks.'
Journal Reference:
1. James Y. S. Tan, Zengguang Cheng, Johannes Feldmann, Xuan Li, Nathan Youngblood, Utku E. Ali, C. David Wright, Wolfram H. P. Pernice, Harish Bhaskaran. Monadic Pavlovian associative learning in a backpropagation-free photonic network. Optica, 2022; 9 (7): 792 DOI: 10.1364/OPTICA.455864
Sabine Hossenfelder, Science needs reason to be trusted, Nature Physics, Volume 13 (2017), pp 316-317, doi:10.1038/nphys4079
As far as I can tell Dennett has said absolutely nothing about how phenomenal consciousness allegedly emerges from (or "is identical to") the brain--which is to say why neural activity should be accompanied by a first-person perspective. So there's not much to agree with, really. (Yes, I read _Consciousness Explained_ and found it comically mistitled.)
@@Khomyakov.Vladimir cool, now explain the relevance of the paper or article, instead of attempting to get by with what you think the title implies.
fantastic presentation, thank you
the efficiency thing really got driven home to me when the warehouse i work at switched to laser guided robots for most of the work. they have lasers all over them a GPS unit ect ect and they are quite slow, and cant deal with anything unexpected. throw a stick in front of it and it just stops and sounds a alarm. it cant go around it. Me on the other hand, in a very basic sense do the same job with 2 cameras and 2 microphones, and deal with almost anything. All powered by 2 cheese burgers and a few cups of coffee.
"Damned Boston Dynamics robot is acting up again. Guess I'll just do it myself with ease once I finish these doritos"
Robots and lasers? Sounds like a scary combo for those who have watched old SF movies. 😂
@@Storin_of_Kel 😆
Oooo Mr. Fancy has robots that don't bump into him. So.... *NOT* Amazon?
Theyre buying into the Shiny Object Syndrome
I think you gave a great overview, but this is a topic that could be explored for hours. Let me use one simple example.
There are many reasons that we build digital logic circuits the way we do. One of the biggest is scalability. The amount of circuitry that we put into one of these devices is massive. And yet, we test and prove each unit correct in testing. We want each and every digital integrated circuit to work within specifications that are compatible with all the others of its type. An 8051 is an 8051 (and as a joke I used to tell young engineers if you need more than an 8051, 22V8 and a 555 to make something it was probably too complex).
The same is not true with people. We expect them to work differently from one another. We have this emergent phenomenon called consciousness (assuming it is emergent). As far as I can tell, we greatly struggle to explain emergent things in our universe. Why does a pile of the right chemicals not self organize into life? Why does it?
I think this notion of computation and thinking as the same thing is probably incorrect. Where did Einstein (yeah that guy) get his (maybe) original notions from? You can't analyze yourself into a vision. You can vision things and then analyze them. All of this leads me to believe the answers are way more complicated than we think they are.
Excellent observations.
@NewTube Channel Okay. Since it is simple. Which language is most easily stored in the human brain? How do you measure it?
@NewTube Channel I agree that human memory and processing are not like a digital computer. I have seen no science (not belief) about the whole body storage idea. It would imply that amputees or quadriplegics would lose significant memory, processing, or language skills. This is not obvious to me.
What’s most amazing about neurons is that they probably evolved from cells that were just really good at communicating with other cells. All of our thoughts exist because of a strange adaptation that came along somewhere between sponges and animals with bilateral symmetry. A completely random event perhaps, who can say for sure.
That’s the biggest difference between something we build and something that just occurs naturally. When humans want to build a circuit we have a goal in mind and each part has a purpose (even if it’s not that well made), but in nature there is no goal. Things that live will continue to live by passing on their traits, and if they accumulate errors in their genetic coding perhaps that increases or decreases the odds of that. Animals and all other life on earth have arrived at their current forms by chance, not even the goal of survival is consciously considered in what forms emerge but rather plays a role passively. The way a nervous system is laid out is so alien to how we would design something that it would probably be easier to understand something built by aliens. Why? Because at least it too was designed.
After having brain surgery for an aneurysm, I swore that someone had installed WinXP into my head because it kept needed to be updated.
Great analysis, thank you. Might I add another suggestion that brains are give purpose by emotions and that emotions are driven by needs and dopamine rewards. In other words we are programmed by our physical and social environment while computers are given purpose by us. A person without emotions ( such an anomaly has happened) cannot survive without guidance.
Are not what you call emotions merely functions that react mechanically automatically without any participation of any part of your common presence that is not only that function?
Computability is dual to non computability -- the duality of Godel's theorem!
Analog is dual to digital -- binary or on is dual to off.
Problem, reaction, solution -- the Hegelian dialectic.
Solving problems is based upon the Hegelian dialectic.
Thesis is dual to anti-thesis creates the converging thesis or synthesis -- the time independent Hegelian dialectic.
Action is dual to reaction -- Sir Isaac Newton (the duality of force), forces are dual.
Concepts are dual to percepts -- the mind duality of Immanuel Kant.
Analytic (Mathematics) is dual to synthetic (Physics) -- Immanuel Kant.
The brain is a duality machine!
Duality creates reality.
"Always two there are" -- Yoda.
The left hemisphere is dual to the right hemisphere -- brain dominance, watch the spinning dancer:-
www.medicaldaily.com/right-your-eyes-science-behind-famous-spinning-dancer-optical-illusion-336122
Master (Lordship) is dual to slave (bondsman) -- the Hegelian dialectic.
Absolute truth is dual to relative truth -- Hume's fork.
@@vhawk1951kl
Peter, emotions can be thought of as a calculation, but it is based on complex factors such as our needs, social context, current mood and baked in survival biases we inherit from evolution. What makes this different from a computer is that it is too complex to get a consistent result. This is analogous to the weather. In principle and as a statistical large picture, we can predict the weather a few days out, but not on a micro climate level.
Your videos are always informative and the straight-faced jokes are part of what makes them highly watchable and enjoyable. Thanks so much for the time and effort you put into making them.
Agree! I still wonder how she manages to keep a completely straight face, considering how fun some of those jokes are.
@@LeanAndMean44 Well, she IS German. lol
Sabine, this is a great video. It makes me wonder, is there an upper limit on how much a human can learn? It seems like every generation builds so much new technology on previous generations' work, but it often requires knowing everything that came before. So I'm curious if we'll ever hit some limit.
Unless the size of your skull is infinite, then the amount of information you can store in it is definitionally finite. While information may be intangible, storing that information and processing it definitely takes physical resources, and physical resources always, always have limitations.
Arguably, physics sets an upper limit to the amount of processing power. A computer the size of a planet is affected by how fast you can move information, which is restricted to the speed of light. A computer the size of the universe would take many billions of years to access and process information.
I suspect that every person who watches Sabine’s videos knows there is a finite limit somewhere. But I’m curious if we will hit some other limit first, or maybe the brain doesn’t even work that way. Maybe, as we learn things, we automatically discard things too. I don’t know, but I’m curious what that limit might be if it actually exists in any practical terms.
@@ChiltonWebb I'm pretty sure we do replace things unless there's some deep archive in my brain that I don't have the keys to
Computability is dual to non computability -- the duality of Godel's theorem!
Analog is dual to digital -- binary or on is dual to off.
Problem, reaction, solution -- the Hegelian dialectic.
Solving problems is based upon the Hegelian dialectic.
Thesis is dual to anti-thesis creates the converging thesis or synthesis -- the time independent Hegelian dialectic.
Action is dual to reaction -- Sir Isaac Newton (the duality of force), forces are dual.
Concepts are dual to percepts -- the mind duality of Immanuel Kant.
Analytic (Mathematics) is dual to synthetic (Physics) -- Immanuel Kant.
The brain is a duality machine!
Duality creates reality.
"Always two there are" -- Yoda.
The left hemisphere is dual to the right hemisphere -- brain dominance, watch the spinning dancer:-
www.medicaldaily.com/right-your-eyes-science-behind-famous-spinning-dancer-optical-illusion-336122
Master (Lordship) is dual to slave (bondsman) -- the Hegelian dialectic.
Absolute truth is dual to relative truth -- Hume's fork.
Because "I" am always bigger than everything you can demonstrate on to "me", there is no end in learning! Even the beginning is always ahead! 😉
I really enjoyed this one Sabine, thank you. I make computer games, and trying to make a computer program do unexpected positive things is a very interesting area of study.
I wonder if the first indications that our computers are approaching human capability will be the sudden appearance of backwards baseball caps and refusals to perform....
I wonder what a "computer tantrum" will be like?
@@John.0z They take many forms. The minor tantrums require a reboot, but major ones require returning the computer to factory defaults.
@@John.0z basically the whole system acting as a printer?
@@user-sl6gn1ss8p With some of the printers I have known, that would be about right.
Computability is dual to non computability -- the duality of Godel's theorem!
Analog is dual to digital -- binary or on is dual to off.
Problem, reaction, solution -- the Hegelian dialectic.
Solving problems is based upon the Hegelian dialectic.
Thesis is dual to anti-thesis creates the converging thesis or synthesis -- the time independent Hegelian dialectic.
Action is dual to reaction -- Sir Isaac Newton (the duality of force), forces are dual.
Concepts are dual to percepts -- the mind duality of Immanuel Kant.
Analytic (Mathematics) is dual to synthetic (Physics) -- Immanuel Kant.
The brain is a duality machine!
Duality creates reality.
"Always two there are" -- Yoda.
The left hemisphere is dual to the right hemisphere -- brain dominance, watch the spinning dancer:-
www.medicaldaily.com/right-your-eyes-science-behind-famous-spinning-dancer-optical-illusion-336122
Master (Lordship) is dual to slave (bondsman) -- the Hegelian dialectic.
Absolute truth is dual to relative truth -- Hume's fork.
Thank you for a both informative video and the addition of dry humor. Gotta love the reference of innate fears of spiders, snakes and circus clowns 😅🙃
A Computer funnels a huge amount of memory through a small static matrix of deterministic decisions (processor). A Brain is a dynamic and huge self modifying matrix combining memory and computation. What we consider AI is a relatively small previously trained probability matrix doing deterministic decisions we cannot explain from looking at it.
Neural network research is just one branch of AI.
Computability is dual to non computability -- the duality of Godel's theorem!
Analog is dual to digital -- binary or on is dual to off.
Problem, reaction, solution -- the Hegelian dialectic.
Solving problems is based upon the Hegelian dialectic.
Thesis is dual to anti-thesis creates the converging thesis or synthesis -- the time independent Hegelian dialectic.
Action is dual to reaction -- Sir Isaac Newton (the duality of force), forces are dual.
Concepts are dual to percepts -- the mind duality of Immanuel Kant.
Analytic (Mathematics) is dual to synthetic (Physics) -- Immanuel Kant.
The brain is a duality machine!
Duality creates reality.
"Always two there are" -- Yoda.
The left hemisphere is dual to the right hemisphere -- brain dominance, watch the spinning dancer:-
www.medicaldaily.com/right-your-eyes-science-behind-famous-spinning-dancer-optical-illusion-336122
Master (Lordship) is dual to slave (bondsman) -- the Hegelian dialectic.
Absolute truth is dual to relative truth -- Hume's fork.
Brahman (thesis, creator God) is dual to Shiva (anti-thesis, destroyer God) synthesizes Vishnu (preserver God) -- Hegel, Hinduism.
I have always thought of computers as translating machines. Humans figured out how to build machines that could translate problems into data and instruction that manipulated that data into something that makes sense to the human at the other end of the process. All of that design came from our ideas and needs. The brain is the product of evolution and is, as a result ,capable of things we struggle to understand while it is still not able to help you remember why you are in the kitchen. It's beer, wine and/or cheese... you are welcome. IBM '68-'99 Love your videos.
That's a pretty interesting approach.
However, it's still true that the base function of computer is to add 1s and 0s - so it is a calculating machine very fundamentally.
What computers evolved from are devices designed to stepwise automate physical tasks. Things like mechanical systems to automate weaving fabric. However complicated they get in terms of abstractions and appearance, at the core is still a sequential stream of instructions to perform.
The computational model that we arrived at is extremely useful for many things, but not so much for simulating a biological brain.
@@Jim-Tuner I think that's machines. Mechanical machines and later electrically running machines were built to replace manual work. Computers were in the beginning really meant for shortening the time of tedious math operations.
Software was added late in the game to further specify mechanical operations. That's sort of the marriage of machines and computers.
@@susanne5803 People say that, but there is more to the story than that. Take a look at the 1803 jacquard loom and its control cards. The mechanism of the loom was the very influential in the work of the Babbage on his difference engine. the loom was a very basic computer with a sequential instruction set which a generated a mechanical result.
In my opinion, its not possible to seperate the history of automated mechanical operations from the history of computing. There are people who differ and who consider "computers" strictly electronic devices created quite late for math calculations. But those electronic devices were in fact using mechanical structures in their design from far earlier.
@@Jim-Tuner Basil Bouchon used something akin to punch cards already 1725 for looms. He took the punch stripes from organs. Pinned drums were also used for barrel organs and music boxes. I still think of computers and machines as different. In a way computers are a specific type of machine helping with math operations.
For everyone in the comment section saying 'humans aren't computers', the Turing machine WAS LITERALLY MODELLED ON WHAT HUMANS CAN DO, like actually look at Turing original paper.
Excellent video as always 👏🏻💐
Great episode, I think once we crack the “feeling” problem , I.e once we work out the brain represents feelings, we can emulate human brain in computers.
I wonder if you understand that when you use the word "we" it indicates, points to, or means, the user of the term - that is*you*sunshine and his immediate interlocutor, and since you have no immediate interlocutor, he can only be referring to yourself or say "I" when you employ the term or word "we".
Self evidently you can experience nothing of the experiences of others insofar as they are their own experiences and thus "we" is and can only possibly be, imaginary or fanciful To what exactly are you referring when you speak of the "squealing problem", and why is it "a problem"?
What exactly do you seek to convey by feeling? - Sensation or some sort of mechanical reaction in the emotional function?
Presumably you mean once we figure out how to make a computer have a first-person perspective and thus capable of experiencing a feeling or anything else. That is, indeed, something of a toughie.
@@halfacanuck It might help the both of you if the both of you clearly understood that the term "we" indicates the user of the term and his immediate interlocutor and since you are short of immediate interlocutors to the tune of any at all, you are referring to yourselves so were better suited to say "I" rather than the imaginary "we"
@@vhawk1951kl You surely understand that English speakers use "we" idiomatically in this kind of discussion to mean "humankind".
@@vhawk1951kl The first definition of "we" in Merriam-Webster's dictionary is "I and the rest of a group that includes me". Thus linkin543210 and I (henceforth "we") are using it correctly, because we mean humankind, to which group we both (presumably) belong. Your sense of "we" as in "you and I" is the second definition which in this case is not the correct one. Your grammatical policing is in this case unwarranted.
Brilliant video, Sabine! Thanks! 😃
Two things I was thinking about... The first are false memories. I never saw something like that happen to a computer!
The other is that I lost 75% of my field of view, because of a surgery I did when I was 13 years old. I don't see from my left eye, neither from the right half of my right eye. (I had a tumor between the two most important glands of the human body, I don't remember their names in English right now.)
Either way, it's weird... Because I remember effects of 3d vision, like those tricks you do with your fingers... But I don't remember any other difference. And I didn't noticed from the beginning, I just realized that I wasn't seeing with my left eye when I blinked (still in the ICU). Really weird!
Anyway, stay safe there with your family! 🖖😊
There is something called bit flip, a bit flip can be caused by cosmic radiation for example, switching a bit that was 0 to 1 or the other way round. That's similar to false memory I believe.
@@IngTomT Interesting. I didn't know about that!
@@IngTomT Computers can definitely have random memory errors, but that isn't really the same as false memories. What happens in your brain is that your brain doesn't actually store memories with a lot of detail, instead you only remember the important bits and your brain tries to fill in the rest in a way that makes sense but isn't exactly the same as what really happened. And then the new modified version of the memory overrides the original, so you keep getting small changes that add up over time and can eventually give you a completely different memory than what actually happened. This also means that counterintuitively the more often you remember something the less accurate it probably is, even if it seems more vivid to you.
You don't really get the same kind of effect in a computer for two reasons. First, we do compress data but when we uncompress it it's a relatively deterministic process and we don't really try to fill in missing details the way the brain does. I could see this changing pretty soon though, since stuff like AI upscaling for images is getting pretty good. More importantly though, since the data and the processing are done in different parts of the computer we always have to copy the data over to the CPU before doing anything to it and the original version isn't affected. So you can get things like JPEG artifacts in images, but they don't build on each other over time like in the brain. You'd have to deliberately save the processed data back to the disk for that to happen, which usually isn't what we want to do.
It's not actually cosmic rays. It's because there were radioactive isotopes in the ceramic they used for the chip casing!
They are more careful about their materials now, and it's only a problem for electronics operating in hazardous zones.
the way you deliver your jokes and punchlines is so god damn good I wish I could give more than one Like thanks for all you do
Shaahin..... Catch yourself on
Action potentials are binary. A neuron either fires or it doesn’t.
I think it's hilarious how we can discount our own biology and pretend to know better than billions of years of evolution. Sure, I may forget why I'm in the kitchen sometimes, but I'm also maintaining perfect homeostasis and processing unfathomable amounts of sensory information while being both assaulted and fueled by these super nachos in front of me
39. There are 39 seasons of Dr. Who.
Haven't seen the video yet. Going to watch it now, but I'm guessing yeah it's probably a kind of biological computing device. Only it works very differently from our binary computers we use in everyday life, since not only are multiple neurons connected to multiple other neurons. The signals and its effects on each other also vary wildly. This makes it far superior to most modern day computing. I wonder how many years it will take before we have something artificial (either silicon, biological or something else) that can match our brains.
Uhm, but you do know that computers are outperforming us in many tasks already, like image recognition for instance? You’re right, our brains work differently, but we’re on a good way to be outperformed by computing devices in any given task.
@@thisismyalias yeah I know, but those are pretty much single task oriented neural networks which still use binary computing. The real wonders will start happening if we have some kind of synergy between them. If you were able to actually program a brain I'm sure it would vastly outperform anything out there today.
I just saw the video and the thing I'm most curious about is whether and how quantum computing has anything to do with us being conscious. I believe if we get far enough there will exist 'artificial' consciousness. We as humans always like to believe we are apart from nature and that our intelligence and, being conscious, is reserved for us humans. I believe this is due to the evolution of our big old ego and a bit of hubris. I believe consciousness is just a gradient shared between animals in the animal kingdom, including ourselves. I even believe that different people experience different kinds and values or a spectrum of consciousness. I mean just look at how easily our consciousness is altered not only by the smallest amount of substances, but even by emotion or time for example.
There is a big chance that we will quickly be the inferior intelligence if we were to develop an A.I. that is able to iterate on itself. I hope don't experience it, but looking at the biggest picture possible I'd be okay with humanity being a "boot-loader" for something beyond our capabilities.
@@DeltaNovum I agree with everything you said. Regarding the question if quantum physics might be involved; I strongly believe that to be the case to some degree, although I do not believe in true randomness but some variant of the hidden variables theorem (and if not local, then global).
And yes, let’s hope to not be around when AGI comes into existence and treats us the same way we treat everything we deem irrelevant or unworthy in our way forward.
@@thisismyalias indeed ^^. And I carry the same assumptions you do. You probably agree with Sabina about free will as well?
I assume our reality to be deterministic and that true free will is not really a thing. In my daily life I just follow my biology and live like I do have free will, but deep down I believe it not to be true.
I know many way more intelligent people than me don't agree. But as with many things in the science community I think fear of death and the ego are things that make us biased from time to time :).
@@DeltaNovum no but ai cant think on its own so we might never become inferior although it can also cant develop any physical laws or invent or discover anything regarding science and technology also our human brains are much more versatile as u have seen in the video and another reason is that an ai can be shut down by an emp attack but our brains survives it. u know that no computer can store information permanently becaz space radiation flicks 1s and 0s in storage device so pal i don't think ai can even surpass us in day to day reasoning and common sense and thus we will never become inferior, mark the point ai cant think on its own or invent anything but can make copies of its own which is not lethal, ai is also meant to help humans for specific purposes🤓 have a nice day dude
❤️👍❤️
I respect Penrose's work in cosmology, but he's talking out of his butt when it comes to consciousness.
Who isn't?
Why, because him and Hameroff figured there might be a way to store quantum States at room temperature, or the very likely possibility the brain is far more capable due to the microtubules being the basic processing nodes, instead of the higher level neuron "scientis" are focused on. Or the genious assumption that entanglement is very likely a basic function of our brain. Etc etc, Penrose is the closest intelligent scientist we have to Einstein. Sabine would wish she could stand in his shadow. I bet she doesn't even realise the another fundamental difference between computers and brains, brains don't really have a filesystem for looking up information, all information is passed to relevant regions, these regions correlate all incoming info and autonomously report if correlation exceeds certain threshold. That architecture is nowhere to be found in any current AI system, current AI is just at infant level, not even close to human brain capacity not now, not in a 100 years. Since these AI networks basically have no understanding of what they are "interpreting" it is a great danger to use them in safety critical systems, probably some high pricd will be paid, if people use this technology in critical cases. That is something Penrose is trying to tell us, and I have a lot more confidence in Penrose than some numbnut that owns this channel or her "clever" respondents.
I am disappointed this video focused on shallow technicalities rather than consciousness.
Love these broadcasts science delivered in an entertaining and challenging manner.
"you'll finally have something to talk about" @7:45
Sabine cracks me up
I have to say, I'm loving the addition of small humorous statements, etc to your videos!!! A really nice touch!
Im very glad you take the time to speak English, I've learned so much from your channel. I know its painful, its a secondary language for me too.
In the interest of preserving history: I was in college before hand-held electronic calculators were inexpensive enough for starving students to afford. So I used a slide rule in my engineering and physics classes. Sabine, that's pronounced "slide-rule" not slide-ruler. I'm 77 years old. Your videos help me exercise my brain. Thanks so much for all the work and time........Radio valve
I too remember the slide rule but it was very expensive. My first interaction with a calculator was a mechanical swedish calculator (I forget the make) which could do the basic arithmetic with lots of noise. The electronic calculator I have used during my Ph.D. days had nixie tubes and discrete transistors. I have not seen calculators that had valves but did see a russian computer that used valves.
Computability is dual to non computability -- the duality of Godel's theorem!
Analog is dual to digital -- binary or on is dual to off.
Problem, reaction, solution -- the Hegelian dialectic.
Solving problems is based upon the Hegelian dialectic.
Thesis is dual to anti-thesis creates the converging thesis or synthesis -- the time independent Hegelian dialectic.
Action is dual to reaction -- Sir Isaac Newton (the duality of force), forces are dual.
Concepts are dual to percepts -- the mind duality of Immanuel Kant.
Analytic (Mathematics) is dual to synthetic (Physics) -- Immanuel Kant.
The brain is a duality machine!
Duality creates reality.
"Always two there are" -- Yoda.
The left hemisphere is dual to the right hemisphere -- brain dominance, watch the spinning dancer:-
www.medicaldaily.com/right-your-eyes-science-behind-famous-spinning-dancer-optical-illusion-336122
Master (Lordship) is dual to slave (bondsman) -- the Hegelian dialectic.
Absolute truth is dual to relative truth -- Hume's fork.
Excellent. Thank you, Prof Sabine
I speculate it is possible to build individual neurons using programmable Uni-junction transistors without programming them. They are configured with a capacitor across the anode, and the gate, and powered in the micro-amp range with a reverse biased signal diode across the gate, and ground. A multi-meg-ohm resistor across B+, and the anode, and another capacitor across the anode, and ground. Each neuron is connected to other neurons via multi-meg-ohm resistors to their respective gates. Each neuron fires an electrical signal depending on the amount of energy that is stored in the gate capacitor that is electrostatically charged in proportion (synapses) to the voltage variation of output from sensors that are connected to the capacitor gate combination of each neuron . In other words, each neuron fires when a threshold is reached that is determined by previous firings of neurons, or many of them. Each neuron is inconsequential individually, but as a collective act as intelligent deterministic execution of process depending on the extent of sensor input. More sensor input, more intelligent process (decision making pathways that output to various machine functions, ie. motors, lights, sound, camera, etc., analogous to eyes, ears, touch, taste ) output from the multitude of neurons. More neurons, more sensory input, more intelligence. Its kind of a programmable resistor of sorts. I'm probably out of my mind, but I think it would work using millions of these hardware configured neurons. Memory is a function of the neuronal collective that functions in real time as a whole, never as discrete individual units of memory (one & zeros) that a microprocessor accesses for instance.
Congrats to (almost) 500k subscribers!!
OMG natural selection for turning off the lights when leaving the room. That's pure gold :D :D :D
There is nothing to argue about. From mathematical perspective: 1. Our brain is dynamical system. 2. Any dynamical system can be represented (not just approximated) by two layer recurrent neural network. 3. Neural networks can be arbitrary precise implemented on any Turing machine. 4. Classical computers implement Turing machine. QED The question is how slow it would be. It is a concern since our brains seems to implement Non-deterministic Turing Machine. Though, we are on verge on second computing revolution. Not quantum, it’s first. I am about DNA computers. They are implementable and implement non-deterministic Turing machine therefore brain will be possible to effectively simulate on it. So the question is not whether we are computers or not, we are, any dynamical system can be implemented on computer and we can at least manually do computers job, so we present equivalent models of computations. The question is how to effectively model brain. And we cannot do it effectively on Turing machine, only on Non-Deterministic Turing machine.
I really like to start my day with a cup of tea and a vidio or two like yours to put me in a frame of mind to face the days challenges.
what an excellent video and explanation, I truly feel like all the answers I was looking for were answered! you are fire
Thanks for all your videos!
I've never wanted to have pre marital relations more in my life. I love it when you talk nerdy to me..... the human brain is capable of multiple perceptions at a time. And if you're in is high enough you can practically see the future.
“Saving energy benefits survival, which is what I said to my kids when they leave the lights on “ 😂😂😂😂, brilliant.
Man I wish my grandma was a computer. That'd be just so cool.
Great video. The point Sabine makes about general purpose vs. specified purpose is a really good one. This is why I'm very skeptical that humans will develop generalized AI. There's really no market for one, and it's probably too complicated to do if we just decided we wanted to do it.
Lets make a petition to have Sabine be a temporal companion to the new Doctor
This animation is SO GOOD!
If you watch infants' and children's reacting to events in cartoons, you realise that they're based on quite sophisticated understanding of natural laws, especially physics, even if they could not verbalise them. A species that spent part of its evolution in trees had better understand gravity from an early age. Laughter is a combination of a scream at danger, and a reassuring coo, so it indicates the subject's reaction to an "impossible" situation.
Computability is dual to non computability -- the duality of Godel's theorem!
Analog is dual to digital -- binary or on is dual to off.
Problem, reaction, solution -- the Hegelian dialectic.
Solving problems is based upon the Hegelian dialectic.
Thesis is dual to anti-thesis creates the converging thesis or synthesis -- the time independent Hegelian dialectic.
Action is dual to reaction -- Sir Isaac Newton (the duality of force), forces are dual.
Concepts are dual to percepts -- the mind duality of Immanuel Kant.
Analytic (Mathematics) is dual to synthetic (Physics) -- Immanuel Kant.
The brain is a duality machine!
Duality creates reality.
"Always two there are" -- Yoda.
The left hemisphere is dual to the right hemisphere -- brain dominance, watch the spinning dancer:-
www.medicaldaily.com/right-your-eyes-science-behind-famous-spinning-dancer-optical-illusion-336122
Master (Lordship) is dual to slave (bondsman) -- the Hegelian dialectic.
Absolute truth is dual to relative truth -- Hume's fork.
Brahman (thesis, creator God) is dual to Shiva (anti-thesis, destroyer God) synthesizes Vishnu (preserver God) -- Hegel, Hinduism.
You're a remarkable woman. Thank you for your efforts.
Edit: Your English is better than most native Americans. English can be a confusing disaster at times.
I'll wager you're a great Mom.
Haha, Sabine casually spitting out death threats to her kids, when they forget to turn off the light, cracked me up.
Some of this seems to explain why I feel so confused much of the time, I think. Thanks for another informative vid -- and the humorous parts.
Thank you, you subjects are always interesting. At 77 going on 78, I need all the “interesting “ I can get!
I really like Sabine. She's so enjoyable to learn from and her sense of humor is AWESOME. Lol!
Hi Sabine, John from Australia here, your English is great, better than a lot of native English speakers that I have known. Great content, three thumbs up! I live near a nuclear reactor, hehehehe.
By analogy with human designed neural networks, brain memory is most likely to be embodied in the pattern of the strengths of interconnections between countless neurons, which may be as simple as how many axons (activating connections) are passing from one neuron to another. These strengths develop as a result of 'learning', by analogy training the network. Perhaps the more often one neuron is stimulated to stimulate another, the stronger the interconnection becomes. Inputs take the form of the stimulation of a selection of targeted neurons, these then fire off to stimulate other neurons they are connected to. The pattern of weights or strengths of these interconnections activate interconnected feedback loops of neurons within the network. The result is a stimulus of a small number of neurons will tip the neural network into a stable state in which a particular pattern of neurons is active. This would form a response or a memory. Underneath it is the pattern of connection strengths which embody memory or response. This is in turn learned by training the network.
Stumbled across your channel, and love your explanations, and the humour
My brain is often a source to frustration. It forgets important information, and it can't shut down pain signals from health issues, that will never get better. It's age related. In other words, it's falling apart. Just like the rest of the body. And unlike a computer, the worn out drive can't be cloned to a fresh drive. Don't get old and sick. It's not worth it. Superb video.
Your English is better than most native English speakers, by the way.
Oh boy, my favourite topic! Stuart Hameroff and Penrose. Bring it on!
Great video!! Thanks for making it!
Hello another great Video, watched in full, but one question remained: what about emotion?
YOU'RE ABSOLUTELY RIGHT GENIUS PROFESSOR.
The faculty of learning is an important aspect of what the organic brain does, and remains an important frontier for computers. I expect that in this century, computers will begin to surpass humans in the ability to learn autonomously and autodidactically.
Your English, including your accent, is marvelous.
Thanks!
On that last note of thought, I learn greatly from TH-cam, objective (as objective as any human can be) creators like you, and even AI like the new Chatbot ChatGPT. I don’t need to worry about personal agendas, how much time the person has, and I don’t need to listen to the condescending and intimidating tone which some people express now and then.
... a machine that's very good at adapting to new situations with new problems! Gotta a love a smart and well-communicating person with that kind of humor.
OH THANK YOU so much for making this video.