Same. When they talked about machines being unable to understand the subtleties of human interactions, I immediatly thought of my own way of thinking and how it more resembles the machine.
I took a linguistics class with a professor who'd worked on very early language software, trying to teach computers how grammar works in the 70s. They thought it would be the simple, given how mathematically language tends to be, but when they started working with them they discovered that the social and idiopathic nature of language made it virtually impossible.
@@rmt3589 to a degree. Ai has come a long way in being able to interpret this information. It's not perfect yet. And it cannot be perfect because if it were perfect it would be wrong, ironically. There is no English grammar. Every human being uses language in their own way, according to their own set of rules, for their own purposes. An ai would need to consult with every Englishspeaker, past and present, to come to a complete proficiency in English. And even if they could, such a machine would be useless because that's not how humans learn language and any process that learned language like that would not lead to outcomes that result from the way humans learn language.
@@nickscurvy8635 I think the understanding of "perfect" and "complete proficiency" are a bit off. There is English grammar, and even teenagers can use perfect grammar. Complete proficiency doesn't mean perfect knowledge. Once you can communicate and understand every idea perfectly effective, you have complete proficiency. Being able to observe language and body language, understand it and how it changes the meaning of the words, and recommunicate a response; this is what a complete proficiency is. Perfect is where it can perform its processes perfectly and without error to the most accurate precision needed. This precision only needs to be precise enough that any error is within the noise, as any perfect process will have an error within the noise. In this case, that noise is each person's specific style and the minimums of understandability, anything more "perfect" than this either is 0 difference or an imperfect.
@@rmt3589 "even teenagers can use perfect grammar" no they can't because such a grammar doesn't exist. English teachers attempt to teach about the existence of unicorns so that teenagers can go hunt for them. Not every English speaker has gotten a chance to vote on the correct english grammar. I certainly didn't. Therefore one cannot possibly exist. Otherwise, based on who's authority does the grammar exist? Traditionally, teaching proper grammar has generally been part of the educational arm of genocide and racial oppression. That's the function ot serves. From France to Spain Spain to China to the USA. Teenagers a certainly proficient in their native language. A human is proficient in language well before their teenage years. Language acquisition is rapid and early for humans.
@@rmt3589 and yes our systems today are well sophisticated enough to, say, proofread for us, in a non automated way. They can work as spellchecker and proofreader. They can also also non automated work at classifying tone of online content (for example flagging posts yhay might be threatening violencr. They are less proficient at that but they can be useful.. certainly their accuracy levels are far below "noise". But even trivial conversation is beyond the reach of even the most sophisticated AI. As soon as you make conversation two way it breaks down. I think ai researchers started down the wrong path by trying to teach the form of language without understanding the function of language. They sorta put the cart before the horse.
I really like David Deutsch’s comment on the Turing test: in one of his popular books, he said that the approach of simulating a human chatting is in no way moving us closer to simulating a human thinking. the problem (he said) is that we lack an _explanation_ of what consciousness is. if we had one, then at worst a very slow AI could be produced, and could have been produced on equipment that is obsolete now.
I think a large part of the problem is that many people still subscribe, either knowingly or subconsciously, to the outdated notion of Cartesian duality. In that framework, "consciousness" is very difficult to define, if not impossible, beyond just "an abstract feeling of subjectivity"--which is not a useful nor clear definition. On the other hand, if you look at things neurologically, consciousness becomes much more well-defined (along with every mental process) as some specified pattern, or class of patterns, of neural signals. Once that's realized, defining "consciousness" stops being an insurmountable task of trying to physicalize a metaphysical concept and instead becomes simply looking for patterns of activity in the brains of people who are conscious vs those who are not.
IceMetalPunk You're definitely on to something... Cartesian duality is obviously wrong; but it's difficult to replace... If you look at things neurologically, then "minds are something that brains do" - but it is difficult to say whether you've excluded the important part of the conceptual structure by your chosen epistemological limits. I also thought the idea of slow versus fast consciousness was interesting... does consciousness have to occur in the 'now' ?
ThePeaceableKingdom It's not really difficult to say. When every mental process we've ever looked at the brain for has had a neurological basis, and we haven't found a single empirical piece of evidence for any other mental basis, to think there might be some "more important foundation" for mental processes than neurology is to totally ignore the null hypothesis. And the null hypothesis is foundational because without it, literally any claim which either has no evidence or cannot have evidence is equally valid, no matter how much they contradict anything else.
IceMetalPunk After Newton, reality had to be a clockworks because that's all the model would allow. In a neurological model consciousness has to be neurological because that's all the model will allow. Any empirical experience the model doesn't allow can't be explained within the model, and therefore is either excluded or simply denied. What could the neurologic model make of something like a prophetic dream?... Empirically, one can find folk who will tell you they've had 'em...
I remember watching (on TV) a documentary about Deep Blue, and they SPECIFICALLY clarified that unlike other computers at the time, Deep Blue did NOT simply examine all possible combinations for the next few turns, and pick the one that turns out best at the end of them. Instead, it tried to narrow down the LIKELY combinations the opponent was to make in response to its next move (out of all moves that IT could make), devised a counter against THOSE combos, and repeated the process as it would potentially go on for the next few moves. The difference in that new approach is that it can look at all LIKELY combos for the next 10 to 15 turns, compared to exploring "all" options that would fill in the memory at only 4 or 5 moves ahead (at best) - not enough to beat a chess master. It took a few more seconds per move in order to make all the computations that determine how likely a move is, but since time per turn isn't exactly a requirement in chess, the sacrifice was more than worth it. With all that said, the point is valid regardless. Yes, AI tends to be good, if not better, than humans, but only at a specific narrow task that has clear rules defined.
the new google Alpha zero engine works a bit differently than typical chess engines. It constantly plays games against itself and learns form the moves it makes. I assume it also uses a degree of brute force calculation aswell. When it played at an engine tournament I don't think it lost a single game, half wins, and half draws.
Me: Okay artificial intelligence program, please analyze the difference between time flies like an arrow and fruit flies like a banana. ChatGPT-4: The sentence "Time flies like an arrow; fruit flies like a banana" is a classic example of a garden path sentence and a play on words involving syntactic ambiguity and homonymy. Let's break it down: 1. Time Flies Like an Arrow Meaning: Literal: Time moves quickly and directly, similar to how an arrow flies through the air. Structure: "Time" (subject) "flies" (verb) "like an arrow" (prepositional phrase indicating manner) 2. Fruit Flies Like a Banana Meaning: Literal: Fruit flies (a type of insect) are attracted to bananas. Structure: "Fruit flies" (subject, noun phrase referring to the insect) "like" (verb meaning 'to be fond of' or 'to be attracted to') "a banana" (object of the verb "like") Analysis: Syntactic Ambiguity: The sentence "Fruit flies like a banana" initially leads the reader to interpret "fruit flies" as a verb phrase (like "time flies"), but it is actually a noun phrase. Homonymy: The word "flies" serves as a verb in "Time flies like an arrow" and as part of the noun phrase "fruit flies" in "Fruit flies like a banana." Play on Words: The sentence is a pun, using the dual meanings of "flies" and the syntactic similarity to create humor and confusion. Key Differences: Subject and Verb Usage: In "Time flies like an arrow," "time" is the subject, and "flies" is the verb. In "Fruit flies like a banana," "fruit flies" is the subject, and "like" is the verb. Interpretation: "Time flies like an arrow" is a metaphorical statement about the passage of time. "Fruit flies like a banana" is a literal statement about the behavior of fruit flies. This linguistic play demonstrates how the same words can have different meanings and structures, leading to different interpretations.
AFAIK, Deep Blue did use pattern-matching. Yes, very complex one, but I'm sure I've heard that it's knowledge base was built on a lot of logged and analysed chess games.
Watching this in 2023 and paused it to ask ChatGPT the difference between "time flies like an arrow" and "fruit flies like a banana" and it instantly gave me a far better analysis of the difference than I would have been capable of. Vorsprung durch technik, folks. We let the genie out of the bottle.
That's a fun prompt. I tried it for myself, but it wasn't so smart this time. For the phrase 'time flies like an arrow' it said '"flies" refers to the insect'. Oh my.
It's interesting to note that Turing's formulation was 'Could you tell whether the entity at the other end of the teletype was male or female?' which would have been a natural topic at about 6 minutes in when we get to the subtleties of communication... but that would bring us uncomfortable close to Turing's own actual scandal. (And yay! Cumberbatch, etc...) 'Disambiguation' is such a lovely word! But when you experience it, it's so Zen, isn't it? - that sudden flash of insight... When you realize the meaning of Groucho Marx's joke, "Time flies like an arrow; fruit flies like a banana"...
analyze the difference between "time flies like an arrow" and "fruit flies like a banana"\ "Time flies like an arrow" and "fruit flies like a banana" are two different phrases that use a similar structure but have completely different meanings. "Time flies like an arrow" is a metaphor that suggests that time moves quickly, just like an arrow moves fast through the air. The phrase is often used to emphasize that time passes by quickly and that we should use it wisely. On the other hand, "fruit flies like a banana" is a play on words that uses the same structure but has a completely different meaning. This phrase is a humorous statement that pokes fun at the common fruit fly. It suggests that fruit flies are attracted to bananas, which is a well-known fact, and that they move quickly, just like time. In summary, "time flies like an arrow" is a serious metaphor that emphasizes the fleeting nature of time, while "fruit flies like a banana" is a humorous statement that plays on the similarity in structure between the two phrases to make a joke about fruit flies. - ChatGPT 3.5 Mar 23 Version
are you asking me? I don't have a link to provide you, but an argument could go like this: if you believe that the human brain in its complexity is entirely incomprehensible to humans, what you are implicitly saying is that even given unlimited time and resources, human methods of science will never lead to an understanding of how the human brain/conciousness works. that seems like an unrealistic claim to me, given advances in neurosciences etc... if you google it you'll probably find smarter people than me arguing about this topic. also often in nature very complex structures can arise from a simple set of rules. I'm not saying intelligence is like this, but I don't see why one should a priori exclude the possibility that it is.
Me: Can you analyze the difference between "time flies like an arrow" and "fruit flies like a banana"? ChatGPT: Certainly! "Time flies like an arrow" is a proverb that means time passes quickly. It is often used to express the idea that time seems to pass more quickly as one gets older. "Fruit flies like a banana" is a statement about the behavior of fruit flies, which are small insects that are attracted to overripe or rotting fruit. The statement means that fruit flies are often found near bananas, as they are attracted to the scent of the fruit. The main difference between the two phrases is that "time flies like an arrow" is a proverb that uses the metaphor of time passing quickly to express a common observation about the passage of time. In contrast, "fruit flies like a banana" is a factual statement about the behavior of a particular type of insect.
Thumbs up from Hungary! I love this channel, helped me a lot in my studies in HighSchool, very interesting topics. I will be applying to Cambridge next year, but I am seriously considering Nottingham as well, just because of the 60 Symbols and Computerphile channels. A school with such knowledgeable and energetic professors can't be a bad place anyhow :) Do you have any Hungarian relations, or just picked the language for it's different pronounciation?
I think one thing that is kind of overlooked about the turing test by most is that you probably would have to repeatedly give it. Because humans adapt, and learn, and change. We may not perceive something as what it is until we know what signs to look for. Perhaps, say, we make a machine and it passes the test for a reasonable sample size of people. Then a few years later it fails miserably because people have figured out the quirks. I don't think one test is enough. You'd need not just an A.I. that could fool people today, but one that could continue to evolve to fool people forever. But at that point, it's frightening to think what that could mean.
I'm not sure if would be considered quantum effects, but neurotransmitters aren't shot out across synapses like a photon torpedo. Once they are 'let go' (broken chemical bond?) by one neuron, Brownian motion takes over and some distribution of the neurotransmitters reach the other brain cell. Certainly random. Am I wrong?
Interestingly, chatGPT replies to the say something in another language question like "To translate a phrase or sentence into Hungarian, you can use a translation tool such as Google Translate or consult a Hungarian-English dictionary."
LOLz, looking at the"recent" comments it might be time to revisit this interview with Professor Brailsford. He's always been my favourite and I'd really like to hear this commentary on more recent developments, even if they are not "trendy" ;-) Plz; I apologies if I sound like a teenager, but I have to say, I really like the way he brings things back down to Earth. PS, I wrote that last sentence before realising I was reflecting Prof Brailsford; thumbs up computerphile for influencing not only thinking but apparently language :-)
What does Professor Brailsford think about the I.B.M. Watson computer? A device that does I think does a good job of analyzing context and parsing language. It may not be turing test perfect, in fact instead of the random answers that chatbots give it may be too specific and too correct at some points
It seems, at least on the surface, like the process Roger Penrose used to arrive at the conclusion that "consciousness has something to do with quantum mechanics" is "consciousness is mysterious, quantum mechanics is mysterious, therefore they are related."
I am more interested in stopping the Hal9000 scenario, do you think it will help to watch that or is there something else you might suggest? Thumbs me up if you have a suggestion so I get notified please.
my favourite quote : 7:16 "Ok, artificial intelligence program... analyse the difference between 'time flies like an arrow' and 'fruit flies like a banana'..." - Proffessor Brailsford
Chat 3.4 response to (Analyse the difference between, time files like an arrow and fruit flies like a banana): The phrases "Time flies like an arrow" and "Fruit flies like a banana" are examples of linguistic devices known as puns or wordplay. They play with the multiple meanings of words or phrases to create humorous or thought-provoking statements. Let's analyze the difference between these two phrases: "Time flies like an arrow": This phrase utilizes a metaphor, comparing the passage of time to the flight of a fly or an arrow. It implies that time moves swiftly, just as a fly or an arrow quickly travels through the air. The phrase also suggests that time should be valued and used efficiently, as it can be easily lost or wasted. The pun lies in the wordplay between "flies" (referring to the insects) and "flies" (indicating movement through the air). "Fruit flies like a banana": This phrase also involves a pun, but with a different context. Here, the wordplay is based on the multiple meanings of the term "fruit flies." On one hand, "fruit flies" refers to small insects that are attracted to ripe or decaying fruits. On the other hand, it can be interpreted as "fruit flies" being the action of fruit (as in fruits moving) in a manner similar to how a banana, a type of fruit, might move or be thrown. The phrase cleverly combines these two meanings, creating a humorous effect. In summary, the main difference between the two phrases lies in the objects being compared and the wordplay being used. "Time flies like an arrow" focuses on the metaphorical concept of time passing quickly, drawing a parallel between time and the swift flight of a fly or an arrow. On the other hand, "Fruit flies like a banana" utilizes a play on words involving the multiple meanings of "fruit flies," combining the insect with the action of fruit, specifically a banana. Both phrases demonstrate the creative use of language to evoke amusement or provoke thought.
I've considered the thought before that our brains harness some quantum mechanical properties. But then I observed that thought and it quickly disintegrated... Bad jokes aside, I truly have considered the thought, but I'm no neuroscientist nor a theoretical physicist, so I have no quantifiable education in the matter. But from what I've learned reading various scholarly articles and watching various educational videos on youtube, it doesn't seem completely unlikely that our brains might make use of some quantum mechanical properties. Take for instance our lightening-quick ability to process sound waves, solo out speech, grammatically analyze said speech, and then - the thing computers have yet to even come close to doing - understand those words. That is, as the Professor touched on, we are taking those words, the sentence structure, along with other various clues such as body language, and we put all of that information into context, or we apply it to everything that's going on around us. (For instance, someone mentions a plane crash. Immediately we think, "Was there a plane crash in the news?" "Was there a movie that recently came out about a plane crash?" etc.) Finally, after subconsciously coming up with this list of possible meanings for the airwaves that just graced our eardrums, we subconsciously pick the one that makes the most sense. There is so much more than simple pattern recognition going on that I don't think computers will ever be able to become 100% indistinguishable from humans. Until someone makes a computer that is capable of keeping track of current events, understanding abstract speech like similes and metaphors, that can perceive and make sense of nonverbal cues like body language, can correctly analyze mood and meaning based on tonal differences, and is capable of complex analysis of n-dimensional space in regard to context - and all that in the matter of milliseconds - I refuse to believe that computers will ever become indistinguishable from humans. Finally, to bring it all back to quantum mechanics, it would make sense if the reason we are capable of such incredibly complex analysis is by means of quantum mechanics. Needless to say, the human mind is not just a complex Turing machine.
I don't know how confirmed it is, but it has been claimed that the efficiency of photosynthesis has to do with quantum mechanical effects. You should find it just by googling "quantum mechanics photosynthesis". If true, then it would make sense that other biological processes rely on QM effects. But if it's true, I don't know why some kind of computer couldn't be made to do the same thing.
Here it's time to mention the well known sci-fi based quantum computers. Their bits (also known as qubits) have got the ability to have multiple states at once(1 and 0 to be more precise). Probably our brain might have that property. To process multiple information "blocks" on the same "lane". If you don't really understand what I just said then please tell me and I'll try to reformulate.
ChaSeR Right. If quantum computers can exist, then it is possible to have macroscopic effects from QM properties. It wouldn't prove that consciousness relies on quantum effects, but it would be a proof of the principle.
I recall reading somewhere that there was a Turing Test for game AI, and some neural-network-based bot managed to fool the testers into thinking they were playing against a human opponent.
One of the most difficult things to consider when thinking about a computer that can match a human mind is that the human mind holds outrageous amounts of data. Not just conscious memory but also subconscious and active analysis. I would think, to make a computer that sophisticated we would have to play on the advantage of computers being ageless rather than their efficiency. Why not start with instinct, and move from there. Go into associations, after all that unless we discover that nature vs nurture simply does not exists, we might have to hard code preferences to mimic human behavior.
I think this is really missing the point of what the Turing test is about. The test was Turings answer to the question of whether a machine could ever be said to be truly intelligent. His idea was that the only way you can measure a concept like intelligence is through its performance. You cant look inside the box and find intelligence by analysing its parts, you just have to accept that if every way you look at it it looks smart, you might as well call it smart. Even if one brain is made of spongy tissue and one is made of c++, if they can do the same thing, they are equally intelligent. The conversation test was just an example.
My prompt: Hi ChatGPT-4 I got a riddle for you: "Analyze the difference between time flies like an arrow and fruit flies like a banana." ChatGPT-4 Response: This is a classic example of a linguistic ambiguity, often used to illustrate the complexity of language understanding. The sentence "Time flies like an arrow; fruit flies like a banana" is a humorous play on words that relies on the multiple meanings of the words "flies" and "like". 1. "Time flies like an arrow": In this sentence, "flies" is used as a verb, and "like" is used as a preposition to indicate similarity. The phrase is a metaphor suggesting that time moves swiftly and directly, similar to how an arrow flies. 2. "Fruit flies like a banana": Here, the sentence structure is the same, but the meanings of "flies" and "like" have changed. "Fruit flies" is now a noun referring to a type of small insect, and "like" is a verb meaning "is fond of". So, the sentence is saying that fruit flies are fond of bananas. The riddle is amusing because it uses the same sentence structure to express two completely different ideas, playing on the different meanings of the words.
I'd really love to hear more about the subject of artificial intelligence by Professor Brailsford. I think, if the machines and the AI programs just get complex enough by becoming powerful enough (in calculations per time), so they can closely approximate the human ambiguity, while being programmed with the knowledge about the world, it should be possible to create something that talks and responds almost like a human being. Of course, like with the travelling salesman problem, there will be this 5% of error margin, but I think that is also something which can be reduced to "good enough" by other clever algorithms for pattern recognition. There is so much in the human mind that works based on pattern recognition, and if we have machines with enough layers of pattern recognition connected to the actual "processing engine", it will be possible to closely approximate a human being. Maybe nowadays, we still need the power of several mainframes for that, so it is still impractical, but as always, technology progresses and will provide more computational power on less space and energy consumption. So, in order to fool an average human being, it would not be necessary to understand any quantum effects in the brain that are possibly there (or maybe not). In the end, I think, we may even find out that there cannot be something like a perfectly conclusive Turing test (if that isn't an accepted fact already). Professor Brailsford, please talk more about this subject!
Ok, I'm no expert ether but it's my understanding that computers and organic brains work in very different ways. As I understand it a neuron is much more complicated than the mass of tiny switches that make up a CPU.
Disthron ***** The argument does not say that CPUs and brain cells work the same way, far from it. It acctually is a realization of how more advanced the brain is than any machine: it says that the brain is a machine that processes information (just as a computer) so optimized in its hardware that is holds the CPU, RAM and HD in a liter-worth of space, at about 37°C and it uses food to work. In fact, the whole "humans are machines" argument goes along the same lines: humans are machines, but so optimized that, somehow, we can doubt and make mistakes, all using just logical computations (mental proceses).
***** This actually is a very interesting concept in AI research. Some computer scientists developed an algorithm called "deep learning" (or something along the lines) that did just that: learn from input and started recognizing patterns. Actually, facebook has this (or a very similar version of it) implemented in its face recognition function, and each time you feed facebook with a new photo, the algorithm learns. Google translate uses the same principle, this is why both of them (facebook's face recognition and google's translate) are getting better every day. This is in fact the comeback that AI is been having for the past year: the rupture from the paradigm of inteligence being (as Turing imagined it) a big database of info, to being an algorithm that learns from input (much like a human being).
***** I suppose you could consider each receptor to be one digital input. The problem is that whether a certain receptor has a bound molecule or not isn't determined directly by the "upstream" neuron in the synapse. It's a probabilistic function of a continuous concentration gradient of that chemical. The previous neuron can't control whether each receptor definitely has a bound molecule (1) or not (0), all it can do is increase or decrease the concentration of the chemical it releases. As such, the flow of information is analog, not digital.
***** If you consider "it has a finite number of molecules to release, so it's digital" accurate, then you're in for trouble when you start talking about other topics. Because there is a smallest length--that is, spacetime is discretely quantized--so if that's all it takes for you to consider something digital, then there's no such thing as analog, since everything is discrete on a certain scale.
Yeah, every time I read a new article about how we keep getting closer to real AI, I always think the same thing. Just because computers keep getting faster doesn't mean they are any closer to getting smarter. As the Professor noted, they are great at brute-forcing their way through certain tasks. But it's the software that will make them true AI, and I can't say that I am certain that humans are smart enough to write such code.
The Turing Test really was a child of it's age. The 1950's were the heyday of behaviorism -- a school of psychology that only looks at behavior, and dismisses all talk of feelings, thoughts, and other "internal" states as unscientific. From that view, the question about whether the computer *understands* what you're typing is unimportant or simply meaningless -- to everyone before and after, it's crucial to whether or not we'd consider this kind of AI to be possible. Also, look up "Chinese Room" on wikipedia to read more about this debate.
here is a question I would like brady to ask mathematicians, scientists and computer scientists: what is their opinion on brute forcing to achieve a solution? the rubiks cube "god number" which was found to be 20 was not solved via a mathematical solution. mathematics narrowed down the number of puzzles that required solving but huge computational power was used to brute force the answer that was 20. Is this answer satisfying to a mathematician or computer scientist knowing that brute force was used to prove a problem like this?
Much better than your last video about AI but it would be nice to hear some expert ( Geoffrey Hinton, Andrew Ng, ...) on the field of deeplearning and what he says about the future of AI and how the current best techniques work. But I know it might be a bit hard to get them...^^
Even if quantum effects are involved, you can set every possible wave-function to a different state in a Turing machine. The wave-function is continuous, but you can divide it up into pixels, and make those minute enough that any quantum effects are irrelevant. However, while you could theoretically simulate a human brain like this, it would be far too inefficient for even the best computers. Then again, if moors law continues, we might be able to run such a simulation in a million billion billion billion years. (assuming 4 picometer cells, a 1liter brain and 64 levels of probability that an electron is in a cell)
CharGPT3.5 on “Analyze the difference between ‘Time flies like an arrow’ and ‘Fruit flies like a banana:” “The difference lies in the interpretation of the phrases: 1. ‘Time flies like an arrow’ suggests that time passes quickly and consistently, akin to the swift movement of an arrow. 2. ‘Fruit flies like a banana’ is a play on words, where ‘fruit flies’ refers to the insect, and the sentence humorously suggests that fruit flies have a preference for bananas, which is both literal and figurative.” Perhaps that “10 years” figure wasn’t too optimistic after all!
I'm interested in the turing test for chess. Determine if a chess player is computer by analyzing a finished game. It's important for detecting cheaters.
Let’s raise a simple question: and what about human who fails to pass the Turing test (like for example man with Down's syndrome, an infant, etc)? How we should call a human who fails to pass the Turing test? What word/term we should use for denoting a human who fails to pass the Turing test? If a human fails to pass Turing test then this raises a simple question: “does such human have consciousness or not?”. As we can clearly see from the above examples, the Turing test is unable to determine if the object has consciousness or not. There is not a single scientific tool which would be able to test for the existence of consciousness….. More details can be found at “Neurocluster Brain Model” site.
Could something similar to Schrödinger's Cat be applied to the Turing Test, as in, "Whoever, or whatever, is on the other end is both man _and_ machine until observed."
I don't think so. Schrodinger's cat isn't primarily about the act of observation. It has more to do with a concept called quantum superposition. Think about the single particle double slit experiment. It seems like a single photon goes through both slits, interferes with itself, and then lands on the screen. Asking which slit the photon actually went through doesn't even make any sense. The answer isn't "both slits" and it's not "neither slit." It's much more mysterious than that. The photon's position is not defined if unobserved. You might say it doesn't even have a position until observed. That's the key difference between this and Schrodinger's cat. In Schrodinger's cat, there is a quantum superposition (from the radioactive element) so the state of the isolated radioactive element and cat system isn't defined until observed. In the Turing test situation, there isn't any quantum superposition happening. Whether what you're talking to is or isn't a human is a definite property. It's not an issue of what you know, it's an issue of whether what you're talking about is actually defined.
did deep blue win (@2:50) by brute force, i could've sworn it actually did use some pattern rec... but i'm too lazy to google (plus i'm at work)... any one want to fill me in?
In Turing's time, it was thought that if anyone ever could program a computer to beat a Grand Master in chess, then it would be intellegent. Then we did it and immediatly everyone said, well THAT isn't REAL intellegence. But, if we could program a computer to play Go... Then we did that. So, they said, if we could translate natural language. Ah, Google translate? Whoops.
The problems of subtlety, metaphor, and common usage will be the ones that prevent strong A.I. from being realized. Unless you can create a program that can learn, experience, remember, synthesize, generate and understand abstractions, you can never artificially construct consciousness.
What about artificial neural networks? They're incredibly good at learning and they're more generalized, also the brain is largely an incredibly deep neural network. Could artificial neural networks be the key to making a more human-like artificial intelligence?
MrMiljaker But how's that different from a human brain? The human brain has a predefined number of inputs. Nobody's born with 3 eyeballs. The amount of inputs into the brain is constant. The amount of outputs is constant. And it's not true at all that the inputs have to be "specific", either. For example, let's say I made an ANN that can learn to play Super Mario World. All I need to do is give it inputs from the display on the screen and outputs to the controller. In fact, after I built it, it'd be capable of learning to play *any* Super Nintendo game after that because it has the proper inputs and outputs. With a deep enough ANN, it could learn the difference between games as well. So a single ANN could learn to play the entire library of Super Nintendo games. If you gave it visual inputs less specific to the Super Nintendo display, like eyeballs, and outputs less specific to the Super Nintendo controller, like hands, it could learn to play _any_ video game. Obviously, the less specific your neural net is to the problem, the deeper it has to be to compensate. But in my hypothetical I proposed, depth isn't a factor. In this hypothetical your neural network can be as deep as you want it.
Deep Blue had the benefit of analysing all Kasparov's matches with the back up of computer programmers and chess experts . Kasparov was facing an opponent on his own basically. It was a narrow victory for the Deep Blue in the end
Good thing computers speak in series of clearly distinguishable energy levels, not in terrible puns that only work in a language composed of what seems like 95 % homophones ...
Computers will never become as fast as the mind until the way a computer processes information is not through the software changing, but instead through the hardware changing. Digital may be efficient and simple, but it doesn't, and will never, take into account the incredibly complexity and amount of information that can fit into a tiny space.
Me: hello, define the difference between these 2 sentences. 1. Time flies like an arrow 2. Fruit flies like a banana. ChatGPT: The first sentence, "Time flies like an arrow," is a metaphor that compares the passage of time to the motion of an arrow. The second sentence, "Fruit flies like a banana," is a statement about the behavior of a type of insect known as fruit flies, which are attracted to ripe bananas and other fruits. In this sentence, "like" is being used in the sense of "enjoy" or "prefer," rather than as a comparison.
i guess there is a big difference between AI and VI. AI are sentient beings with knowledge of their existence. VI (virtual intelligence) are just simulations that can act like a sentient being but not being one.
I guess the argument against the Turing test is, if you make a machine that can pass the Turing test, then you've made a machine that's really good at passing the Turing test, and it says nothing about general intelligence.
Kasparov was strong enough to beat Deep Blue back in the days of the match, he just failed to do so for many reasons some of them unknown and suspicious... Its today's computers that are like Professor Brailsford explained, nobody has a chance against them not even chess champions.
If someone gave me the Turing test, they would think I was a machine.
invisibledave comment was made by a bot
Same. When they talked about machines being unable to understand the subtleties of human interactions, I immediatly thought of my own way of thinking and how it more resembles the machine.
@@user-eu5ol7mx8y wow u so special
Believe it or not humans have failed the Turing test 😂😂 idiots right
false.
I just love how Professor Brailsford explains stuff :)
I took a linguistics class with a professor who'd worked on very early language software, trying to teach computers how grammar works in the 70s. They thought it would be the simple, given how mathematically language tends to be, but when they started working with them they discovered that the social and idiopathic nature of language made it virtually impossible.
It's easy now, just use GPT3/GPT-J, and mark punctuation as words. (Eg "." Is a word that means "end of sentence")
@@rmt3589 to a degree. Ai has come a long way in being able to interpret this information.
It's not perfect yet. And it cannot be perfect because if it were perfect it would be wrong, ironically. There is no English grammar. Every human being uses language in their own way, according to their own set of rules, for their own purposes. An ai would need to consult with every Englishspeaker, past and present, to come to a complete proficiency in English.
And even if they could, such a machine would be useless because that's not how humans learn language and any process that learned language like that would not lead to outcomes that result from the way humans learn language.
@@nickscurvy8635 I think the understanding of "perfect" and "complete proficiency" are a bit off. There is English grammar, and even teenagers can use perfect grammar.
Complete proficiency doesn't mean perfect knowledge. Once you can communicate and understand every idea perfectly effective, you have complete proficiency. Being able to observe language and body language, understand it and how it changes the meaning of the words, and recommunicate a response; this is what a complete proficiency is.
Perfect is where it can perform its processes perfectly and without error to the most accurate precision needed. This precision only needs to be precise enough that any error is within the noise, as any perfect process will have an error within the noise. In this case, that noise is each person's specific style and the minimums of understandability, anything more "perfect" than this either is 0 difference or an imperfect.
@@rmt3589 "even teenagers can use perfect grammar" no they can't because such a grammar doesn't exist. English teachers attempt to teach about the existence of unicorns so that teenagers can go hunt for them. Not every English speaker has gotten a chance to vote on the correct english grammar. I certainly didn't. Therefore one cannot possibly exist. Otherwise, based on who's authority does the grammar exist? Traditionally, teaching proper grammar has generally been part of the educational arm of genocide and racial oppression. That's the function ot serves. From France to Spain Spain to China to the USA. Teenagers a certainly proficient in their native language. A human is proficient in language well before their teenage years. Language acquisition is rapid and early for humans.
@@rmt3589 and yes our systems today are well sophisticated enough to, say, proofread for us, in a non automated way. They can work as spellchecker and proofreader. They can also also non automated work at classifying tone of online content (for example flagging posts yhay might be threatening violencr. They are less proficient at that but they can be useful.. certainly their accuracy levels are far below "noise".
But even trivial conversation is beyond the reach of even the most sophisticated AI. As soon as you make conversation two way it breaks down. I think ai researchers started down the wrong path by trying to teach the form of language without understanding the function of language. They sorta put the cart before the horse.
I would love to see this interview updated.
I really like David Deutsch’s comment on the Turing test: in one of his popular books, he said that the approach of simulating a human chatting is in no way moving us closer to simulating a human thinking. the problem (he said) is that we lack an _explanation_ of what consciousness is. if we had one, then at worst a very slow AI could be produced, and could have been produced on equipment that is obsolete now.
"if we had one, then at worst a very slow AI could be produced, and could have been produced on equipment that is obsolete now."
interesting idea...
I think a large part of the problem is that many people still subscribe, either knowingly or subconsciously, to the outdated notion of Cartesian duality. In that framework, "consciousness" is very difficult to define, if not impossible, beyond just "an abstract feeling of subjectivity"--which is not a useful nor clear definition. On the other hand, if you look at things neurologically, consciousness becomes much more well-defined (along with every mental process) as some specified pattern, or class of patterns, of neural signals. Once that's realized, defining "consciousness" stops being an insurmountable task of trying to physicalize a metaphysical concept and instead becomes simply looking for patterns of activity in the brains of people who are conscious vs those who are not.
IceMetalPunk
You're definitely on to something... Cartesian duality is obviously wrong; but it's difficult to replace...
If you look at things neurologically, then "minds are something that brains do" - but it is difficult to say whether you've excluded the important part of the conceptual structure by your chosen epistemological limits.
I also thought the idea of slow versus fast consciousness was interesting... does consciousness have to occur in the 'now' ?
ThePeaceableKingdom
It's not really difficult to say. When every mental process we've ever looked at the brain for has had a neurological basis, and we haven't found a single empirical piece of evidence for any other mental basis, to think there might be some "more important foundation" for mental processes than neurology is to totally ignore the null hypothesis. And the null hypothesis is foundational because without it, literally any claim which either has no evidence or cannot have evidence is equally valid, no matter how much they contradict anything else.
IceMetalPunk
After Newton, reality had to be a clockworks because that's all the model would allow.
In a neurological model consciousness has to be neurological because that's all the model will allow. Any empirical experience the model doesn't allow can't be explained within the model, and therefore is either excluded or simply denied.
What could the neurologic model make of something like a prophetic dream?... Empirically, one can find folk who will tell you they've had 'em...
Me: My name is Ron
AI: Hello
Me: Whats my name?\
AI: Your name is cat.
Kat
Typical. The AI's have clearly been out drinking.
No ron, go find becky
ok?
man there is a way Brailsford talks that just captures me ...... couldn't stop watching
I remember watching (on TV) a documentary about Deep Blue, and they SPECIFICALLY clarified that unlike other computers at the time, Deep Blue did NOT simply examine all possible combinations for the next few turns, and pick the one that turns out best at the end of them.
Instead, it tried to narrow down the LIKELY combinations the opponent was to make in response to its next move (out of all moves that IT could make), devised a counter against THOSE combos, and repeated the process as it would potentially go on for the next few moves. The difference in that new approach is that it can look at all LIKELY combos for the next 10 to 15 turns, compared to exploring "all" options that would fill in the memory at only 4 or 5 moves ahead (at best) - not enough to beat a chess master. It took a few more seconds per move in order to make all the computations that determine how likely a move is, but since time per turn isn't exactly a requirement in chess, the sacrifice was more than worth it.
With all that said, the point is valid regardless. Yes, AI tends to be good, if not better, than humans, but only at a specific narrow task that has clear rules defined.
Professor Brailsford is great, he's so passionate that you can't help but be interested!
false.
the new google Alpha zero engine works a bit differently than typical chess engines. It constantly plays games against itself and learns form the moves it makes. I assume it also uses a degree of brute force calculation aswell. When it played at an engine tournament I don't think it lost a single game, half wins, and half draws.
Yes it used a neural net system *and* tree searching.
It played a 100 game match* with Stockfish 8. Alpha Zero scored 28 wins, 0 losses, 72 draws.
That worked well for WOPR.
I would love to know what he would think about this topic right now ...
Me: Okay artificial intelligence program, please analyze the difference between time flies like an arrow and fruit flies like a banana.
ChatGPT-4: The sentence "Time flies like an arrow; fruit flies like a banana" is a classic example of a garden path sentence and a play on words involving syntactic ambiguity and homonymy. Let's break it down:
1. Time Flies Like an Arrow
Meaning:
Literal: Time moves quickly and directly, similar to how an arrow flies through the air.
Structure:
"Time" (subject)
"flies" (verb)
"like an arrow" (prepositional phrase indicating manner)
2. Fruit Flies Like a Banana
Meaning:
Literal: Fruit flies (a type of insect) are attracted to bananas.
Structure:
"Fruit flies" (subject, noun phrase referring to the insect)
"like" (verb meaning 'to be fond of' or 'to be attracted to')
"a banana" (object of the verb "like")
Analysis:
Syntactic Ambiguity: The sentence "Fruit flies like a banana" initially leads the reader to interpret "fruit flies" as a verb phrase (like "time flies"), but it is actually a noun phrase.
Homonymy: The word "flies" serves as a verb in "Time flies like an arrow" and as part of the noun phrase "fruit flies" in "Fruit flies like a banana."
Play on Words: The sentence is a pun, using the dual meanings of "flies" and the syntactic similarity to create humor and confusion.
Key Differences:
Subject and Verb Usage:
In "Time flies like an arrow," "time" is the subject, and "flies" is the verb.
In "Fruit flies like a banana," "fruit flies" is the subject, and "like" is the verb.
Interpretation:
"Time flies like an arrow" is a metaphorical statement about the passage of time.
"Fruit flies like a banana" is a literal statement about the behavior of fruit flies.
This linguistic play demonstrates how the same words can have different meanings and structures, leading to different interpretations.
AFAIK, Deep Blue did use pattern-matching. Yes, very complex one, but I'm sure I've heard that it's knowledge base was built on a lot of logged and analysed chess games.
Watching this in 2023 and paused it to ask ChatGPT the difference between "time flies like an arrow" and "fruit flies like a banana" and it instantly gave me a far better analysis of the difference than I would have been capable of.
Vorsprung durch technik, folks. We let the genie out of the bottle.
That's a fun prompt. I tried it for myself, but it wasn't so smart this time. For the phrase 'time flies like an arrow' it said '"flies" refers to the insect'. Oh my.
Love this guy :) So energetic!
the last few minutes of this made me laugh way more than it should have.
It's interesting to note that Turing's formulation was 'Could you tell whether the entity at the other end of the teletype was male or female?' which would have been a natural topic at about 6 minutes in when we get to the subtleties of communication... but that would bring us uncomfortable close to Turing's own actual scandal. (And yay! Cumberbatch, etc...)
'Disambiguation' is such a lovely word! But when you experience it, it's so Zen, isn't it? - that sudden flash of insight... When you realize the meaning of Groucho Marx's joke, "Time flies like an arrow; fruit flies like a banana"...
*****
Yes!
and where exactly does complexity go when you abstract it?...
*****
Well, in my personal experience, usually more 'heat' than 'light'...
I would ask a computer what it itself would ask another computer during a Turing test
It might ask what is the machines intent. So what is your intent?
analyze the difference between "time flies like an arrow" and "fruit flies like a banana"\
"Time flies like an arrow" and "fruit flies like a banana" are two different phrases that use a similar structure but have completely different meanings.
"Time flies like an arrow" is a metaphor that suggests that time moves quickly, just like an arrow moves fast through the air. The phrase is often used to emphasize that time passes by quickly and that we should use it wisely.
On the other hand, "fruit flies like a banana" is a play on words that uses the same structure but has a completely different meaning. This phrase is a humorous statement that pokes fun at the common fruit fly. It suggests that fruit flies are attracted to bananas, which is a well-known fact, and that they move quickly, just like time.
In summary, "time flies like an arrow" is a serious metaphor that emphasizes the fleeting nature of time, while "fruit flies like a banana" is a humorous statement that plays on the similarity in structure between the two phrases to make a joke about fruit flies. - ChatGPT 3.5 Mar 23 Version
If the human brain were so simple we could understand it, we would be so simple, we couldn't.
Emerson M. Pugh
axxization Yes, that's the guy. Thanks.
it's a poignant phrase but not necessarily true
Proof that it isn't true? Post a link so I can understand why.
are you asking me? I don't have a link to provide you, but an argument could go like this: if you believe that the human brain in its complexity is entirely incomprehensible to humans, what you are implicitly saying is that even given unlimited time and resources, human methods of science will never lead to an understanding of how the human brain/conciousness works. that seems like an unrealistic claim to me, given advances in neurosciences etc... if you google it you'll probably find smarter people than me arguing about this topic. also often in nature very complex structures can arise from a simple set of rules. I'm not saying intelligence is like this, but I don't see why one should a priori exclude the possibility that it is.
Me: Can you analyze the difference between "time flies like an arrow" and "fruit flies like a banana"?
ChatGPT:
Certainly! "Time flies like an arrow" is a proverb that means time passes quickly. It is often used to express the idea that time seems to pass more quickly as one gets older.
"Fruit flies like a banana" is a statement about the behavior of fruit flies, which are small insects that are attracted to overripe or rotting fruit. The statement means that fruit flies are often found near bananas, as they are attracted to the scent of the fruit.
The main difference between the two phrases is that "time flies like an arrow" is a proverb that uses the metaphor of time passing quickly to express a common observation about the passage of time. In contrast, "fruit flies like a banana" is a factual statement about the behavior of a particular type of insect.
Thumbs up from Hungary!
I love this channel, helped me a lot in my studies in HighSchool, very interesting topics. I will be applying to Cambridge next year, but I am seriously considering Nottingham as well, just because of the 60 Symbols and Computerphile channels. A school with such knowledgeable and energetic professors can't be a bad place anyhow :)
Do you have any Hungarian relations, or just picked the language for it's different pronounciation?
Also Thumbs up for the HTC One! :))
where did you end up mate?
ok?
ok?
I think one thing that is kind of overlooked about the turing test by most is that you probably would have to repeatedly give it. Because humans adapt, and learn, and change. We may not perceive something as what it is until we know what signs to look for. Perhaps, say, we make a machine and it passes the test for a reasonable sample size of people. Then a few years later it fails miserably because people have figured out the quirks. I don't think one test is enough. You'd need not just an A.I. that could fool people today, but one that could continue to evolve to fool people forever. But at that point, it's frightening to think what that could mean.
I'm not sure if would be considered quantum effects, but neurotransmitters aren't shot out across synapses like a photon torpedo. Once they are 'let go' (broken chemical bond?) by one neuron, Brownian motion takes over and some distribution of the neurotransmitters reach the other brain cell.
Certainly random. Am I wrong?
Interestingly, chatGPT replies to the say something in another language question like "To translate a phrase or sentence into Hungarian, you can use a translation tool such as Google Translate or consult a Hungarian-English dictionary."
LOLz, looking at the"recent" comments it might be time to revisit this interview with Professor Brailsford. He's always been my favourite and I'd really like to hear this commentary on more recent developments, even if they are not "trendy" ;-) Plz; I apologies if I sound like a teenager, but I have to say, I really like the way he brings things back down to Earth. PS, I wrote that last sentence before realising I was reflecting Prof Brailsford; thumbs up computerphile for influencing not only thinking but apparently language :-)
What does Professor Brailsford think about the I.B.M. Watson computer? A device that does I think does a good job of analyzing context and parsing language. It may not be turing test perfect, in fact instead of the random answers that chatbots give it may be too specific and too correct at some points
It seems, at least on the surface, like the process Roger Penrose used to arrive at the conclusion that "consciousness has something to do with quantum mechanics" is "consciousness is mysterious, quantum mechanics is mysterious, therefore they are related."
If you want to know more about Deep Blue watch "Game Over:Kasparov and the machine" here on YT.
I am more interested in stopping the Hal9000 scenario, do you think it will help to watch that or is there something else you might suggest? Thumbs me up if you have a suggestion so I get notified please.
my favourite quote : 7:16
"Ok, artificial intelligence program... analyse the difference between 'time flies like an arrow' and 'fruit flies like a banana'..." - Proffessor Brailsford
No longer funny ;-)
Chat 3.4 response to (Analyse the difference between, time files like an arrow and fruit flies like a banana):
The phrases "Time flies like an arrow" and "Fruit flies like a banana" are examples of linguistic devices known as puns or wordplay. They play with the multiple meanings of words or phrases to create humorous or thought-provoking statements. Let's analyze the difference between these two phrases:
"Time flies like an arrow":
This phrase utilizes a metaphor, comparing the passage of time to the flight of a fly or an arrow. It implies that time moves swiftly, just as a fly or an arrow quickly travels through the air. The phrase also suggests that time should be valued and used efficiently, as it can be easily lost or wasted. The pun lies in the wordplay between "flies" (referring to the insects) and "flies" (indicating movement through the air).
"Fruit flies like a banana":
This phrase also involves a pun, but with a different context. Here, the wordplay is based on the multiple meanings of the term "fruit flies." On one hand, "fruit flies" refers to small insects that are attracted to ripe or decaying fruits. On the other hand, it can be interpreted as "fruit flies" being the action of fruit (as in fruits moving) in a manner similar to how a banana, a type of fruit, might move or be thrown. The phrase cleverly combines these two meanings, creating a humorous effect.
In summary, the main difference between the two phrases lies in the objects being compared and the wordplay being used. "Time flies like an arrow" focuses on the metaphorical concept of time passing quickly, drawing a parallel between time and the swift flight of a fly or an arrow. On the other hand, "Fruit flies like a banana" utilizes a play on words involving the multiple meanings of "fruit flies," combining the insect with the action of fruit, specifically a banana. Both phrases demonstrate the creative use of language to evoke amusement or provoke thought.
OMG. For all these years I've misunderstood the fruit fly joke. I'm glad I wasn't in charge of talking to any AIs.
you didn't misunderstand ...it is ambiguous:)
This is interesting, because you would be a false negative of a Touring Test! :o That is how AI will take control, perhaps? hehehe
I've considered the thought before that our brains harness some quantum mechanical properties. But then I observed that thought and it quickly disintegrated...
Bad jokes aside, I truly have considered the thought, but I'm no neuroscientist nor a theoretical physicist, so I have no quantifiable education in the matter. But from what I've learned reading various scholarly articles and watching various educational videos on youtube, it doesn't seem completely unlikely that our brains might make use of some quantum mechanical properties.
Take for instance our lightening-quick ability to process sound waves, solo out speech, grammatically analyze said speech, and then - the thing computers have yet to even come close to doing - understand those words. That is, as the Professor touched on, we are taking those words, the sentence structure, along with other various clues such as body language, and we put all of that information into context, or we apply it to everything that's going on around us. (For instance, someone mentions a plane crash. Immediately we think, "Was there a plane crash in the news?" "Was there a movie that recently came out about a plane crash?" etc.) Finally, after subconsciously coming up with this list of possible meanings for the airwaves that just graced our eardrums, we subconsciously pick the one that makes the most sense.
There is so much more than simple pattern recognition going on that I don't think computers will ever be able to become 100% indistinguishable from humans. Until someone makes a computer that is capable of keeping track of current events, understanding abstract speech like similes and metaphors, that can perceive and make sense of nonverbal cues like body language, can correctly analyze mood and meaning based on tonal differences, and is capable of complex analysis of n-dimensional space in regard to context - and all that in the matter of milliseconds - I refuse to believe that computers will ever become indistinguishable from humans.
Finally, to bring it all back to quantum mechanics, it would make sense if the reason we are capable of such incredibly complex analysis is by means of quantum mechanics. Needless to say, the human mind is not just a complex Turing machine.
I don't know how confirmed it is, but it has been claimed that the efficiency of photosynthesis has to do with quantum mechanical effects. You should find it just by googling "quantum mechanics photosynthesis". If true, then it would make sense that other biological processes rely on QM effects.
But if it's true, I don't know why some kind of computer couldn't be made to do the same thing.
Here it's time to mention the well known sci-fi based quantum computers. Their bits (also known as qubits) have got the ability to have multiple states at once(1 and 0 to be more precise). Probably our brain might have that property. To process multiple information "blocks" on the same "lane". If you don't really understand what I just said then please tell me and I'll try to reformulate.
ChaSeR Right. If quantum computers can exist, then it is possible to have macroscopic effects from QM properties. It wouldn't prove that consciousness relies on quantum effects, but it would be a proof of the principle.
ChatGPT sends its regards
I recall reading somewhere that there was a Turing Test for game AI, and some neural-network-based bot managed to fool the testers into thinking they were playing against a human opponent.
Another excellent video!
The WHY motivations to portions of the conversation can quickly lead to a background not existing.
Exactly sir! Excellent answer!
Can we get the professor's opinion on IBM's Watson and Jeopardy?
probably the greatest explanation of my life
One of the most difficult things to consider when thinking about a computer that can match a human mind is that the human mind holds outrageous amounts of data. Not just conscious memory but also subconscious and active analysis.
I would think, to make a computer that sophisticated we would have to play on the advantage of computers being ageless rather than their efficiency. Why not start with instinct, and move from there. Go into associations, after all that unless we discover that nature vs nurture simply does not exists, we might have to hard code preferences to mimic human behavior.
I think this is really missing the point of what the Turing test is about.
The test was Turings answer to the question of whether a machine could ever be said to be truly intelligent. His idea was that the only way you can measure a concept like intelligence is through its performance. You cant look inside the box and find intelligence by analysing its parts, you just have to accept that if every way you look at it it looks smart, you might as well call it smart. Even if one brain is made of spongy tissue and one is made of c++, if they can do the same thing, they are equally intelligent. The conversation test was just an example.
My prompt:
Hi ChatGPT-4 I got a riddle for you:
"Analyze the difference between time flies like an arrow and fruit flies like a banana."
ChatGPT-4 Response:
This is a classic example of a linguistic ambiguity, often used to illustrate the complexity of language understanding. The sentence "Time flies like an arrow; fruit flies like a banana" is a humorous play on words that relies on the multiple meanings of the words "flies" and "like".
1. "Time flies like an arrow": In this sentence, "flies" is used as a verb, and "like" is used as a preposition to indicate similarity. The phrase is a metaphor suggesting that time moves swiftly and directly, similar to how an arrow flies.
2. "Fruit flies like a banana": Here, the sentence structure is the same, but the meanings of "flies" and "like" have changed. "Fruit flies" is now a noun referring to a type of small insect, and "like" is a verb meaning "is fond of". So, the sentence is saying that fruit flies are fond of bananas.
The riddle is amusing because it uses the same sentence structure to express two completely different ideas, playing on the different meanings of the words.
I enjoyed the time flies example
I'd really love to hear more about the subject of artificial intelligence by Professor Brailsford. I think, if the machines and the AI programs just get complex enough by becoming powerful enough (in calculations per time), so they can closely approximate the human ambiguity, while being programmed with the knowledge about the world, it should be possible to create something that talks and responds almost like a human being. Of course, like with the travelling salesman problem, there will be this 5% of error margin, but I think that is also something which can be reduced to "good enough" by other clever algorithms for pattern recognition. There is so much in the human mind that works based on pattern recognition, and if we have machines with enough layers of pattern recognition connected to the actual "processing engine", it will be possible to closely approximate a human being. Maybe nowadays, we still need the power of several mainframes for that, so it is still impractical, but as always, technology progresses and will provide more computational power on less space and energy consumption. So, in order to fool an average human being, it would not be necessary to understand any quantum effects in the brain that are possibly there (or maybe not). In the end, I think, we may even find out that there cannot be something like a perfectly conclusive Turing test (if that isn't an accepted fact already). Professor Brailsford, please talk more about this subject!
5:55 I lost it. Haha
Simply fascinating.
Ok, I'm no expert ether but it's my understanding that computers and organic brains work in very different ways. As I understand it a neuron is much more complicated than the mass of tiny switches that make up a CPU.
Disthron ***** The argument does not say that CPUs and brain cells work the same way, far from it. It acctually is a realization of how more advanced the brain is than any machine: it says that the brain is a machine that processes information (just as a computer) so optimized in its hardware that is holds the CPU, RAM and HD in a liter-worth of space, at about 37°C and it uses food to work. In fact, the whole "humans are machines" argument goes along the same lines: humans are machines, but so optimized that, somehow, we can doubt and make mistakes, all using just logical computations (mental proceses).
***** This actually is a very interesting concept in AI research. Some computer scientists developed an algorithm called "deep learning" (or something along the lines) that did just that: learn from input and started recognizing patterns.
Actually, facebook has this (or a very similar version of it) implemented in its face recognition function, and each time you feed facebook with a new photo, the algorithm learns. Google translate uses the same principle, this is why both of them (facebook's face recognition and google's translate) are getting better every day. This is in fact the comeback that AI is been having for the past year: the rupture from the paradigm of inteligence being (as Turing imagined it) a big database of info, to being an algorithm that learns from input (much like a human being).
***** Thats exactly why the human brain is way waaaay more advanced than a machine: a super-optimized computer.
*****
I suppose you could consider each receptor to be one digital input. The problem is that whether a certain receptor has a bound molecule or not isn't determined directly by the "upstream" neuron in the synapse. It's a probabilistic function of a continuous concentration gradient of that chemical. The previous neuron can't control whether each receptor definitely has a bound molecule (1) or not (0), all it can do is increase or decrease the concentration of the chemical it releases. As such, the flow of information is analog, not digital.
*****
If you consider "it has a finite number of molecules to release, so it's digital" accurate, then you're in for trouble when you start talking about other topics. Because there is a smallest length--that is, spacetime is discretely quantized--so if that's all it takes for you to consider something digital, then there's no such thing as analog, since everything is discrete on a certain scale.
Yeah, every time I read a new article about how we keep getting closer to real AI, I always think the same thing. Just because computers keep getting faster doesn't mean they are any closer to getting smarter. As the Professor noted, they are great at brute-forcing their way through certain tasks. But it's the software that will make them true AI, and I can't say that I am certain that humans are smart enough to write such code.
The Turing Test really was a child of it's age. The 1950's were the heyday of behaviorism -- a school of psychology that only looks at behavior, and dismisses all talk of feelings, thoughts, and other "internal" states as unscientific. From that view, the question about whether the computer *understands* what you're typing is unimportant or simply meaningless -- to everyone before and after, it's crucial to whether or not we'd consider this kind of AI to be possible. Also, look up "Chinese Room" on wikipedia to read more about this debate.
Could you post an URL to the book by Roger Penrose that Professor Brailsford is talking about?
here is a question I would like brady to ask mathematicians, scientists and computer scientists:
what is their opinion on brute forcing to achieve a solution?
the rubiks cube "god number" which was found to be 20 was not solved via a mathematical solution. mathematics narrowed down the number of puzzles that required solving but huge computational power was used to brute force the answer that was 20. Is this answer satisfying to a mathematician or computer scientist knowing that brute force was used to prove a problem like this?
Can you have a turing test to find out if you're playing against a human or a robot at chess?
can there be something on this channel like calculator unboxing for numberophile? perhaps software/app 1st use?
Much better than your last video about AI but it would be nice to hear some expert ( Geoffrey Hinton, Andrew Ng, ...) on the field of deeplearning and what he says about the future of AI and how the current best techniques work. But I know it might be a bit hard to get them...^^
So even incredibly intelligent people like Roger Penrose fall prey to the quantum-physics-equals-magic fallacy.
You so clever. You write book now.
This was amazing and interesting!
Even if quantum effects are involved, you can set every possible wave-function to a different state in a Turing machine. The wave-function is continuous, but you can divide it up into pixels, and make those minute enough that any quantum effects are irrelevant. However, while you could theoretically simulate a human brain like this, it would be far too inefficient for even the best computers. Then again, if moors law continues, we might be able to run such a simulation in a million billion billion billion years. (assuming 4 picometer cells, a 1liter brain and 64 levels of probability that an electron is in a cell)
CharGPT3.5 on “Analyze the difference between ‘Time flies like an arrow’ and ‘Fruit flies like a banana:”
“The difference lies in the interpretation of the phrases:
1. ‘Time flies like an arrow’ suggests that time passes quickly and consistently, akin to the swift movement of an arrow.
2. ‘Fruit flies like a banana’ is a play on words, where ‘fruit flies’ refers to the insect, and the sentence humorously suggests that fruit flies have a preference for bananas, which is both literal and figurative.”
Perhaps that “10 years” figure wasn’t too optimistic after all!
I'm interested in the turing test for chess. Determine if a chess player is computer by analyzing a finished game. It's important for detecting cheaters.
Let’s raise a simple question: and what about human who fails to pass the Turing test (like for example man with Down's syndrome, an infant, etc)? How we should call a human who fails to pass the Turing test? What word/term we should use for denoting a human who fails to pass the Turing test? If a human fails to pass Turing test then this raises a simple question: “does such human have consciousness or not?”. As we can clearly see from the above examples, the Turing test is unable to determine if the object has consciousness or not. There is not a single scientific tool which would be able to test for the existence of consciousness….. More details can be found at “Neurocluster Brain Model” site.
i was really confused by "fruit flies like a banana" - dammit i'm a robot!
Could something similar to Schrödinger's Cat be applied to the Turing Test, as in, "Whoever, or whatever, is on the other end is both man _and_ machine until observed."
I don't think so. Schrodinger's cat isn't primarily about the act of observation. It has more to do with a concept called quantum superposition. Think about the single particle double slit experiment. It seems like a single photon goes through both slits, interferes with itself, and then lands on the screen. Asking which slit the photon actually went through doesn't even make any sense. The answer isn't "both slits" and it's not "neither slit." It's much more mysterious than that. The photon's position is not defined if unobserved. You might say it doesn't even have a position until observed. That's the key difference between this and Schrodinger's cat. In Schrodinger's cat, there is a quantum superposition (from the radioactive element) so the state of the isolated radioactive element and cat system isn't defined until observed. In the Turing test situation, there isn't any quantum superposition happening. Whether what you're talking to is or isn't a human is a definite property. It's not an issue of what you know, it's an issue of whether what you're talking about is actually defined.
did deep blue win (@2:50) by brute force, i could've sworn it actually did use some pattern rec... but i'm too lazy to google (plus i'm at work)... any one want to fill me in?
In Turing's time, it was thought that if anyone ever could program a computer to beat a Grand Master in chess, then it would be intellegent.
Then we did it and immediatly everyone said, well THAT isn't REAL intellegence.
But, if we could program a computer to play Go... Then we did that.
So, they said, if we could translate natural language.
Ah, Google translate? Whoops.
well fruit *would* fly pretty similar to bananas, just saying haha
The problems of subtlety, metaphor, and common usage will be the ones that prevent strong A.I. from being realized. Unless you can create a program that can learn, experience, remember, synthesize, generate and understand abstractions, you can never artificially construct consciousness.
What about artificial neural networks? They're incredibly good at learning and they're more generalized, also the brain is largely an incredibly deep neural network. Could artificial neural networks be the key to making a more human-like artificial intelligence?
MrMiljaker But how's that different from a human brain? The human brain has a predefined number of inputs. Nobody's born with 3 eyeballs. The amount of inputs into the brain is constant. The amount of outputs is constant.
And it's not true at all that the inputs have to be "specific", either. For example, let's say I made an ANN that can learn to play Super Mario World. All I need to do is give it inputs from the display on the screen and outputs to the controller. In fact, after I built it, it'd be capable of learning to play *any* Super Nintendo game after that because it has the proper inputs and outputs.
With a deep enough ANN, it could learn the difference between games as well. So a single ANN could learn to play the entire library of Super Nintendo games.
If you gave it visual inputs less specific to the Super Nintendo display, like eyeballs, and outputs less specific to the Super Nintendo controller, like hands, it could learn to play _any_ video game.
Obviously, the less specific your neural net is to the problem, the deeper it has to be to compensate. But in my hypothetical I proposed, depth isn't a factor. In this hypothetical your neural network can be as deep as you want it.
Deep Blue had the benefit of analysing all Kasparov's matches with the back up of computer programmers and chess experts . Kasparov was facing an opponent on his own basically. It was a narrow victory for the Deep Blue in the end
At 6:48 my Google Home said "I don't speak that language yet."
***** Could you please do a video on esoteric programming languages?
What happened to Brady...?
Good thing computers speak in series of clearly distinguishable energy levels, not in terrible puns that only work in a language composed of what seems like 95 % homophones ...
It's just an example. Equally ambitious, context sensitive questions exist outside of puns.
Do one about IMB's Watson (extrabits).
My phone picked up that "Okay Google".
It seems very likely that the source of self and consciousness will be found to be quantum.
You: What is your name
AI: John
You: That is my name also! What is my name?
AI...
Computers will never become as fast as the mind until the way a computer processes information is not through the software changing, but instead through the hardware changing. Digital may be efficient and simple, but it doesn't, and will never, take into account the incredibly complexity and amount of information that can fit into a tiny space.
What would we do next when the perfect A.I., which can and will imitate a human completely, emerges?
6:50 your intonation's a bit weird :P
Me: hello, define the difference between these 2 sentences.
1. Time flies like an arrow
2. Fruit flies like a banana.
ChatGPT: The first sentence, "Time flies like an arrow," is a metaphor that compares the passage of time to the motion of an arrow. The second sentence, "Fruit flies like a banana," is a statement about the behavior of a type of insect known as fruit flies, which are attracted to ripe bananas and other fruits. In this sentence, "like" is being used in the sense of "enjoy" or "prefer," rather than as a comparison.
chat gpt is smarter than me
This guy gets it
time flies like an arrow vs. fruit flies like a banana ... love it
“Any AI smart enough to pass a Turing test is smart enough to know to fail it.” - Ian McDonald
So, the machine performed the Divine Move from Go, but in chess?
And here we are with gpt 4
7:30 those aren't fruit flies.
The turing test? More like the enduring test
you should see the classified stuff they did in the 1970s with a.i.
i guess there is a big difference between AI and VI. AI are sentient beings with knowledge of their existence. VI (virtual intelligence) are just simulations that can act like a sentient being but not being one.
... Apparently I am a computer, I understood those two sentences like a computer would
I guess the argument against the Turing test is, if you make a machine that can pass the Turing test, then you've made a machine that's really good at passing the Turing test, and it says nothing about general intelligence.
i think he's talking about orch or at the end
It is so interesting that it seems the human brain isn't processing linearly but everyone assumes it is
Kasparov was strong enough to beat Deep Blue back in the days of the match, he just failed to do so for many reasons some of them unknown and suspicious...
Its today's computers that are like Professor Brailsford explained, nobody has a chance against them not even chess champions.
I don't think brute force is the correct name for what Deepblue did, that kind of makes it look no better than a permutation based algorithm.
So I guess we're going have to wait for quantum computers to see if a machine can be ironic or sarcastic...
he beat him along time ago not beat him off along time ago
Fruit flies like a banana
I chortled on all levels
will be sing this video for my AI class thanks!