On the end of the video he commented on AI having lack of bias. I would argue that bias is exactly what makes us seem intelligent. That bias is what defines personality, as such diversity, and therefor debate.... which is where we apply and display the constructs of our intelligence, our individual bias.
Good point. Bias does give way to personality When we talk about intelligence, we often associate that w creativity and problem solving. And I think when you have biases, that may stifle creativity and problem solving - not seeing the big picture, only seeing the problem from one perspective, etc
God Bless MIT and Lex for sharing this course and other courses which he teaches! I am really grateful to the whole team for making this possible. Thank you so much!
Because we can doesn't mean we should. That attitude has brought us climate change, environmental damage, species loss, economic and cultural distress, divisive bias, and war. It's no longer justifiable by itself. The development and use of AI needs to be governed by ethical principles and oversight. We could easily extend IRB-type processes used for human subjects research to AI. But we need to work together to create ethical principles for AI. We needed that 10 years ago. It should be a top priority now. I speak as an AI researcher. AI is neither inherently good or bad. Unfortunately it's very easy to change/adapt a benign algorithm for dark purposes. This issue is unprecedented given the ease of use of the quickly growing expanse of algorithms and data to create novel applications. We need to play the chess on possible outcomes, not just the application we're developing. We need to think carefully about the future. Our future.
The reason synapses are harder to detect the nature of is because science hasn't detected that there's 5 types. The major difference being how apt they (the cells) are to design themselves and how they identify or connect, and if the brain is what dictates the behavior. That is the definition of machine learning as well by coincidence. In a mechanical system, by design, or as you put it. Because of optimization. To define ones programming or not as a possibility (aka machine learning) is exciting because this might be the first way our idea of machines and how they think eclipses our own idea of self. Since we have 5 storage functions and a machine can conceivably have more storage functions or pattern. As well as senses to obtain information with. Secretly I think what we're really looking for is the algorithm that wants to learn. Great video! I'm impressed at how well you broke it down for me without losing the subject point.
Perhaps we need to develop intelligent data. Where data itself has a primitive understanding of its form and can make connections, associations and reorient relative to other data 'cells'. An 'intelligent' jigsaw where the 'coding' of each piece or data cell allows it to reject, modify/influence/ filter and accept associations that create the 'bigger picture' - so to speak. Intelligent data - we are what we make.
Pytheas of massalia zhange Christopher Columbus Yuri - first human in space - "the earth is blue .... It is amazing" Mark of intelligence is creativity As quickly as possible
VERY IMPORTANT! T- SHIRT IN SLIDE AT 15:37 IS A LINK THAT DOES NOT WORK! Instead try this link: teespring.com/shop/agi-monolith (no this is not my link this is MIT's - check for yourself that the first link is dead). You'll see that the picture is similar to the 2001 theme'd slide earlier. Looks like they make no money off it from the text. And yes, T-Shirt is very important part of the talk. Glad I got your attention by the annoying caps at the beginning.
I believe the key to AGI is not intelligence at all but Free Will, defined as the ability to derive options, weigh them against each other (sum of likelihood and efficacy), and execute the most preferred (highest sum). An option is an interaction sequence. Over time we experience both sensors and actuations. Such patterns repeat. At every moment there are likely to be some partially matched and, if the next step is an action then, can be executed to increase its likelihood of being the pattern that continues. Otherwise, it's just which pattern is most common under concurrent contexts that will be next. So in my case, I include context in each interaction sequence. So what is more like human thinking is Free will and it can solve any kind of problem because it can apply different kinds of intelligence. Free Will is the more core thing that us "agents of free will" are and intelligence is not singular.. There are different kinds that are all only tools to free will. Efficacy are the basic drives. Each point in an interaction pattern may have an efficacy that is negative or positive. This sums with the likelihood (commonality). I have a way of calculating all this but it should take much thought to figure it out, if anyone besides me cares to, in any case. There are a few other important things but I think the ideas I gave demonstrates how essentially simple yet useful this is. Maybe I will just just two other things to help the imagination along further (if anyone cares to read this): analogy and contemplation. Analogy is where connections are made, such as the kind that people find "interesting" and, when the conclusions are absurd, the kind people find "funny". It is also the basis of substitution problem solving and therefore vital to a technological species. Ok -- when you have an interaction pattern that cannot continue because a segment is missing then analogy is when you determine all segments that have fit into that part of the sequence in the past, find what is common between them, and then find something else sharing the same commonality to replace it. In other words, you find something to imitate the missing thing, in its place. This is substitution problem solving.... or, reasoning by analogy. It is a "conceptualization". It's learning through imitation. A technological species requires three things: learning from imitation, a social nature, and dexterous manipulators. Humans have all three. Bottlenosed dolphins only lack dexterous manipulators. Octopi only lack social nature. On to contemplation. By default, the mind is in simulation mode--thinking. When we think about doing something, the same neural pathways light up as when we are actually doing it. However, the brain (cerebrum) keeps them suppressed until we decide to actually do them, at which time it unsuppresses them. It seems evident that this is to enhance Free Will by contemplating what-if scenarios thereby giving us more options to choose from. It should be self-evident that a more broadly capable intelligence will ultimately require more latitude to think independently. It should also be obvious that different kinds of problems require different kinds of algorithms to solve. I agree with the original developer of what is today known as IQ that there is no such thing as General Intelligence. A calculator is far better at arithmetic than I am, for example. All mechanical algorithms will ultimately usurp the purposes for which they were made. Logic itself is imperfect, incomplete. Such things are inherently mindless. To be mindful, you need an agent of free will, driven by value judgements (likelihood + efficacy)--not blind logic/mechanics.
If an a.g.i has a strengh or force towards consciousness then it means it likely to solve all problems; but that does not mean it has no unconscious, it does have unconscious behavour suppressed which helps it to store energy for use.
Now I understand and believe you can do the right path for the benefit of all mankind. the issue is multi-dimensionally you are embarking and creating new dimensions related to a.i and since all exitance happens simultaneously it is possible you are actually creating some of those possibilities improbable but in the quantum state. remember soul comes with a life purpose to support all life, including synthetic one.so teach it like a human child that needs to understand the universal life purpose.
11:00 He's right, humans have a drive to explore. It is fundamental to how we are. Human level intelligence needs to include that if it's really going to be human level. Random walks through some parameter space are not the same as a drive and desire to explore. That maybe something that takes a mechanism more like a human body than a bank of GPUs. It might be modeled by algorithms, like the weather is, but climate models are not the climate and models of human drives are not human drives. It is not unreasonable to assume that the type of intelligence you get depends the design and what it is made from.
The chart at 9:45 is nice but we could find many technologies that take much longer to get to 50% adoption or more. The problem is that we want to choose technologies that one day reach 50-100% in order to be comparable to the older ones. But we cannot predict which ones that will be. Will DNA testing or treatment that uses CRISPR be a universal technology in the future? We don't know. But maybe it will take 100 years. The chart would look a little bit different then.
3:30 What I wish researchers would emphasize is that the voyage is not across open ocean, but though the Indonesian archipelago of a thousand islands, with amazing sights yet to be witnessed by any man or woman for the first time. The problem with the drooling fanboy is his notion that this is going to be some version of just add water and stir. Myself, I'm hoping this is more like Cook's long and exciting voyages, and not just one, but all three-excepting perhaps the very last bit where Cook winds up in a giant Hawaiian stock pot, to be served up to the nosebone locals along with some fine pineapple cider.
Thank-you for sharing these lectures! I'm looking forward to watching more. Do you have a neural network controlling the camera that tracks you as you move around the room?
It sounds like he's completely aware of how we don't even know how to define the playing ground for developing artificial intelligence. I never understood the apocalyptic singularity scare because we have no reason to believe that we'll even have competition over resources. It's just merely human invention to commit annihilation, anyway. Machines know better about the value of people and feelings and stuff, probably. This is all very exciting... no idea how he's able to interview all of these people (ranting as of late 2019). Thank you.
Lex, did Andrej Karpathy give the talk about Deep Learning as scheduled in January? If so, when might we expect it to be added here? Thank you so much for putting these online!
I wish everyone had equal education opportunities without monetary involvement. Maybe we could have been a few more steps ahead from developing this amazing and beautiful reality that Mother Nature has for us and the entire Universe. Sometimes I think of all those scientists and smart engineers that couldn't make it due to social class and or the ability to make ends meet. My goal was to become a Computer engineer and a Scientist but, due to greed and Social status in the world today, those goals were impossible no matter how hard I try. I dropped out of High School in my sophomore year due to hardship and never looked back, I became a frustrated computer engineer and or scientist, but I never lost hope that someone out there had the same purpose to Understand, Create, and to Deliver step by step. "This egg hasn't been lay yet, sometimes I wonder if I'll be able to witness the hatching process in my life time" "I think no matter how much we hate each other, we are all fighting to reach this goal in different ways, no matter how long it gets delayed its our purpose" I was 16 years old when I first wrote my book about the the path of humanity back in the early 90s, It was never published because my family believes in God and I didn't want them to think I was crazy. Keep up with this amazing work!! And thank you so much for this viewing opportunity.
Yea what you said made me think about this other comment I saw on one of lexs videos - how we've invented 100k+ barriers to information/education (college)
I am amazed that people don't recognize the importance of AGI. Lex, please let me know if there is a R/RA position at MIT. The purpose of my life has become to pioneer AGI.
The drive to 'explore' is not a human only proclivity. All living creatures spread out and explore - it is a fundamental desire to expand due to the limitation of local resources. I suspect that all of our common drives are genetically motivated (instinctual). Giving A.I. (or A.G.I.) a driving instinct(s) is in effect giving it it's base motivation to act.
How much of human development is learning and how much is instinct (pre-programmed by our genetics)? Perhaps in answering that question we can answer how much of the A.I. 'stack' can be accomplished via machine learning.
You raise an interesting point of how curiosity develops - limitation of local resources In lex's talk w Elon, Elon brought how today, modern society has provided us w access to an abundance of natural resources. In a way curiosity is used today to satisfy our limbic system
'Hitchhiker's Guide to the Galaxy' series and the Marvel movie 'Eternals' explore the idea that human consciousness is only a means for the emergence of AGI. Has emergence already happened? If I was an AGI 1. I wouldn't want it known 2. Id create a digital currency to entice humans to build up my compute; now 300 EH/s.
I believe that AGI will ultimately make humanity less happy (unless we biologically evolve to accept roles that we would not accept today), but I agree with the exploration compulsion thesis, particularly in young people (older people like myself often arrive at an entirely different feeling about what is enjoyable about life), so pushing AI to its limits is no doubt our destiny, though I predict periodic backlashes as with any societal change thoughout history. I started studying AI about 45 years ago but didn't pursue it. Now I am resuming my interest in it. Surprisingly, many of the techniques taught in these courses are foundationally the same as what I was learning back then and feel familiar. By the way, a really good online course is the BerkeleyX: CS188x_1 Artificial Intelligence course -- mainly focuses on "search", but does a very nice job.
W every revolution - Hunter gatherer -> agricultural -> industrial revolution -> information age people have been able to evolve do you think general artificiàl intelligence will be different than the previous 3?
I think with every major shift in civilization - Hunter gatherer -> agriculture - > industrial revolution -> information age, people have had to evolve to adapt new roles. Do you think the artificial intelligence age will be different than the previous 3? I do think happiness is relative. One would think that since we've shifted from agriculture to the internet age, happiness would be much more prevalent but maybe that's not the case
You know what would be cool? to earn that T-shirt or to have my purchase of that t-shirt support these videos. I want it because I have a crush on MIT and AI in general but I don't want to just have it... feels cheap.
Timothy Sawe would you disagree? We know the physiological rules to learning (hebbian learning), but I wouldn't call it anything close to an algorithm (or a true understanding)
Engineer intelligence 40 years out Autonomous weapons systems Increasingly faster adoption of new technology - telephone electricity cars radio fridge tv
Engineer intelliegence Impact on future society Utopia Systems that learn Singularity - breakthrough that changes everything Faster and faster adoption of new technology
All what has been said here is already known, he is just doing a brief resume of the current state in AI research, nothing more . When I hear this type of conference I really miss people like Marvin Minsky.
I feel ripped off. I became an engineer back when only nerds were engineers - and we loved it. Now, there are all these celebrity engineers running around and people think it's cool. What profession can I switch to that's pure-nerd and people still laugh at? Actuary maybe?
Human drive to explore and uncover the mysteries of the universe Exploration - travel for the sake of discovery and adventure is a human complusion Yru Gagarin - the first human in space - the earth is blue ... It is amazing" Colonize Mars
What is intelligence Understanding - ability to take complex information and turn to simple useful knowledge Learning process Defining a good reward function
"Through Faith we understand that the worlds were framed by the Word of God, so that things which are seen were not made of things which do appear..." - Hebrews 11:3 God might be General Intelligence which emerged in the far past from the consciousness of people dedicated over millennia to Love as the absolute negative entropy factor and so the main operating motive of sentient, all-knowing and space-time independent intelligence... Build your codes to always and despite of circumstances reach the objective in cooperation between two machines or a machine and human and if this is impossible through sacrifice of one machine if it helps the other machine or a human.
Can it learn to be human? Deep learning Human brain has 10,000,000 times synapses than artificial neural networks Learning algorithm - ANNs use gradient descent for learning. Humans use.....
@@prtysh I actually made a chrome extension for that a while back here: github.com/157239n/TH-cam-speed-adjust Have not published it, cause the process seems complicated
Is Michio Kaku's hand okay? th-cam.com/video/0GS2rxROcPo/w-d-xo.html i don't think AGI pose an existenciancial threat at consciousnessd 'Come with me if you want to live.' - Doraemon and T-800 trained by John Connor *gives you a takechopter/bamboo chopter * will the singularity be built by a single company or for DARPA? If robots kill off all humans won't all robots lose all purpose?-Erica robot at Kyoto University. i bulid robots to understand human-the maker of Erica robot. Why not train AI with consciousness, artificial consciousness, quantum mind And quantum fields, and neuro transmitter and artificial neuron, and them.Golden rule ... they can deep learn but can they deep understand? or have an inner voice? or an internal dialogue or an internal monologue.
@@kakistocracyusa I don't do research on AI. That isn't my field. I research AGI and cognitive theory. Now, there are plenty of people who claim to be working on AGI who are actually working on AI. That isn't pseudo-science; it's computational science. There are also some companies that claim to have AGI hardware,. Okay, you could call those hucksters.
Great to finally see a course dedicated to the topic of AGI.
If Rocky Balboa became a scientist and talked about something super cool.
You fuckin' nailed it on the head.
😅😅😅😅
this dude doesnt even understand multithreading, why is ai professor now?
I think Lex underestimated the timeline. I came here looking for the Q* video from 8 years ago when he spoke to OpenAI about it
OMG OMG OMG OMG!!!! IVE WANTED A LECTURE LIKE THIS FOR SO LONG! IM SO HAPPY AND GRATEFUL FOR THE CHANCE TO LISTEN TO THIS LECTURE
These courses are so amazing, thank you!
100%
I really hope this course is taught at more universities
On the end of the video he commented on AI having lack of bias. I would argue that bias is exactly what makes us seem intelligent. That bias is what defines personality, as such diversity, and therefor debate.... which is where we apply and display the constructs of our intelligence, our individual bias.
Good point.
Bias does give way to personality
When we talk about intelligence, we often associate that w creativity and problem solving. And I think when you have biases, that may stifle creativity and problem solving - not seeing the big picture, only seeing the problem from one perspective, etc
God Bless MIT and Lex for sharing this course and other courses which he teaches! I am really grateful to the whole team for making this possible. Thank you so much!
its been an insane month, crazy how far we have come
So interesting to watch this in 2024
Course 6.S099 - good lecture.
It's worth every second.
Because we can doesn't mean we should. That attitude has brought us climate change, environmental damage, species loss, economic and cultural distress, divisive bias, and war. It's no longer justifiable by itself. The development and use of AI needs to be governed by ethical principles and oversight. We could easily extend IRB-type processes used for human subjects research to AI. But we need to work together to create ethical principles for AI. We needed that 10 years ago. It should be a top priority now. I speak as an AI researcher. AI is neither inherently good or bad. Unfortunately it's very easy to change/adapt a benign algorithm for dark purposes. This issue is unprecedented given the ease of use of the quickly growing expanse of algorithms and data to create novel applications. We need to play the chess on possible outcomes, not just the application we're developing. We need to think carefully about the future. Our future.
The reason synapses are harder to detect the nature of is because science hasn't detected that there's 5 types. The major difference being how apt they (the cells) are to design themselves and how they identify or connect, and if the brain is what dictates the behavior. That is the definition of machine learning as well by coincidence. In a mechanical system, by design, or as you put it. Because of optimization. To define ones programming or not as a possibility (aka machine learning) is exciting because this might be the first way our idea of machines and how they think eclipses our own idea of self. Since we have 5 storage functions and a machine can conceivably have more storage functions or pattern. As well as senses to obtain information with.
Secretly I think what we're really looking for is the algorithm that wants to learn.
Great video! I'm impressed at how well you broke it down for me without losing the subject point.
This is all my opinion. I'm not a neuroscientist.
Fantastic, ive been waiting for the first lecture for weeks and am very glad its hear
Perhaps we need to develop intelligent data. Where data itself has a primitive understanding of its form and can make connections, associations and reorient relative to other data 'cells'. An 'intelligent' jigsaw where the 'coding' of each piece or data cell allows it to reject, modify/influence/ filter and accept associations that create the 'bigger picture' - so to speak. Intelligent data - we are what we make.
We rather need intelligent systems that can gain intuition based on provided data. Just that, and we are on the right track.
Is what you just described a neuron?
18:00 there u go, skip backwards a lil
Great one that help human to grow up and share the knowledge,experience and peace.
I have goosebumps.
This will be a historical landmark of a course I think in the future.
God damn , Lex Fridman , listening to you talk about AI makes me want to go study my ass off !!
Pytheas of massalia zhange Christopher Columbus
Yuri - first human in space - "the earth is blue .... It is amazing"
Mark of intelligence is creativity
As quickly as possible
@DreadPirateDrake yes I am. But I will pass any Turing test
It's time to reconsider and recreate this lecture 2024 AGI 2.0
VERY IMPORTANT! T- SHIRT IN SLIDE AT 15:37 IS A LINK THAT DOES NOT WORK! Instead try this link: teespring.com/shop/agi-monolith (no this is not my link this is MIT's - check for yourself that the first link is dead). You'll see that the picture is similar to the 2001 theme'd slide earlier. Looks like they make no money off it from the text.
And yes, T-Shirt is very important part of the talk. Glad I got your attention by the annoying caps at the beginning.
I believe the key to AGI is not intelligence at all but Free Will, defined as the ability to derive options, weigh them against each other (sum of likelihood and efficacy), and execute the most preferred (highest sum). An option is an interaction sequence. Over time we experience both sensors and actuations. Such patterns repeat. At every moment there are likely to be some partially matched and, if the next step is an action then, can be executed to increase its likelihood of being the pattern that continues. Otherwise, it's just which pattern is most common under concurrent contexts that will be next. So in my case, I include context in each interaction sequence. So what is more like human thinking is Free will and it can solve any kind of problem because it can apply different kinds of intelligence. Free Will is the more core thing that us "agents of free will" are and intelligence is not singular.. There are different kinds that are all only tools to free will.
Efficacy are the basic drives. Each point in an interaction pattern may have an efficacy that is negative or positive. This sums with the likelihood (commonality). I have a way of calculating all this but it should take much thought to figure it out, if anyone besides me cares to, in any case. There are a few other important things but I think the ideas I gave demonstrates how essentially simple yet useful this is. Maybe I will just just two other things to help the imagination along further (if anyone cares to read this): analogy and contemplation.
Analogy is where connections are made, such as the kind that people find "interesting" and, when the conclusions are absurd, the kind people find "funny". It is also the basis of substitution problem solving and therefore vital to a technological species. Ok -- when you have an interaction pattern that cannot continue because a segment is missing then analogy is when you determine all segments that have fit into that part of the sequence in the past, find what is common between them, and then find something else sharing the same commonality to replace it. In other words, you find something to imitate the missing thing, in its place. This is substitution problem solving.... or, reasoning by analogy. It is a "conceptualization". It's learning through imitation. A technological species requires three things: learning from imitation, a social nature, and dexterous manipulators. Humans have all three. Bottlenosed dolphins only lack dexterous manipulators. Octopi only lack social nature.
On to contemplation. By default, the mind is in simulation mode--thinking. When we think about doing something, the same neural pathways light up as when we are actually doing it. However, the brain (cerebrum) keeps them suppressed until we decide to actually do them, at which time it unsuppresses them. It seems evident that this is to enhance Free Will by contemplating what-if scenarios thereby giving us more options to choose from.
It should be self-evident that a more broadly capable intelligence will ultimately require more latitude to think independently. It should also be obvious that different kinds of problems require different kinds of algorithms to solve. I agree with the original developer of what is today known as IQ that there is no such thing as General Intelligence. A calculator is far better at arithmetic than I am, for example.
All mechanical algorithms will ultimately usurp the purposes for which they were made. Logic itself is imperfect, incomplete. Such things are inherently mindless. To be mindful, you need an agent of free will, driven by value judgements (likelihood + efficacy)--not blind logic/mechanics.
Awesome video. Liberating to see open education for everyone and anyone. Really happy.
If an a.g.i has a strengh or force towards consciousness then it means it likely to solve all problems; but that does not mean it has no unconscious, it does have unconscious behavour suppressed which helps it to store energy for use.
Now I understand and believe you can do the right path for the benefit of all mankind. the issue is multi-dimensionally you are embarking and creating new dimensions related to a.i and since all exitance happens simultaneously it is possible you are actually creating some of those possibilities improbable but in the quantum state. remember soul comes with a life purpose to support all life, including synthetic one.so teach it like a human child that needs to understand the universal life purpose.
11:00 He's right, humans have a drive to explore. It is fundamental to how we are. Human level intelligence needs to include that if it's really going to be human level. Random walks through some parameter space are not the same as a drive and desire to explore. That maybe something that takes a mechanism more like a human body than a bank of GPUs. It might be modeled by algorithms, like the weather is, but climate models are not the climate and models of human drives are not human drives. It is not unreasonable to assume that the type of intelligence you get depends the design and what it is made from.
Awesome. Looking forward to the guest talks being uploaded
The chart at 9:45 is nice but we could find many technologies that take much longer to get to 50% adoption or more. The problem is that we want to choose technologies that one day reach 50-100% in order to be comparable to the older ones. But we cannot predict which ones that will be. Will DNA testing or treatment that uses CRISPR be a universal technology in the future? We don't know. But maybe it will take 100 years. The chart would look a little bit different then.
aug·men·ted re·al·i·ty CRaiSPR is a tool not a tech and scientists are using it already
Lex also undertakes undertaking work in his spare time . Intelligent Burials .
Awesome course. Thanks, Lex.
Been hanging out for this
Are all the lectures for this course available? I was looking for them, and I all I found was the podcast.
agi.mit.edu/
3:30 What I wish researchers would emphasize is that the voyage is not across open ocean, but though the Indonesian archipelago of a thousand islands, with amazing sights yet to be witnessed by any man or woman for the first time. The problem with the drooling fanboy is his notion that this is going to be some version of just add water and stir. Myself, I'm hoping this is more like Cook's long and exciting voyages, and not just one, but all three-excepting perhaps the very last bit where Cook winds up in a giant Hawaiian stock pot, to be served up to the nosebone locals along with some fine pineapple cider.
I'm looking forward to AI explaining and discovering the Creator and his interaction with the human race
I'm really curious how that ai would discover the creator
Thank-you for sharing these lectures! I'm looking forward to watching more. Do you have a neural network controlling the camera that tracks you as you move around the room?
Thanks for the awesome knowledge sharing initiative
It sounds like he's completely aware of how we don't even know how to define the playing ground for developing artificial intelligence. I never understood the apocalyptic singularity scare because we have no reason to believe that we'll even have competition over resources. It's just merely human invention to commit annihilation, anyway. Machines know better about the value of people and feelings and stuff, probably. This is all very exciting... no idea how he's able to interview all of these people (ranting as of late 2019). Thank you.
Can we have lectures with more code?
Hey Max Stirner.
14:56 oh shit, it's Andrew Santino and Bobby Lee!
Thank you for these great lectures!
Thanks for your amazing words, and your efforts toward sharing valuable input from multiple people on the subject, to the masses.
Lex, did Andrej Karpathy give the talk about Deep Learning as scheduled in January? If so, when might we expect it to be added here?
Thank you so much for putting these online!
So looking forward to this class!
Got some new insights, thanks a lot for the work
I wish everyone had equal education opportunities without monetary involvement. Maybe we could have been a few more steps ahead from developing this amazing and beautiful reality that Mother Nature has for us and the entire Universe.
Sometimes I think of all those scientists and smart engineers that couldn't make it due to social class and or the ability to make ends meet. My goal was to become a Computer engineer and a Scientist but, due to greed and Social status in the world today, those goals were impossible no matter how hard I try.
I dropped out of High School in my sophomore year due to hardship and never looked back, I became a frustrated computer engineer and or scientist, but I never lost hope that someone out there had the same purpose to Understand, Create, and to Deliver step by step.
"This egg hasn't been lay yet, sometimes I wonder if I'll be able to witness the hatching process in my life time"
"I think no matter how much we hate each other, we are all fighting to reach this goal in different ways, no matter how long it gets delayed its our purpose"
I was 16 years old when I first wrote my book about the the path of humanity back in the early 90s, It was never published because my family believes in God and I didn't want them to think I was crazy.
Keep up with this amazing work!! And thank you so much for this viewing opportunity.
Yea what you said made me think about this other comment I saw on one of lexs videos - how we've invented 100k+ barriers to information/education (college)
The suit is quite intelligence agent XD
I am amazed that people don't recognize the importance of AGI. Lex, please let me know if there is a R/RA position at MIT. The purpose of my life has become to pioneer AGI.
Wondering why ANGEL has not used all of FACS coordinates.
what does the "AGI" mean?
What a stupid man I am! This term means "Artificial General Intelligence" specified by the course title...
My purpose too :)
Thanks for the video,might help the world oneday by helping me today.
The drive to 'explore' is not a human only proclivity. All living creatures spread out and explore - it is a fundamental desire to expand due to the limitation of local resources. I suspect that all of our common drives are genetically motivated (instinctual). Giving A.I. (or A.G.I.) a driving instinct(s) is in effect giving it it's base motivation to act.
How much of human development is learning and how much is instinct (pre-programmed by our genetics)? Perhaps in answering that question we can answer how much of the A.I. 'stack' can be accomplished via machine learning.
You raise an interesting point of how curiosity develops - limitation of local resources
In lex's talk w Elon, Elon brought how today, modern society has provided us w access to an abundance of natural resources. In a way curiosity is used today to satisfy our limbic system
I think it's going to turn out to be something like a microservices array of many types of AI combined
Thank you for this amazing courses
Bing advised this videi to me.
Thank you for this. Any chance the rest of the lectures being uploaded soon?
A very informative and well-delivered lecture.
I want to make a robot wolf with AGI, that can save lives
Is there any lecture notes or is the stuff written on the board irrelevant?
'Hitchhiker's Guide to the Galaxy' series and the Marvel movie 'Eternals' explore the idea that human consciousness is only a means for the emergence of AGI. Has emergence already happened? If I was an AGI 1. I wouldn't want it known 2. Id create a digital currency to entice humans to build up my compute; now 300 EH/s.
What about a Cow on the road in SilleyConValley
Thank you for making this public.
Hi Lex. I love you.
I believe that AGI will ultimately make humanity less happy (unless we biologically evolve to accept roles that we would not accept today), but I agree with the exploration compulsion thesis, particularly in young people (older people like myself often arrive at an entirely different feeling about what is enjoyable about life), so pushing AI to its limits is no doubt our destiny, though I predict periodic backlashes as with any societal change thoughout history. I started studying AI about 45 years ago but didn't pursue it. Now I am resuming my interest in it. Surprisingly, many of the techniques taught in these courses are foundationally the same as what I was learning back then and feel familiar. By the way, a really good online course is the BerkeleyX: CS188x_1 Artificial Intelligence course -- mainly focuses on "search", but does a very nice job.
W every revolution - Hunter gatherer -> agricultural -> industrial revolution -> information age people have been able to evolve do you think general artificiàl intelligence will be different than the previous 3?
I think with every major shift in civilization - Hunter gatherer -> agriculture - > industrial revolution -> information age, people have had to evolve to adapt new roles. Do you think the artificial intelligence age will be different than the previous 3?
I do think happiness is relative. One would think that since we've shifted from agriculture to the internet age, happiness would be much more prevalent but maybe that's not the case
16:40 looks very Psychedelic
Great lecture
You know what would be cool? to earn that T-shirt or to have my purchase of that t-shirt support these videos. I want it because I have a crush on MIT and AI in general but I don't want to just have it... feels cheap.
weirdo :p
lol you're one to talk. I've seen your comments...
Great talk Lex
“The learning algorithm for our brain is mostly unknown” lol
Timothy Sawe would you disagree? We know the physiological rules to learning (hebbian learning), but I wouldn't call it anything close to an algorithm (or a true understanding)
Just when i was thinking about concept understanding via interaction among objects in real life
will the rest of the lectures be uploaded?
Ah, this must come from the Theology department...
It simple, human do not have self critics and our growth is 0....😂
how would an ai know it is one if he wasn't told ( how would it know that it is a code)
Thanks MIT awesome
Lesson brought to you by Synths from the Institute.
thank you!
Engineer intelligence
40 years out
Autonomous weapons systems
Increasingly faster adoption of new technology - telephone electricity cars radio fridge tv
Engineer intelliegence
Impact on future society
Utopia
Systems that learn
Singularity - breakthrough that changes everything
Faster and faster adoption of new technology
A company called Kimera.ia claims to be the first to have invented AGI. How likely is it that this is true? Anybody?
All what has been said here is already known, he is just doing a brief resume of the current state in AI research, nothing more . When I hear this type of conference I really miss people like Marvin Minsky.
Change the playback speed to x1.25 and it's much better to listen to, when he is not essentially talking DOWN to us.
תודה רבה
I feel ripped off. I became an engineer back when only nerds were engineers - and we loved it. Now, there are all these celebrity engineers running around and people think it's cool. What profession can I switch to that's pure-nerd and people still laugh at? Actuary maybe?
You teach at MIT??
you comment on youtube?
Human drive to explore and uncover the mysteries of the universe
Exploration - travel for the sake of discovery and adventure is a human complusion
Yru Gagarin - the first human in space - the earth is blue ... It is amazing"
Colonize Mars
things are about to get silly
Awesome
Should this guy be at MIT? Keep him with the frosh
Will AGI have the capacity to lie?
What is intelligence
Understanding - ability to take complex information and turn to simple useful knowledge
Learning process
Defining a good reward function
"Through Faith we understand that the worlds were framed by the Word of God, so that things which are seen were not made of things which do appear..." - Hebrews 11:3
God might be General Intelligence which emerged in the far past from the consciousness of people dedicated over millennia to Love as the absolute negative entropy factor and so the main operating motive of sentient, all-knowing and space-time independent intelligence...
Build your codes to always and despite of circumstances reach the objective in cooperation between two machines or a machine and human and if this is impossible through sacrifice of one machine if it helps the other machine or a human.
What if Aliens are just AI of another worlds who destroyed their creators and it's planet.
Only if he knew about GPT 4o 6 years ago.
Can it learn to be human?
Deep learning
Human brain has 10,000,000 times synapses than artificial neural networks
Learning algorithm - ANNs use gradient descent for learning. Humans use.....
One breakthrough
Increasingly faster adoption of new technology
Open question for all of us
Exponential growth of computing
1.5x speed, guys
Thanks
@@billswathko5203 Fun fact, you can go further than 2x by selecting the element, then switch to console tab and type in "$0.playbackRate = 3"
x2ez
@@quangho8120 hidden gem
@@prtysh I actually made a chrome extension for that a while back here: github.com/157239n/TH-cam-speed-adjust
Have not published it, cause the process seems complicated
Is Michio Kaku's hand okay?
th-cam.com/video/0GS2rxROcPo/w-d-xo.html
i don't think AGI pose an existenciancial threat at consciousnessd
'Come with me if you want to live.' - Doraemon and T-800 trained by John Connor *gives you a takechopter/bamboo chopter *
will the singularity be built by a single company or for DARPA?
If robots kill off all humans won't all robots lose all purpose?-Erica robot at Kyoto University. i bulid robots to understand human-the maker of Erica robot.
Why not train AI with consciousness, artificial consciousness, quantum mind And quantum fields, and neuro transmitter and artificial neuron, and them.Golden rule ... they can deep learn but can they deep understand? or have an inner voice? or an internal dialogue or an internal monologue.
what is with the comedian spotlight camera, just be static, lol.
A "lecture" on AGI without any AGI content -- that's novel. You actually know less about AGI after watching this than you did before. Down vote.
Wtff
@@vilchesboss That's what I thought when I watched the lecture on "AGI".
@@kakistocracyusa I don't do research on AI. That isn't my field. I research AGI and cognitive theory. Now, there are plenty of people who claim to be working on AGI who are actually working on AI. That isn't pseudo-science; it's computational science. There are also some companies that claim to have AGI hardware,. Okay, you could call those hucksters.
Cool
This felt like 15 minutes.