I really enjoyed this conversation with Daniel. Here's the outline: 0:00 - Introduction 2:36 - Lessons about human behavior from WWII 8:19 - System 1 and system 2: thinking fast and slow 15:17 - Deep learning 30:01 - How hard is autonomous driving? 35:59 - Explainability in AI and humans 40:08 - Experiencing self and the remembering self 51:58 - Man's Search for Meaning by Viktor Frankl 54:46 - How much of human behavior can we study in the lab? 57:57 - Collaboration 1:01:09 - Replication crisis in psychology 1:09:28 - Disagreements and controversies in psychology 1:13:01 - Test for AGI 1:16:17 - Meaning of life
Hi Lex, i have a wild theorie, whe're i would like to think you about. There are 2 Moments in the dat where you deel to experience reality. During the dat, when you are not sleeping and during the night when you are deeaming. The interesting one ofcourse are dreams. When you dreams, you bever think this is a dreams, you think what you are deeaming is real. It's only after you wake up that you think, and you Remember you're dreams, that you think it's only a dreams, it's not reality. But what if deeaming is more then just trying to recove from a Busy day. What if it's alsof a sort of warning system. I have some mental issues, what causes let's say that im sometimes rational , my old self, but sometimes, emotional , not my old self. I have sometimes, mostly orientee dreams, where i meant and kiss a girlfriend. And the wierd thing is , i then meer her that dat it the following dat/weeks and i light or light not kiss her.Like our conscious/nature Will give us a sneak leek into a possible Future through our dreams. As if 95% of our Future is allready dererminee , but there is still a chance to changer it. Would this be a practical example of hedelbergs u certainty principle/the theory of many world's. Ithe sneak peel actually helper me ones, how to het out of a difficile situation.
I was thinking the same thing. How is he getting all these great guests right off the bat? I imagine the MIT cred must help some.... Also, Lex has great questions and interview style. Congrats to Lex on a great podcast with great guests.
For those of you that don’t know a lot about Daniel is that he never lets sunk cost fallacy get the best of him. Working with him on his book, he would delete and start from scratch without thinking about the effort he put in writing that section at first place. And it’s been a feature of him in his research life, he would start a research and get deep into it but then he would just start something new and not stick to it just coz he had spent years doing it. That’s one of the things I learned from Daniel and since I learned that I stop reading books anywhere I feel like its getting boring or its not worth continuing, I go out of theatre if I dont like the movie after 10-15 minutes, I close down business ideas if they don’t show results within a time frame. It has changed my life. One last thing I learned from doing my research and work was that you should always have an end in mind, as to when you will stop doing something if you don’t get the results you were expecting, it helps you overcome sunk cost fallacy a lot easier. Thank you so much Daniel for your work with Amos to help us humans make better decisions!
"I don't know enough of philosophy to answer that." That is the mark a true thinker: someone who knows they don't know and won't proffer half baked ideas as facts. Amazing. 10/10
I was struck how well informed Daniel Kahneman is about current progress in AI, Chess, and so much more. I want to still be that sharp in 35 years from now.
"What makes the experiencing self happy and what makes the remembering self happy are different things." this is the most profound thing I have heard in some time. another phenomenal podcast, thank you Lex!
I just want to let you know, I really enjoy listening to your podcast. Your questions feel like they come from the desire to learn and know more. Thanks for making this!
Daniel I'm so grateful for your book. Encountering it is one of the luckiest things in my life. I'm looking forward to your new book. I hope you are always in good health. Great episode. Thank you Lex.
Awesome guest again Lex! Always keep the big questions in the interviews; it's inspiring to see the variety of outlooks on life held by the great minds of our time.
I read his book "Thinking, Fast and Slow" a couple of times and took many notes, but I must say watching him explain his thoughts in a different manner and how he arrived to his conclusions is bringing even more depth and wisdom to an already amazing work. Thank you Lex Fridman for your always apt and interesting questions and rigorous preparation in advance to pull out the most value out of your stellar and exemplary guest-list.
The guests Lex both wants to talk with and is able to get on his show says a lot about the level of depth and intelligence Lex has. What a brilliant guest..I could have listened to this conversation for hours! When Daniel Kahneman said: "What makes the experiencing self happy and the remembering self happy, are 2 different things" What a doorway of thought that subject opens.
This interview with Daniel Kahneman is THE BEST youtube interview I've ever seen, not just about the topic of AI or philosophy of mind and consciousness, but hands down in terms of being interesting and informative and utterly relevant to my thinking (and scholarly projects at the moment). Thank you, Lex! I applaud your work and efforts with these video interviews.
The most profound line for me was regarding how people no longer need to remember things. This shook me and made a chill run down my spine because my immediate thought was, those in control of the knowledge create the narrative.
I don't understand by your saying"those in control of the knowledge". What I took away was that you don't need the "truth" to make up a fictional narrative.
Clearly you have not read '1984', you should, everyone should: "Who controls the past controls the future: who controls the present controls the past." George Orwell (1948)
(I’m a nobody to challenge this man but...) I’m not sure it’s right to say that we no longer need to remember facts. Surely knowledge accumulated over a long time is the foundation of wisdom - like the wisdom that Kahneman talks about in following his gut with research. You can’t Google that.
Lex is such an amazing interviewer. I'm always impressed by the quality of his questions and the obvious respect these incredibly accomplished guests have for him and his intellect. Top notch.
57:58 wow (- I like how Kahneman blinks affirmingly after each statement). - Lex is asking the best questions, and is always ready to receive more information than was sent. - I just wished it “clicked” more often with his guests. (He also has the best guests, as numerous commentators already mentioned). - I’m so glad to have “found” this channel.
Lex, Thank you for these insightful conversations. You are the Terry Gross of AI interviews for depth and quality. Mention of Yoshua Bengio's work was perhaps missed inadvertently while citing those who are trying to solve reasoning tasks with machines. Interestingly Yoshua Bengio started his recent 2019 NeurIPS talk citing the influence of Prof. Kahneman's book on him. In fact Yoshua's view of deep learning as System 1 deep learning and System 2 deep learning is the direct influence of Prof. Kahneman's thinking. There is a difference worth noting in Yann LeCun's approach (who Prof. Kahneman mentioned) and Yoshua Bengio's approach to solve System 2 Deep learning tasks (at least from what we can glean from their recent talks). While both are focussing on self-supervised learning (learning by predicting/reconstructing missing parts of input), Yann LeCunn's approach is for models to learn by predicting what happens next in input space. Yoshua Bengio's approach is to learn by predicting what happens next in an abstract space - not directly in input space. The input space and abstract space maps to what Prof. Kahneman refers to in the middle of your conversation as "experience" (input space) and "memory" (abstract space). Prof. Kahneman even elaborates that our memory of experiences is not a full replay of experience but a compressed version(low dimensional version in ML speak) of it. Yoshua's approach is to make prediction in that abstract space and learn from that. When we see a person let go of a pen - we predict it will fall, not the exact position it will fall. When we plan our trip to a place, we plan not the actual experience of the trip but the salient aspects of it. Making prediction in a low dimensional (System 2 representations) abstract space that is anchored in representations learned from perception (System 1 representations) is based on the assumption that changes in the world can be explained by a few causal variables - making predictions in such a space helps the model learn representations that capture causality, which system 1 Deep learning lacks. If this assumption is true, then predicting what happens in abstract space has the benefit of learning causal variables that are invariant to underlying changes in distribution in the input space. The training objective for predicting in the abstract space leverages off the changes in the underlying input space distribution as the means to learn its representation and its prediction performance serves as a metric to evaluate its learned representations. However, there are challenges to predicting in the abstract low dimensional space - specifically what would the training objective exactly be (Yoshua elaborates this in his 2019 NeurIPS talk). Relevant links Yoshua's Dec 2019 NeurIPS talk - ruclips.net/video/FtUbMG3rlFs/видео.html Yann LeCun's Dec 2019 talk - ruclips.net/video/A7AnCvYDQrU/видео.html A practitioner's view that examines the above two approaches in some detail, as well as other plausible approaches to solve System2 tasks - qr.ae/TJZ0d5
I’ve never seen a better selection of guests than you’ve had. The Russian historian you had on changed my appreciation for a whole side of the world. Kahneman changed the world, and you’re changing the world. Do you read ad’s? Yes. Don’t let that make you feel like you’re just some average podcaster or media personality. Truly grateful.
Such a humble and knowledgeable scientist. He says I don't know or I'm not sure so many times, and yet, he doesn't shy away from having an opinion based on intuition. You can feel his commitment and love towards reason and science. Thanks, Lex!
I listened to this interview twice. I felt honored to hear it. Daniel Kahneman’s deeply thought out and carefully nuanced responses to Lex’s questions amazed me. “Thinking, Fast and Slow,” was one of the most life-changing books I ever read. My humble comments: 1. How much of the “fast” thinking is based on evolutionary hardware (which we cannot yet change) and how much of it is based on: a) inputs from the environment, including learning; and b) the habits we create as we live? In short, how can we use the brain we have in a maximized way? 2. As I understood Sartre’s philosophy we choose the meaning of our existence. But which part of us does this? The fast or the slow thinking part? (Bit of Sartre trivia for those who speak French-Sartre and his lover and life-long companion, Simone De Beauvoir, addressed each other with the formal “Vous”). 3. Does AI need memory in the form of a human narrative, that is, a timeless, but sequential, series of events? Is it important to eliminate time from the narrative? 4. What if we had a respected association of statisticians give their mark of approval (like the “UL” certification on electrical appliances) to studies of human behavior or nutrition that news sources wish to publicize before waiting for follow up studies? We have all read news articles about studies that show that some spice or food reduces cancer or some other health risk, or that some behavior makes you more successful in life-and yet the interesting results are sometimes just random and based on small sample size. Respected journalists could at least state that the study meets some basic requirement of statistical significance. Thank you. William L. Ramseyer
DAMN. Daniel freaking Kahneman, I am so excited for this. His book "thinking fast and slow" took over my life for a solid year, and I still think of these ideas in relation to machine learning. This is my pre-watch comment, thank you in advance!!!
Many thanks for this video , Lex. As usual, I am very much excited about the topics discussed. Please do not stop doing this incredible podcast. Good luck for you!
The only thing really missing with the podcast is a better forum to discuss things and connect with people who enjoy this kind of content. Traditionally social media is just such a waste of time to me. I haven't even listened to this podcast yet but I would love to hear what brilliant people who listen to it think about the ideas. I am literally going to take notes on this one. I love Kahneman so much.
I watch Lex occasionally and it’s hilarious RUclips recommends this to me as I’m about to begin “Thinking Fast and Slow”. Going to listen to this before I read the book I think! Thanks as always Lex!
Lex thank you for having this guest. I bought “thinking, fast and slow” a few years back and loved it. There is so much insight into how the mind works in that book. It seems like it would really help with AI concepts. (I believe I got turned on to it by reading something else by him and Tversky before that.) Daniel is such an interesting guy. Great interview!
Nice conversation. Like a good conversation between good friends. As much as I like detailed discussions about specific subjects, this one show the humanity in us, which is much needed. Thanks for sharing 👍🏼
Great conversation - thanks for doing this amazing series. I can't help notice how good the video and sound are - can you share the setup you use (video camera, lights, mikes)?
This is the first time I've heard of or seen this man, but it is so glaringly obvious I've been missing out on some profound knowledge and perspective. Brilliant.
"There was a time when people read books. And you could assume that your friends had read the same books that you had read. So there was a great deal of sharing." Damn, that beats the hell out of twitter and instagram.
I think I can see a bit of a proud smirk on your face at the beginning when you say Daniel Kahneman is on the podcast! You earned it if so! Great work lately!
"When you think about something it looks more important than it really is" "No, I don't think meaning is all that important. Personnaly, in those nazi camps, i'd just give up and die" "Whether you get a good science collaborator is mostly luck" "I've never seen Instagram" "The why [purpose of humanity] is hopeless, really" Kahneman is savage. Those are some heavy-ass hits on ideas of hard work vs luck and meritocracy.
58:09 Aaawww 😊 One can really see how Mr. Kahneman re-experiences the bliss of the collaboration with Amos Tversky and sees it before his inner eyes. So sweet, I wanna give him a hug 🤗 Thanks for your work. I really appreciated the book 'Thinking Fast and Slow'. Greetings from Germany 😚
At 34:44. and pleased to hear Daniel address the point that Go! or Chess have very specific constraints, while road travel via vehicle, at all ranges of legal allowed speeds (with far greater variations at speeds in excess of legal limits) involves many choices made within similar levels of constraint, but those constraints are not at all absolute, and in fact vehicular control can engage a need for decisions under a far greater variety of conditions or options and constraints.
Fantastic channel. Bravo Guests I’d be thrilled to see you have a discussion: -Douglas Hofstadter -Robert Sapolsky -George Lakoff -Peter Thiel -Alex “Sandy” Pentland
I haven't listened to the podcast yet and judging by the comments I'm very excited to hear it. I'm subscribing now. Thanks to Rogan's podcast. Great to meet you Lex.
Fabulous chat and wonderful insights. "On political, religious and climate change, people don't change their mind - and by and large there is very little you can do about it"
Conversation with Daniel on AI is a real kick start for a thought process. Lex is a very good person showing a lot of empathy in being a careful listener . I listen to Kahneman and also to Daniel Dannet_ unified theory of information. My interest is in the use of Genetic Algorithms for my area of aviation maintenance and engineering systems and evolve agents _ micro AI . thanks . chris
I have yet to read his book. However, I came to the same conclusion myself that we operate on an automatic system much of the time, whereas other times we actually have to produce, thought and build our ideas much like you would architect a building... Definitely have to read his book now. I love this podcast Lex!
Very good conversation. A bit short for what Kahneman is qualified to speak about. I would like to hear you question him about biases and heuristics more. And the concept of prospect theory
Loved the book! Can you specify why the masses are traditionally wrong in their thinking... is this related to utilization of systems one and/or system two? Thanks Dr. Ken
Yes, excellent suggestion. In the meantime I highly recommend: www.bbc.co.uk/programmes/p06qvj98 "‘Just because you passionately believe in it, that doesn't make it a good idea."
Such a terrific interview - the talk about systems one and two got me thinking about jiu jitsu - perhaps the process of learning and developing in b j j is a particularly acute example of the transmission of skills from the system 2 domain (process learning requiring work) to system 1, where they can be applied instinctively in dynamic situations.
It was indeed an amazing podcast! It was full of novel ideas. I was fascinated to the fact that Daniel Kahneman’s theories are being used to build AI systems. His book is amazing too. Thanks to Lex for asking very interesting and diverse questions.
You both reasoning about pedestrians was so deep. Sometimes I think it is going to be more likely to put compulsory devices on humans than make a perfect reaction algorithm in vehicules. We may end up making humans "robots" enough to be able to interact efficiently without dying.
Hi there! I was wondering around the issue of predicting pedestrian's behaviour mentioned during the conversation and I was thinking about the way "pedestrians" are supposed to behave in the case they have to interact with a train. Given we know that there are physical limitations for a train to stop, "pedestrians" are educated to respect train technology. Could this apply to the overall human behaviour to develop through education as soon as we decided to reduce the overall accident risk globally? Eliminate some predictable variance and control what we can control? More or less the way we also do with airlines accidents? Do you see a future where people are expected to be educated to coexist with machines? Just a thought.
I have a pretty successful integration with the youtube algorithm. To the point that quite often it recommends me the content that would fascilitate my exploration of a yet unborn line of thought. I can have an intuition, a vague idea or a dream and the next day I bump into a video that fully expresses that intuition without me even searching. Algorithm obviously doesn't know me, but it is able to match me with people who were at the same spot before me and essentially gives me their collective advice on what would I like. I don't have a mentor in my life so I have a hard time defining novel problems in my personal development and that's where this recommendation system is of great help. There is a lot of noise in recommendations but at the same time a consistent flow of incredible hits.
![Kehlani - After Hours [Official Music Video]](http://i.ytimg.com/vi/6qatTTscK4c/mqdefault.jpg)
I really enjoyed this conversation with Daniel. Here's the outline:
0:00 - Introduction
2:36 - Lessons about human behavior from WWII
8:19 - System 1 and system 2: thinking fast and slow
15:17 - Deep learning
30:01 - How hard is autonomous driving?
35:59 - Explainability in AI and humans
40:08 - Experiencing self and the remembering self
51:58 - Man's Search for Meaning by Viktor Frankl
54:46 - How much of human behavior can we study in the lab?
57:57 - Collaboration
1:01:09 - Replication crisis in psychology
1:09:28 - Disagreements and controversies in psychology
1:13:01 - Test for AGI
1:16:17 - Meaning of life
Please bring Ellen Langer
AMAZING GUEST! Thank you for bringing this talk to us.
Thanks, as always
You are the man 👍
Hi Lex, i have a wild theorie, whe're i would like to think you about. There are 2 Moments in the dat where you deel to experience reality. During the dat, when you are not sleeping and during the night when you are deeaming. The interesting one ofcourse are dreams. When you dreams, you bever think this is a dreams, you think what you are deeaming is real. It's only after you wake up that you think, and you Remember you're dreams, that you think it's only a dreams, it's not reality. But what if deeaming is more then just trying to recove from a Busy day. What if it's alsof a sort of warning system. I have some mental issues, what causes let's say that im sometimes rational , my old self, but sometimes, emotional , not my old self. I have sometimes, mostly orientee dreams, where i meant and kiss a girlfriend. And the wierd thing is , i then meer her that dat it the following dat/weeks and i light or light not kiss her.Like our conscious/nature Will give us a sneak leek into a possible Future through our dreams. As if 95% of our Future is allready dererminee , but there is still a chance to changer it. Would this be a practical example of hedelbergs u certainty principle/the theory of many world's. Ithe sneak peel actually helper me ones, how to het out of a difficile situation.
I loved watching Kahneman"s wheels turning @ 37:07
Lex, is your expertise in AI or guest recruitment? Damn
Yes.
I was thinking the same thing. How is he getting all these great guests right off the bat? I imagine the MIT cred must help some.... Also, Lex has great questions and interview style. Congrats to Lex on a great podcast with great guests.
Lex is a bit of a hippie, he just doesn't have long hair & always wears a suit ... but philosophically he's not too far off.
@Heath Sims true.
For those of you that don’t know a lot about Daniel is that he never lets sunk cost fallacy get the best of him. Working with him on his book, he would delete and start from scratch without thinking about the effort he put in writing that section at first place. And it’s been a feature of him in his research life, he would start a research and get deep into it but then he would just start something new and not stick to it just coz he had spent years doing it. That’s one of the things I learned from Daniel and since I learned that I stop reading books anywhere I feel like its getting boring or its not worth continuing, I go out of theatre if I dont like the movie after 10-15 minutes, I close down business ideas if they don’t show results within a time frame. It has changed my life. One last thing I learned from doing my research and work was that you should always have an end in mind, as to when you will stop doing something if you don’t get the results you were expecting, it helps you overcome sunk cost fallacy a lot easier.
Thank you so much Daniel for your work with Amos to help us humans make better decisions!
Thanks for sharing.
great comment
I got bored and stopped reading your comment
@@GuildWarsMaude hhahah
And how do you calibrate when you should stop doing something? neither extreme is good, I agree, but the grays are terribly complex to distinguish
"I don't know enough of philosophy to answer that." That is the mark a true thinker: someone who knows they don't know and won't proffer half baked ideas as facts. Amazing. 10/10
I was struck how well informed Daniel Kahneman is about current progress in AI, Chess, and so much more. I want to still be that sharp in 35 years from now.
Yeah me too, being a psychologist at his age and had all that knowledge on AI... Cool
You can say the same for Noam Chomsky. He is older than him. He is like Gandalf in real life, Wise and Sharp.
Gotta keep up with what you find interesting. It helps that life is pretty interesting over all.
2?
Not surprising at all since the progress is negligible if any at all.
Listening to this again. Rest in Peace Daniel Kahneman.
"What makes the experiencing self happy and what makes the remembering self happy are different things." this is the most profound thing I have heard in some time. another phenomenal podcast, thank you Lex!
he was genius to figure all this things out (both with Tversky, don't forget about Tversky, he would get Nobel Prize if he'd be alive :)))
I think there is a complete ted talk of his on this topic itself.
@@PiyushSihag1 better to read his book and some of their journals.
His work is full of mind blowing insights that altered the way I think about human beings.
I am so glad to see that you interviewed this man. His book: Thinking, fast and slow is a gem that everyone should read.
I just want to let you know, I really enjoy listening to your podcast. Your questions feel like they come from the desire to learn and know more. Thanks for making this!
Daniel I'm so grateful for your book. Encountering it is one of the luckiest things in my life. I'm looking forward to your new book. I hope you are always in good health.
Great episode. Thank you Lex.
Awesome guest again Lex! Always keep the big questions in the interviews; it's inspiring to see the variety of outlooks on life held by the great minds of our time.
I read his book "Thinking, Fast and Slow" a couple of times and took many notes, but I must say watching him explain his thoughts in a different manner and how he arrived to his conclusions is bringing even more depth and wisdom to an already amazing work.
Thank you Lex Fridman for your always apt and interesting questions and rigorous preparation in advance to pull out the most value out of your stellar and exemplary guest-list.
The guests Lex both wants to talk with and is able to get on his show says a lot about the level of depth and intelligence Lex has.
What a brilliant guest..I could have listened to this conversation for hours!
When Daniel Kahneman said: "What makes the experiencing self happy and the remembering self happy, are 2 different things" What a doorway of thought that subject opens.
This interview with Daniel Kahneman is THE BEST youtube interview I've ever seen, not just about the topic of AI or philosophy of mind and consciousness, but hands down in terms of being interesting and informative and utterly relevant to my thinking (and scholarly projects at the moment). Thank you, Lex! I applaud your work and efforts with these video interviews.
The most profound line for me was regarding how people no longer need to remember things. This shook me and made a chill run down my spine because my immediate thought was, those in control of the knowledge create the narrative.
Exactly
I don't understand by your saying"those in control of the knowledge". What I took away was that you don't need the "truth" to make up a fictional narrative.
Clearly you have not read '1984', you should, everyone should:
"Who controls the past controls the future:
who controls the present controls the past."
George Orwell (1948)
@Mr. H There's no difference. Either they have something in their direct experience, or they recall a list of symptoms exactly like those on web md
(I’m a nobody to challenge this man but...) I’m not sure it’s right to say that we no longer need to remember facts. Surely knowledge accumulated over a long time is the foundation of wisdom - like the wisdom that Kahneman talks about in following his gut with research. You can’t Google that.
Im glad to be one of the 100,000 people Lex! Thanks for the fantastic conversation! 👍
“...Because time is all that we have got to live… Time is the currency of life...” Daniel Kahneman 43:01 - 43:07.
Time is an illusion - Einstein
oh my god. Daniel Kahneman is one of my Heros. As a psychology Student this seems extremly interesting for me. Thank u so much for sharing.
RIP Kahneman!
Lex is such an amazing interviewer. I'm always impressed by the quality of his questions and the obvious respect these incredibly accomplished guests have for him and his intellect. Top notch.
Thank you so much for these interviews. They are insanely intriguing and curiosity driving particularly in the format they are given.
57:58 wow (- I like how Kahneman blinks affirmingly after each statement). - Lex is asking the best questions, and is always ready to receive more information than was sent. - I just wished it “clicked” more often with his guests. (He also has the best guests, as numerous commentators already mentioned). - I’m so glad to have “found” this channel.
Wow Lex, another Major guest. Thank you for all the great content!
Lex, Thank you for these insightful conversations. You are the Terry Gross of AI interviews for depth and quality.
Mention of Yoshua Bengio's work was perhaps missed inadvertently while citing those who are trying to solve reasoning tasks with machines. Interestingly Yoshua Bengio started his recent 2019 NeurIPS talk citing the influence of Prof. Kahneman's book on him. In fact Yoshua's view of deep learning as System 1 deep learning and System 2 deep learning is the direct influence of Prof. Kahneman's thinking.
There is a difference worth noting in Yann LeCun's approach (who Prof. Kahneman mentioned) and Yoshua Bengio's approach to solve System 2 Deep learning tasks (at least from what we can glean from their recent talks).
While both are focussing on self-supervised learning (learning by predicting/reconstructing missing parts of input), Yann LeCunn's approach is for models to learn by predicting what happens next in input space. Yoshua Bengio's approach is to learn by predicting what happens next in an abstract space - not directly in input space. The input space and abstract space maps to what Prof. Kahneman refers to in the middle of your conversation as "experience" (input space) and "memory" (abstract space). Prof. Kahneman even elaborates that our memory of experiences is not a full replay of experience but a compressed version(low dimensional version in ML speak) of it. Yoshua's approach is to make prediction in that abstract space and learn from that. When we see a person let go of a pen - we predict it will fall, not the exact position it will fall. When we plan our trip to a place, we plan not the actual experience of the trip but the salient aspects of it. Making prediction in a low dimensional (System 2 representations) abstract space that is anchored in representations learned from perception (System 1 representations) is based on the assumption that changes in the world can be explained by a few causal variables - making predictions in such a space helps the model learn representations that capture causality, which system 1 Deep learning lacks. If this assumption is true, then predicting what happens in abstract space has the benefit of learning causal variables that are invariant to underlying changes in distribution in the input space. The training objective for predicting in the abstract space leverages off the changes in the underlying input space distribution as the means to learn its representation and its prediction performance serves as a metric to evaluate its learned representations. However, there are challenges to predicting in the abstract low dimensional space - specifically what would the training objective exactly be (Yoshua elaborates this in his 2019 NeurIPS talk).
Relevant links
Yoshua's Dec 2019 NeurIPS talk - ruclips.net/video/FtUbMG3rlFs/видео.html
Yann LeCun's Dec 2019 talk - ruclips.net/video/A7AnCvYDQrU/видео.html
A practitioner's view that examines the above two approaches in some detail, as well as other plausible approaches to solve System2 tasks - qr.ae/TJZ0d5
I’ve never seen a better selection of guests than you’ve had. The Russian historian you had on changed my appreciation for a whole side of the world. Kahneman changed the world, and you’re changing the world. Do you read ad’s? Yes. Don’t let that make you feel like you’re just some average podcaster or media personality. Truly grateful.
Which russian historian are you talking about? I'd like to see it as well :)
@@delinquenz46 Stephen Kotkin. Sorry I never got your reply until just now!
Such a humble and knowledgeable scientist. He says I don't know or I'm not sure so many times, and yet, he doesn't shy away from having an opinion based on intuition. You can feel his commitment and love towards reason and science. Thanks, Lex!
I listened to this interview twice. I felt honored to hear it. Daniel Kahneman’s deeply thought out and carefully nuanced responses to Lex’s questions amazed me. “Thinking, Fast and Slow,” was one of the most life-changing books I ever read. My humble comments:
1. How much of the “fast” thinking is based on evolutionary hardware (which we cannot yet change) and how much of it is based on: a) inputs from the environment, including learning; and b) the habits we create as we live? In short, how can we use the brain we have in a maximized way?
2. As I understood Sartre’s philosophy we choose the meaning of our existence. But which part of us does this? The fast or the slow thinking part? (Bit of Sartre trivia for those who speak French-Sartre and his lover and life-long companion, Simone De Beauvoir, addressed each other with the formal “Vous”).
3. Does AI need memory in the form of a human narrative, that is, a timeless, but sequential, series of events? Is it important to eliminate time from the narrative?
4. What if we had a respected association of statisticians give their mark of approval (like the “UL” certification on electrical appliances) to studies of human behavior or nutrition that news sources wish to publicize before waiting for follow up studies? We have all read news articles about studies that show that some spice or food reduces cancer or some other health risk, or that some behavior makes you more successful in life-and yet the interesting results are sometimes just random and based on small sample size. Respected journalists could at least state that the study meets some basic requirement of statistical significance.
Thank you. William L. Ramseyer
I found his humbleness to be his key to the light he brought to us all. I’m inspired! Thank you Lex and thank you Daniel
DAMN. Daniel freaking Kahneman, I am so excited for this. His book "thinking fast and slow" took over my life for a solid year, and I still think of these ideas in relation to machine learning. This is my pre-watch comment, thank you in advance!!!
Many thanks for this video , Lex. As usual, I am very much excited about the topics discussed. Please do not stop doing this incredible podcast. Good luck for you!
The only thing really missing with the podcast is a better forum to discuss things and connect with people who enjoy this kind of content. Traditionally social media is just such a waste of time to me. I haven't even listened to this podcast yet but I would love to hear what brilliant people who listen to it think about the ideas. I am literally going to take notes on this one. I love Kahneman so much.
I watch Lex occasionally and it’s hilarious RUclips recommends this to me as I’m about to begin “Thinking Fast and Slow”. Going to listen to this before I read the book I think!
Thanks as always Lex!
Lex thank you for having this guest. I bought “thinking, fast and slow” a few years back and loved it. There is so much insight into how the mind works in that book. It seems like it would really help with AI concepts. (I believe I got turned on to it by reading something else by him and Tversky before that.) Daniel is such an interesting guy. Great interview!
Nice conversation. Like a good conversation between good friends. As much as I like detailed discussions about specific subjects, this one show the humanity in us, which is much needed. Thanks for sharing 👍🏼
I enjoyed this interview-well done Lex and thanks for the work you do!
WOW! Great guest I enjoyed his book and the Fast and Slow approach... Hey invite next Robert Sapolsky!
That would be awesome!
@@fatmamahmoud9433 He will probably destroy the illusion of existing intelligence all together.
Between Thinking fast and slow and Sapolsky’s Behave I know everything there is to know in the world
@@matthewmeyers6917 If you really believe that, then you should read those two books again.
@@leedsdrumacademy Pretty sure Matthew Meyers’ comment was tongue-in-cheek.
First time to hear your intrrview with Mr. Kehneman. Lots to think about and happy for the pleasant and unhurried conversation.
Great conversation - thanks for doing this amazing series. I can't help notice how good the video and sound are - can you share the setup you use (video camera, lights, mikes)?
Loving listening to an elder who has not been infected with social media. Enjoy it while it lasts. Love xxx
Wow. I was always a fan of Daniel Kahneman but this knocks it up a level. Amazing mind.
I absolutely loved how humble and grounded Daniel Kahneman is. No big words, no fluff.
Thanks for having this conversation with a genius like Dr. Daniel Kahneman, really enjoyed it
Thank you Lex and Daniel. It was a pleasure to listen to.
This is the first time I've heard of or seen this man, but it is so glaringly obvious I've been missing out on some profound knowledge and perspective. Brilliant.
Brilliant interview. Some of the very best content on RUclips. Well done mate.
"There was a time when people read books. And you could assume that your friends had read the same books that you had read. So there was a great deal of sharing."
Damn, that beats the hell out of twitter and instagram.
That was great dude -thanks for visiting him and letting him be profound while asking him profound questions.
I'm reading his book right now! Thanks for doing this!
A stupendously good contribution to The Great Conversation as usual Lex. Superb.
Thank you very much Lex. You inspire me every time.
I think I can see a bit of a proud smirk on your face at the beginning when you say Daniel Kahneman is on the podcast! You earned it if so! Great work lately!
"When you think about something it looks more important than it really is"
"No, I don't think meaning is all that important. Personnaly, in those nazi camps, i'd just give up and die"
"Whether you get a good science collaborator is mostly luck"
"I've never seen Instagram"
"The why [purpose of humanity] is hopeless, really"
Kahneman is savage. Those are some heavy-ass hits on ideas of hard work vs luck and meritocracy.
Excellent talk. Thank you for posting.
One of the most captivating podcasts I've ever watched.
58:09 Aaawww 😊 One can really see how Mr. Kahneman re-experiences the bliss of the collaboration with Amos Tversky and sees it before his inner eyes. So sweet, I wanna give him a hug 🤗 Thanks for your work. I really appreciated the book 'Thinking Fast and Slow'. Greetings from Germany 😚
At 34:44. and pleased to hear Daniel address the point that Go! or Chess have very specific constraints, while road travel via vehicle, at all ranges of legal allowed speeds (with far greater variations at speeds in excess of legal limits) involves many choices made within similar levels of constraint, but those constraints are not at all absolute, and in fact vehicular control can engage a need for decisions under a far greater variety of conditions or options and constraints.
Pure gold here... Loved this!
Fantastic channel. Bravo
Guests I’d be thrilled to see you have a discussion:
-Douglas Hofstadter
-Robert Sapolsky
-George Lakoff
-Peter Thiel
-Alex “Sandy” Pentland
So so relevant today. What a great man Daniel was. He will be missed. Very missed.
Just finished watching. Absolutely stunning conversation. Please do another one with him.
Knocking it out of the ballpark w/ these guests & your interview skills!
A quite interesting talk. I couldn't skip a minute.
Lex: Any chance for a podcast with *Carlo Rovelli* on a nature of time and causality?
I haven't listened to the podcast yet and judging by the comments I'm very excited to hear it. I'm subscribing now. Thanks to Rogan's podcast. Great to meet you Lex.
When I read professor’s book, I thought what a wise man he was; after this talk, he is not alone!
Yet another delightful, educational conversation!
Beautiful Conversation....
Sharing information like this is not just about being wise, but also being honest and brave!
I clicked on this so fast! I just started reading his book a few weeks ago!
Fabulous chat and wonderful insights. "On political, religious and climate change, people don't change their mind - and by and large there is very little you can do about it"
I do listen to your podcast several times a day. I am feeling more smart
Conversation with Daniel on AI is a real kick start for a thought process. Lex is a very good person showing a lot of empathy in being a careful listener . I listen to Kahneman and also to Daniel Dannet_ unified theory of information. My interest is in the use of Genetic Algorithms for my area of aviation maintenance and engineering systems and evolve agents _ micro AI . thanks . chris
Sorry i pressed like on my own comment by mistake
Lex's podcasts are excellent. Are there any other comparable podcasts? If so, please share.
I have yet to read his book. However, I came to the same conclusion myself that we operate on an automatic system much of the time, whereas other times we actually have to produce, thought and build our ideas much like you would architect a building... Definitely have to read his book now.
I love this podcast Lex!
I hadn't realized how savvy Kahneman is on AI!
He’s a savvy man on a lot of topics!
Very good conversation. A bit short for what Kahneman is qualified to speak about. I would like to hear you question him about biases and heuristics more. And the concept of prospect theory
Very well narrated .. like the choice words chosen
Loved the book! Can you specify why the masses are traditionally wrong in their thinking... is this related to utilization of systems one and/or system two? Thanks Dr. Ken
Hi Lex, as a continuation to Daniel Kahneman it would be cool if you could get Dan Ariely on. Love your interview style!
I would love to hear a conversation between you and Demis Hassabis !
Yes, excellent suggestion.
In the meantime I highly recommend:
www.bbc.co.uk/programmes/p06qvj98
"‘Just because you passionately believe in it, that doesn't make it a good idea."
Really enjoyed the Podcast Lex. Thanks
This is just too good. Lex has mastered recommendation system, unlike youtube. Just thinking about behavioural economics.
Excellent discussion! Love it! Thx Lex
Such a terrific interview - the talk about systems one and two got me thinking about jiu jitsu - perhaps the process of learning and developing in b j j is a particularly acute example of the transmission of skills from the system 2 domain (process learning requiring work) to system 1, where they can be applied instinctively in dynamic situations.
Lex thank you for everything.
That was amazing thank you!🦋🕊
Now that Lex is posting so much content the rest of youtube just isn't the same anymore. I keep coming back here.
It was indeed an amazing podcast! It was full of novel ideas. I was fascinated to the fact that Daniel Kahneman’s theories are being used to build AI systems. His book is amazing too. Thanks to Lex for asking very interesting and diverse questions.
Another great podcast Lex! Any chance you have plans to interview Ben Goertzel of SingularityNET?
Lex you are amazing and this podcast you have is superb
You both reasoning about pedestrians was so deep. Sometimes I think it is going to be more likely to put compulsory devices on humans than make a perfect reaction algorithm in vehicules. We may end up making humans "robots" enough to be able to interact efficiently without dying.
Always excellent, Lex!
Hi there! I was wondering around the issue of predicting pedestrian's behaviour mentioned during the conversation and I was thinking about the way "pedestrians" are supposed to behave in the case they have to interact with a train. Given we know that there are physical limitations for a train to stop, "pedestrians" are educated to respect train technology. Could this apply to the overall human behaviour to develop through education as soon as we decided to reduce the overall accident risk globally? Eliminate some predictable variance and control what we can control? More or less the way we also do with airlines accidents? Do you see a future where people are expected to be educated to coexist with machines? Just a thought.
Knowledge is power. This is an educational post.
I respect your intro/ explanation to your ads. I subbed and will follow on Twitter.
The ad placements can often RUIN THE FLOW of a piece of content. U Get It!
This might be the greatest podcast ever
I have a pretty successful integration with the youtube algorithm. To the point that quite often it recommends me the content that would fascilitate my exploration of a yet unborn line of thought. I can have an intuition, a vague idea or a dream and the next day I bump into a video that fully expresses that intuition without me even searching. Algorithm obviously doesn't know me, but it is able to match me with people who were at the same spot before me and essentially gives me their collective advice on what would I like. I don't have a mentor in my life so I have a hard time defining novel problems in my personal development and that's where this recommendation system is of great help. There is a lot of noise in recommendations but at the same time a consistent flow of incredible hits.
How do I find the music that is the intro to your podcast? Gorgeous guitar.
that was beautiful , thank you 🙏
fascinating conversation! hello from norway (sociology student)