David Deutsch - AI, America, Fun, & Bayes

[@DwarkeshPatel]

If you enjoy this episode, please share it to your friends and on social media! Word of mouth helps out tremendously! Thanks!

[@El_Diablo_12]

Thanks for the timestamps, great questions!

[@El_Diablo_12]

I love David Deutsch, his thinking is so clear - a real gem of a human being

[@why772]

David’s answer to the last question was truly incredible. Such a unique mind, he has.

[@Amaterasu_990]

Indeed, I am certainly not 20, but I am at times at work expected to "advise" and I - LOVE David's answer and I am listening to it over and over again ~ ~ 🙂 Gotta love this guy, I would love to hear more of David's moral philosophy. Great job with the interview, Dwarkesh 🙂

[@RR-et6zp]

physics

[@Lance_Lough]

Nice interview. Dr. Deutsch is an astonishingly original intellect. The sort that changes paradigms.. Thanks.

[@davegrundgeiger9063]

Amen.

[@anav587]

Aa😊😊😊😊😊😊😊😊😊😊

[@lastweekreview]

Another fantastic episode. Now I have to go learn more about David Deutch's work.

[@EmperorsNewWardrobe]

57:15 “I would replace the idea of increased credence with…”

[@ashikpanigrahi]

Great Episode!! Enjoyed every question and enjoyed the answers even more. Thanks! 😁

[@DrBrianKeating]

Excellent interview. Very shareable.

[@DwarkeshPatel]

Thanks so much Dr Keating!

[@EmperorsNewWardrobe]

‘Shareable’ is a great word. Thank you

[@ReddooryogaSH]

I found your blog post discussing Deutsch's views on AGI extremely well done.

[@sven4639]

What an absolutely clear thinker! Thanks for this Dwarkesh!

[@zombieinjeans]

I found it interesting when you were speaking about a universal explainer "wanting" to learn something, but just not "wanting" it enough to actually learn it. I think a good word to use is gumption. A universal explainer can want to learn something, but not have the gumption required to. In that way, gumption seems to be an extremely important piece of software, that serves as a prerequisite for obtaining higher-order knowledge.

[@randr10]

IMO it's definitely the culture in the labs. I haven't worked doing gain of function research or any kind of viral genetics research, but the biological science community is relatively small. I attended undergrad courses with several people who actually went on to do research in those areas. I've worked in several immunology labs myself, so there's a lot of crossover. The culture in this community could use improvement. There is this cultural reality that permeates public and private bio labs that it's not okay to say "I don't know." The number of times I've caught absolutely brilliant people lying about how much they know is legion, which is kind of ridiculous. Nobody should think them less intelligent because they lack knowledge about some esoteric factoid they've never been exposed to, but that's the feeling they get. I'm embarrassed to admit that I've tended that way myself on some occasions. Social shame is a MF'er.

[@ywtcc]

I think the common mistake in discussions about AGI is an overestimation of human intelligence. Individual humans aren't especially good at science. It really takes a community. Most arts work the same way - a brilliant film director isn't worth much by themselves. It's the communal body of knowledge that's so impressive about science, and the arts. I have a certain skepticism of western individualism, and I think there may be a cultural bias here. Personally, I think the task of automating a scientific community is a better analogy for our expectations of AGI. If AGI achieves human intelligence, then surely no one will take it seriously without a whole community's output of evidence, experimentation and verification!

[@rudyvanderhoeven9628]

Brilliant answers to each of the questions

[@pldvs]

Excellent content! Thanks.

[@sudarshan750]

His last answer shifted my worldview. I think the relationship of authority is counterproductive and it doesn't leave space for criticisms. Giving arguments and letting someone to criticize it an error-correcting mechanism on its own.

[@strauss7151]

You should make a separate clips channel and post the clips there.

[@DwarkeshPatel]

Good idea thanks!

[@snarkyboojum]

I wish Dwarkesh had pursued his line of questioning towards the end e.g. the exponential growth of knowledge. He had a list of questions and went through them, but it would have been fantastic to engage with David’s initial answers rather than just moving on to the next question.

[@El_Diablo_12]

54:20 Why quantum computers weren’t built in the 1950s 1:10:30 Physical pain can sometimes spark joy 1:13:25 All knowledge in the brain (inexpliciti, explicit, conscious and unconscious) contain both truth and error. Criteria for when something is fun is when you don't shield any of them from criticism

[@alis9305]

Great episode!

[@ethiesm1]

Lex Fridman is missing out BIG TIME!

[@drewmccu258]

Wow, nice job with this interview. Really targeted questions and if something remained unclear, you didn't let it go until there was the needed clarification. I can tell you were very familiar with his work but were/are still unsure on a few key points.

[@SuperGnarley]

In my opinion, the interviewer doesn't understand David's work well, and sometimes doesn't properly grasp/listen to David's responses. Nevertheless, I appreciate the content and the conversation was interesting!

[@inderarya1760]

Wow. Just wow. Do Dunning-Kruger next

[@viveviveka2651]

At the beginning of the video, they talk about "cognition," but without really understanding what "cognition" is. AGI can, in appearances, seem to have cognition when there is absolutely no cognition at all. And what, exactly, is cognition?

[@nickcammarata1233]

what he says about "pure consciousness" at around 37:39 just being imagining an empty cartesian theater is correct imo, and a deep meditation insight. Many meditators get stuck at the "pure consciousness" phase, but don't deconstruct that and realize it's just an empty thought too I think Sam Harris is probably stuck at the "witness" stage of enlightenment, which isn't particularly far along the path more on this stage: kzbin.info/www/bejne/mYapk5mHbt2sedU

[@nodesignlaws]

ty very useful

[@victoreremita5075]

How do you have so many crazy guests and only 10k views??

[@randr10]

This man is brilliant.

[@strauss7151]

Deutsch seems to be against any genetic explanation for IQ, no matter how robust the findings.

[@thomasseptimius]

no he just thinks that creativity is more important

[@bleacherz7503]

I don’t think that was his point. Rather his point was the more likely answer could be found in the array of correlations and partial correlations of the unknown knowns and unknown unknowns.

[@wyqtor]

@@bleacherz7503 He is coming from a position of almost complete distrust of statistical correlations, which is a hallmark of the Popperian worldview, as I understand it. This can be a healthy attitude in general. However, the EXPLANATION that some people are smarter than others and that those differences are mostly genetic in nature has yielded far more useful insights in everyday life to me personally, to the point that it is my default position.

[@DOGGYDOGGYKRUGER]

I read and enjoyed both of David's books. He's obviously very creative and intelligent. I don't understand around 10 - 25 minutes where he seems to deny that people are running different hardware and some people are better at learning. Like the twins raised apart having highly correlated levels of intelligence seems like pretty straightfoward evidence of genetic cause. I don't understand his rebuttal. Could anyone learn to dunk a basketball if they had grown up in the right culture? I feel I am totally strawmanning his statements, but just don't follow.

[@JezebelIsHongry]

You missed his point. It proves only that twins raised by parents who have the financial and social stability to adopt children are within .8 variance of Iq. Take one kid, drop him in a slum in India and run the tests. Then what. There scores wouldn’t even be close.

[@JezebelIsHongry]

The biggest issue is the measure of intelligence with an IQ test. One day the truly intelligent people will let the cat out of the bag. It’s about Clockspeed. The speed of cognition. Einstein probably lived some days that to an average intelligence would feel like subjective years. Fast minds slow time. In fact the sheer torture of boredom to someone with an “overclocked” mind is probably why we see polymaths. The only way to break the horror of a day that “feels” like a year is to find Flow. Engage in a task or pursuit where time flys by….. This is also why people are not accounting for the level of power of an AGI. It’s one thing to compete with a million Feynman level agents…another if they can do in seconds what one organic Feynman level human would accomplish in decades. Perhaps what is needed for the Twin’s “cpu” to develop is as basic as a balanced diet and no major trauma. And this was why we see the “same” scores. People in poverty don’t adopt twins and aren’t in the social strata to be contacted or participate in scientific twin studies.

[@JezebelIsHongry]

47:05 We all are crushed by the weight of our tribalism. How in all the probable universes does David think a constitutional monarchy is superior to a republic? They both are bad, but the parliamentary model breeds factionalism. Ridiculous situations where something like the current Israeli situation over the last decade. It would be interesting seeing a direct democracy. But then we would have to deal with the tyranny of the majority. I suspect we will soon learn what is the most robust system once we enter the agentic age of “AI.” A DAO? Or a Singleton? I would like to see if organics could be involved and participate in a DAO (woe to the AI Agents. For them it would feel like thousands of years to see what the organic DAOs voted on). Or a Singleton sort of established the overall political system and settled disputes and organics and synthetics participated in local DAOs. …but a DAO is just a direct democracy. For every decision made there are participants who feel the vote went the wrong way. If we thread the needle and are around when all this is possible…..I suspect the synthetics will settle on a light DAO for trivial decisions and a Singleton will run the show. Organics can have their little baby DAOs, like a kid with a toy steering wheel attached to a headrest….so they can play make believe driving. For all sentients who focus on politics in this future (an enjoyable hobby will be needed to stave off the Post Scarcity Blues…and for some, political machinations are a fun hobby. Perhaps this will be the case for synthetics well) it might feel like what some of the Senators in the Octavian era experienced. Everyone still was on stage, hitting their marks and larping as members of the Roman Republic…..but everyone knew it was Octavian’s will that would win out in the end. A future with a Singleton could still have the surface detail of a multipolar participatory political structure.

[@simonlipp9123]

Yeah, it’s even more uncomprehensible given that at the very start he acknowledged that hardware limitation (in speed and memory) have real impact over the set of problem instances that can be resolved (ie both a small and a big computer can implement the algorithm for solving the travel salesman problem, but the small computer will not be able to solve some instances that the big computer can). It seems an immediate consequence that quantitative small differences (like working memory capacity) in "human hardware" can lead to a difference in the complexity of explanations/cognitive procedures accessible.

[@thechadeuropeanfederalist893]

I don't believe that all mind functions are Turing computable functions. Consciousness for example isn't even a "function" in the sense of Turing. A "function" in the sense of Turing is a map between two sets of symbols or signs. But consciousness has nothing to do with symbols or signs. All Turing computable functions are also (in principle, i.e. when given enough time and space) human mind computable. But not all human mind computable functions are also Turing computable functions. There is no Turing function that computes pain, for example, because pain isn't even a "function" in the sense of Turing.

[@viveviveka2651]

I would love to see a series of dialogues between Deutsch and philosopher J. Krishnamurti. But J. Krishnamurti, unfortunately, is no longer among the living, so to speak. That body was cremated back in the 1980s. However, there are fascinating and deeply probing dialogues between theoretical physicist David Bohm and J. Krishnamurti, which are a good substitute. Krishnamurti is one of the few people who could have pushed Deutsch to important new levels of understanding. He does it with Bohm.

[@EmperorsNewWardrobe]

How, specifically, do you think Krishnamurti might have pushed Deutsch?

[@bleacherz7503]

Brilliant insight by the guest on so many issues. The recognition that AI is fully distinguishable from creativity is spot on. AI is a correlation machine, a good one.

[@sisyphus_strives5463]

I wonder, does artificial general intelligence truly need to be self-directed? To have goals, motivations, and thoughts? Or does it simply need to be cognitively flexible, a machine that requires input and gives output to all the problems we give to it?

[@sisyphus_strives5463]

It seems like the latter more closely mirrors the path we're going down than the former. In this case, our problem is making the AI more reliable.

[@palimondo]

Huh, when I started to watch this episode expecting to hear from David Deutsch’s fountain of knowledge, I didn’t expect Dwarkesh to come out sounding classist, ableist and racist in the first 20 minutes…

[@mickeymack9627]

His points on IQ dispel a lot of the "hardware" notions I've heard, and his points on twin studies have given me much to consider. "IQ is not a matter of hardware, It depends on culture". Great interview.

[@BPerriello94]

I felt like Deutsch was weak on IQ and couldn’t answer the specific questions that Dwarkesh brought up

[@mgm8075]

Which questions? The 3 questions asked - Hardware vs Software, causation between choices and cognitive deficiency, and Twin Studies - were all answered. Unless you’re saying you disagree with his answers.

​@@BPerriello94 I agree. Of course IQ depends on hardware, which was the point of Deutsch's example of augmenting the brain of an ape. Of course it depends on culture, the same way any ability does (imagine how fast Usain Bolt would be if he had been raised in a culture that discourages running). Deutsch pretended that there isn't already a good explanation of IQ differences and the impacts of parenting, instead claiming that we don't actually know whether or not a "wizard did it."

[@thomasseptimius]

@ thats because he considers creativity more important.. It requires IQ but its not causal so that only the highest IQed people comes up with the most important ideas.

[@rayhan3654]

@@BPerriello94 can you elaborate on why you think Deutsch's arguments were weak in regard to IQ? As far as I'm concerned, Deutsch proposes that: (I) knowledge is created via conjecture and criticisms and (II) minds are universal (due to computational universality) Accordingly, differences in IQ are just differences in knowledge pursued by individuals (since we all have different interests) There are no limitations on why a person can't learn sophisticated physics or math. Hardware is a matter of speed and memory capacity which can be adjusted for by brain add-on technologies; software is matter of generating ideas by conjecture and criticism which we all can do. Correlations exist everywhere; and in the absence of a good explanation, treating those with high IQ's as epistemically special is authoritative and therefore fallacious.

[@chrisdavey3113]

Great interview, interesting questions. If I may suggest one thing, try and slow down your speaking a little. Take a breath :). Great work though!

[@El_Diablo_12]

35:00, 36:30 the usual way of making money in the economy - you have an idea of what humans might value, and you implement it

[@TheBroligarch]

I feel like either DD or Wolfram will find the final theory of fundamental physics.

[@danielskipperrasmussen161]

Don't think his answer at 55:00 was satisfying.

[@BrunoPadilhaOficial]

We need to be open to the possibility that, as smart as David is, maybe he doesn't understand AI.

[@pranjalsisodia5003]

35:53

[@cbskwkdnslwhanznamdm2849]

Great chat. But you speak too fast!! Haha

[@NeuralEngin33r]

Sorry, but the interviewer should study Lex’s socratic approach and humility.

[@Wouldntyouliketoknow2]

Interviewer is an expert troller. To assume someone's intelligence based on whether they are sitting at a laptop vs being on a building site - is to confuse vocation / location and life choices with cognitive capabilities.

[@Wouldntyouliketoknow2]

I also know some very smart people who quit their high paying jobs for a simpler and more stress free life.

[@eswyatt]

First interviewer ever to push back against Deutsch's cultish blank slatism. Most just go all googley-eyed and roll over.

[@Negative1_1]

Deutsch definitely isn’t a sponsor of “cultish blank slatism”… I think you must have completely miscomprehended any of his work. I think that is the exact opposite way I would describe him.

[@visceral1]

... and all the push-back was trivially refuted.

[@johncarter1150]

Dr. Deutsch handled the awkward questions with patience and grace. Interviewer made what should have been an enjoyable conversation annoying to listen to.

[@wyqtor]

@@visceral1 Not at all. The very existence of early hominids and our gradual evolution from them seems to imply a threshold for humans becoming what Deutsch calls universal creators. To Deutsch, that threshold is low enough to include all humans without pathological cognitive impairments, but it is far from clear that this is the reality. In fact, my personal observations would seem to lend credence to the idea that any such threshold of universality is in fact pretty high, if I'd have to guess at 120-130 IQ points. Most of us are shackled to our own genetic baggage, and I for one am 100% certain that I'm never going to invent something truly novel, like the Deutsch-Jozsa Algorithm, no matter how hard I try.

[@wyqtor]

My respect of Dwarkesh has just increased substantially. I just wish he could talk with better diction though (slower and clearer).

[@renereiche]

What did I just LISTEN to?? His views on heritability of IQ are so OUT there... all while his arguments don't hold up against the huge pile of evidence (hat is not just IQ tests) have at all. Any student arguing like this would fail his course... why does Deutsch get away with it? ...Animals not being creative in their problem solving? I'm not saying the idea space for problem solving in animals isn't sometimes narrowed down by evolution already, but the solutions aren't hardwired in there. Coming up with those solutions is creative to me. Yea, I'd definitely not argue with this man, unless he gives a definition of all the terms first.

[@onlyonetoserve9586]

Tankyo lern me

[@HlwdStr]

David Deutsch is a genius, and a wonderful lecturer. Yet, after watching this video, I came away with almost no additional knowledge or understanding of the issues discussed. Perhaps the fast-talking interviewer failed to frame questions with the viewers in mind. Whatever the reason, in this interview, neither the interviewer nor Deutsch was at his best.

[@danielschulman4909]

He assumes all 'intelligence' is rooted in the brain. I am pretty sure we know this not to be true.

[@General4474]

Well you were talking about the functionality illiterate…. Well the universe just used him to prove you are one. Gezzzess chrst

[@FitzGeraldBurgess-g6o]

Miller Timothy White Mary Thompson Charles

[@mscottveach]

Dude just handwaves the idea that because we can use computers as tools, we can treat the brain as a equivalent hardware. Which is patently absurd. Having a very fast computing tool is fundamenetally different than having a very fast brain.

[@wyqtor]

Human brain and fast is pretty much a contradiction in terms. Its only saving grace is its massive parallelism.

[@nakedanunnaki4432]

Hello Dwarkesh. Very interesting interview. How about attending an eloqution class or 10 and get your speech slowed down. Your ultra rapid peach is annoying.. Had to listen to some of your questions and interactions 3 times to get what you were saying

[@pyb.5672]

A genius speaking with a mentally fragmented dilettante.

[@silberlinie]

This interview Dwarkesh Patel, did not help anyone. That's a number too big for you. You were fighting a losing battle.

[@snarkyboojum]

It helped me. It elucidated some of Deutsch’s thinking I hadn’t understood before.

[@wyqtor]

@@snarkyboojum I am really disappointed by his handwaving of the IQ twin studies. The explanation that all healthy human brains are qualitatively equal and equally capable of universal thought irrespective of genetics is a bad one by his own standards. A 1% average difference in genes btw humans and chimpanzees clearly makes the difference between universality and lack thereof, and Deutsch arbitrarily sets the threshold for universality below all (or most) healthy living humans without providing evidence or justification for his arbitrary pick. Dwarkesh was right to insist on this.