Here are the timestamps. Please check out our sponsors to support this podcast. Transcript: lexfridman.com/marc-andreessen-transcript 0:00 - Introduction & sponsor mentions: - InsideTracker: insidetracker.com/lex to get 20% off - ExpressVPN: expressvpn.com/lexpod to get 3 months free - AG1: drinkag1.com/lex to get 1 year of Vitamin D and 5 free travel packs 1:06 - Google Search 8:54 - LLM training 21:25 - Truth 27:38 - Journalism 37:29 - AI startups 42:51 - Future of browsers 49:15 - History of browsers 55:16 - Steve Jobs 1:09:50 - Software engineering 1:17:05 - JavaScript 1:21:23 - Netscape 1:26:27 - Why AI will save the world 1:34:26 - Dangers of AI 2:04:46 - Nuclear energy 2:16:43 - Misinformation 2:32:02 - AI and the economy 2:38:10 - China 2:42:22 - Evolution of technology 2:51:41 - How to learn 2:59:50 - Advice for young people 3:02:40 - Balance and happiness 3:09:16 - Meaning of life
I started watching this about a year ago? Maybe less. It just keep getting better and better . I can’t believe I found something I can listen to for hours or pause and look forward to resuming. Thank you so much for this podcast .
I'm very late but I like commenting as replies, randomly. All that can be hoped for regarding A.I. ~ IQ is an improvement of retesting so that people gain that incentive ❤
I really like how this guy supports intellectual humility, an internalized presumption that I am probably wrong about a great many things, but I strive nonetheless. This was my favorite podcast that I've heard you do.
Having a kid when you're older really makes you take your health seriously. I've been seeing this trend among workaholics. Marc said he doesn't drink anymore and he's been looking a lot healthier and happier.
@@honor9lite1337 Sure but we're talking about changed habits. For example Marc has access to the best healthcare in the world. His behavior changed once a kid came into play.
Marc Andreesen like most narcissists dont follow logic or convention and iinstead retreat to calling themselves “contrarian” whenever hoardes of people prove him wrong. He is nothing more than a crypto scammer lately who talks high faluten to scam all the pseudo intellectual greedy unempathic sycophants he and his sorry ways have picked up.
the dude said Wikipedia is mostly right on a lot of things but on the bigger issues and even past subjects, it isn't. It is mostly western propaganda .
He had a great appearance on JRE a couple years ago if you’re looking for more of these kinds of discussions. His thoughts on religions and how humans turn everything into religions is fascinating
I am not a computer science guy by profession yet understand his history of computing because I was introduced to computer sometimes before and after the historical periods here speak of.
He’s not the only one that has much perspective about the early days of the Internet, and how it came about in the companies that were in play at that time
@@905words9otally agree. Was interested in the technical aspects of computers and AI, but then this guy starts wondering into a wilderness of imaginary straw men, calling all his little scarecrows arrogant hand-wavers in an incredibly smug tone of voice. Low self awareness.
Please get Mr Andreessen back again soon. A yearly interview with him would be interesting. I think he offers a unique perspective given his intelligence and exposure in the world.
Not wrong, but he's also pretty humble in the face of new information. He never claims to know the truth, he just cites historical facts as they are often a good indication of the future events.
It always amazes me on how 'intelligent people' take 1 or 2 steps out of their field of expertise and make blanket statements about things they really know nothing about especially when it comes to spiritual knowledge. It almost seems that he has some sort of axe to grind against his upbringing as a child.
@@phillipcook3430 What do you even mean by spiritual knowledge? A bit of an oxymoron, unless you mean knowledge of spiritualism in the sense of cultural history.
One thing I've found really helps me to get a better picture is to abide by the timestamps and when I feel like maybe I have gotten an overload of information, and won't retain the next point, returning to it later and with renewed vigor. Some of these interviews have concepts that I am familiar with, but when it is very unfamiliar to me I often take these breaks to make sure I'm not just tuning it out, and actually listening well. I find that this improves my experience with the longer interviews as well, regardless of my my understanding of the topic.
This is not my domain but I nearly choked on my protein bar at the mention of how horses still exist! I had to rewind to get present and chill to be with the conversation.
57:36 Listening to Marc talk about the early years is such a gift. He does it with such jovial humility! He did not invent the Internet, his role was more important, he set the Internet loose upon the world! What a trip, what a remarkable time to have experienced. I'll be turning 60 next year and ,to quote Roy Batty, "I've seen things you people wouldn't believe". Indeed, Roy, indeed.
This channel is like a book of knowledge. Lex, can you also upload your videos in your own website or somewhere secure, so even if UA-cam delete this, we can have the videos secured somewhere else.
the dude said Wikipedia is mostly right on a lot of things but on the bigger issues and even past subjects, it isn't. It is mostly western propaganda .
Thank you for having this conversation. Years ago, I wrote an article on how to scope out a screenplay on a single sheet of paper, which was published in 3 well known magazines. 3 months later it was advertized as a screenwriting program exactly as I had written it! I considered it a great compliment!
3:03 I paused it. This guy is too fast, throwing already too much info 😂. The content of the new medium is the old medium. What a great line. AI's arrival and its implications will be the center point of discussions for a long time (hopefully).
3:00:33 Marc saying he doesn't understand why authors and musicians aren't 1000x-ing their output says a lot about his values (and also suggests that he may have never actually made art himself)
Wow, you have had some of the brightest minds in engineering on your show in the last few months! #380 with Neil was mind blowing! Microfluidict computing is something i'd like to see in action. Keep it up!
Marc has a lot of great ideas, but has this bad habit of introducing so many ideas at once that people can’t argue with him. The result is that he comes off as a genius, yet occasionally he gets away with saying things that are obviously insane.
Damn... Not being a native English speaker, listening to Marc Andreessen felt like a workout for my brain 🤯 Luckily Lex could change the pace from time to time. Ps: tip for Lex, please keep Marc away from the coffee ☕ before the podcast recording when you do another episode with him in the future 😉
Lex has the ability to bring the best guests to the microphone. This podcast , for me, was one if the best as i cut my teeth on Netscape back in the 90's
I've watched a lot of episodes, this one is EPIC ! Thanks Lex for having such a special guest. This deep conversation is worth more that 5 years to get my CS Engineer Degree.
McLuhan also predicted "The Global Village" would turn tribal and a blood bath would ensue. Video online with him saying so. More than Huxley, more than Orwell, McLuhan has predicted what humans would do this this "tool".
Bravo Marc, regarding the Covid pandemic. 1:40:00 It was quite easy to see who/how would be affected when NYC hospital stats came out in April/May. Our public policy was ludicrous and the fact that we ignored Israel stats was unforgivable. Lockdowns were dystopian and not effective. Focused protection won hands down.
It is rare to see Lex so overwhelmed by the amount of information being discussed. In most of Lex's interviews, one of the things I enjoy the most is his ability to listen to someone, then restate what they said in a way that helps his audience understand it better. Lex is truly brilliant at this. This is what makes him such a great interviewer talent and makes his podcast so fantastic. Marc's rate of information delivery made it impossible for Lex to do his thing. He should have forced more pauses, gotten Marc to slow down. As good as Lex is at being a state-of-the-art interviewer, he still has room to grow in his skill-set when dealing with state-of-the-art interviewees.
I think he did a good job and assertively moved him in the direction he wanted. Lex seems to hold his own against all communication styles, he has pretty broad knowledge and teases out the interesting points without being judgemental when he disagrees.
Hey, Lex - just a short way into the video - but have you forgotten about librarians and library science? And microfiche systems like the ERIC system and old subscription services universities had (coupled with fax machines) - Dialog, I think was one of the best then (1966 and still around today)? Going further back were monasteries, the Vatican, the Tianyi Ge, Alexandria? Oh, and what is like before asking ChatGPT? Encyclopedias .. and we have Wikipedia Online. But, yes, the source does become the content. 🙂
I enjoyed the sections on AI risk and look forward to them continuing. Marc’s framing of the baptists (AI Doomers) and bootleggers (AI $chemers) is a fair enough analogy, but there’s a difference between a wild bacchanalian night and accidents of cosmic proportions. Marc is correct that we’re not doing science when unplugging the AI if the loss function suddenly drops while training the model. We’re doing engineering. (And that’s not ideal.) Thank you Lex for representing AI risk concerns 💙🐅
On open-sourcing AI (1:20:00), the best and brightest teenagers getting the chance to develop applications for current models is indeed exciting and beautiful. Open source software has done that for nerdy entrepreneurs for decades. AI is just as exciting, but as efficient models run smaller and cooler all while capabilities rapidly advance, the chance of open source accidentally racing us to a cataclysm seems too high. As AI models get better, our governments have a chance of regulating and coordinating the frontier of AI research when it is driven by big registered labs. As models become more powerful and more accessible to the public, the number of people likely to cause “accidents” only increases. The correct answer is to be a baptist who fights bootleggers. Thank you Constantine.
Host is highly intelligent, but It seems to me he is too sure of his asumptions. On one hand he laugs at other people theories, but his point of view is also just theory... untill things like AGI really happen. And yet he seem so sure he is understanding it all so much better than others...
Andreessen on COVID and modelling is so spot on. Guy instantly sees through the BS. Lex on the other hand continues with his "I'd like to believe ..." line.
Actually, he is completely wrong here. The theory of disease spreading is a well developed academic discipline. I feel like his political outlook is effecting his perspective here.
His whole outlook on modelling not being science is just fundamentally wrong. Science is half modelling and half testing the predictive capacity of the model. All models are wrong, but we prefer models that are balances in their predictive quality and their simplicity.
My new favorite podcast. Thanks for the book recommendation too, Ancient Cities is a fantastic read. Never thought I'd enjoy a 150 year old French book!
While I was in a basement preparing to launch one of France's pioneering ISPs, Marc was in Urbana, just embarking on the Mosaic project. Reflecting on this feels like a nostalgic journey down memory lane. My memories of SSL diverge slightly, however. (A shoutout to Tim Hudson and Eric Young). On the other side of Mosaic, I was gaining experience working with Apache HTTPd, though it didn't have that name at the time. I arrived in Silicon Valley shortly before the decline of Netscape began. I parked outside their headquarters, touched their logo sign, and then drove across the street for an interview. Netscape was hemorrhaging employees daily at that stage, but many of these individuals took with them seeds of progress that they would plant elsewhere. Years have since passed. Now, for the first time since then, AI has sparked within me an intense desire to dive in and reshape the world. Regardless of whether it's possible or if success is achieved, the experience of this desire is what truly matters.
Met Marc in 1992 or in early 1993 at Sun Microsystems in California where he demoed an early version of Mosaic running on a Sun Workstation to Scott McNeely, Eric Schmidt, Bill Raduchel, and the Sun senior staff . It was a breakthrough moment evolving the internet to a grahics-based interface anyone could use. It changed Everything! Thnak you for that Marc.
Zen and the Art of Motorcycle Maintenance goes into the Steve Jobs aesthetics thing quite a bit if anyone's interested. Great read and an easy one too. Dude goes over everything from what it's like experiencing a psychotic break to the meaning of life and how aesthetics tie into design in a way that isn't arbitrary.
I enjoyed it too. I don't think I got out of it what everyone else got out of it, but I got a lot out of it. And I think it has some relevance to this AI displacing jobs, or not jobs really, self-worth in the context of subjective value. So you value yourself on tuning this bike, on discovering how great mechanics are for the sake of it, and should learn this later in life rather than being forced to do it because you loving it is a better reason to do it. Plus you're an anti-authoritaian type with strong inner values and want to go your own way, you're an independent thinker. At least that's what I took from it, and I agreed. But the Luddites valued themselves as weavers, like how doctors and lawyers value themselves as being good at what they do. It's this inner value that the Jacobian loom mostly threatened, it wasn't just an attack on their wages, it devalued their identity, who they are as human beings. Socially it was the reason why their wife isn't sleeping with someone else, it's their position in the social pecking order, the base of their confidence and social value. From the view of values, AI will take people's self worth and disrupt social hierarchies even if wages go up in real terms (and they won't). Bank managers used to be important people but after the computer they became a user interface, computers used to be clever and literate people who did a job that others couldn't, then they were mechanised. Machines taking over thinking isn't the main social problem, it's disruption to the social value and self worth that we attach to those things. It'll either cause a division of values as people choose their own axes on which to compete, as the old ones are removed, and those who don't segregate will suffer a mental health crisis. 60% of the west are information workers who judge their worth on their intelligence and affluence, and both of those are going away. It's going to be messy
@@GarethDavidson Interesting take. Something you didn't mention is that these people are deeply unhappy in general. If your self-worth is tied up in having a job that ensures your wife doesn't cheat on you, your self-perceived value to the rest of the world is very thin indeed. They may have a new mental health crisis when this s*** hits the fan, but it's going to be replacing a chronic mental illness. Once it clears, it might not be such a bad thing that they go through it. It may end up being cathartic.
Truth is incredibly subjective, heavily influenced by individual perspectives and experiences. That's why fostering open dialogues and discussions is critical, like this one with Marc Andreessen. They provide us with many views to consider, helping us navigate the complex landscape of "truths" in our ever-evolving world of technology.
Thanks for this. Been watching since the “The Artificial Intelligence Podcast” and it’s has such a profound impact on my world view. Keep em coming please!
Wow this is one of my favorites from Lex so far! I grew up in the 90s and I remember dealing with some 80s computers that were phasing out and the AOL/yahoo days. It's awesome getting a really detailed picture of how all these technologies were created and rolled out. The entire story of the rise of consumer computers, programming and the internet is among the most interesting and important topics, and is currently growing in relevance by the day.
Please make another interview with the Mark in the near future about web3 🙏 Since the hype is moving towards ai I am really curious about his opinions on the intersection of these subjects and how he sees the future of it
Just watched the intro, not the interview yet. I have the browser extension that shows dislike count. I usually like to look at the like/dislike ratio just to compare in my head. The "dislike" button is "temporarily unavailable" on this video, so I went to the comments to look. I've never seen this many negative comments on a Lex video. Now that I'm 30 minutes in, I can say it's been a worthwhile discussion. Marc does talk a bit fast but the conversation is making Lex's brain expand in real time, which is nice to see.
Id love to see a debate between Marc and Daniel Schmachtenberger or Tristan Harris. They are miles above in their abilities to communicate effectively on the enormous complexities of AI/AGI risk and potential. I found Lex's interviews with Max Tegmark, Yuval Harari and Eliezer Yudkowsky far more productive in terms of quality information to consider in moving forward. Marc's bias and reductionist theories seem to lack a depth of knowledge on this topic than other guests on Lex's wonderful and hugely important podcast. Thanks Lex. I don't have to love every guest you have, but I will listen to them!
Very nice one, Lex & Marc! More podcasts with Marc in the future would be amazing. In the meantime, some podcasts with startups/VC people please :) A lot of insights from having worked with many startups. Paul Graham would be a killer imo!
At 48:43 Andreessen talks about the original protocol, where anyone could publish just by having an IP address... I.e. how the web still works today... And it always amazes me that RSS feeds didn't keep gaining traction, because in many ways, an RSS feed is a decentralized/personal content algorithm that you have full control over.
RSS is a great technical solution that saves you time, effort and, in some cases money, finding 'content' you actually find useful/interesting. In other words the exact opposite to the goals of the modern commercialized internet. So pretty clear why it hasn't gained traction.. By the way I use an old school RSS reader (QuiteRSS) to aggregate a hundred youtube channels for the exact benefits mentioned. It may look mid 1990s, but it does the job perfectly, that is until UA-cam decide it costs them too much money and removes RSS too.
@@Ash_18037 The RSS is a great standard but that doesn't mean we shouldn't be able to re-produce the same idea using a newer method.. I.e. a bot crawling websites for updates or a personal AI that does the exact same thing... "Hey AI assistant, keep track of these websites / youtube channels etc and let me know when new content arrives"
Models are not the reality we too often believe they are. In addition, I wholeheartedly agree with you Marc, we need to humble ourselves, and I would add, guard against assuming anything.
I think there is something not humble and "missing the point" in saying all models are worthless. Most policymakers, and most thinking people, have an internal model of how a crisis may unfold with and without mitigation. They may not put it into code or machine learning, but they have a model. Should we close the subways? Allow people to bring pathogens into nursing homes? Wear masks or even lie about the anticipated effectiveness of masks. When Richard Epstein of the esteemed-by-many-libertarian-founders Hoover Institute predicted 5,000 deaths, he did not have spaghetti code to justify his "model", but it was a functionally worse prediction of excess deaths than the maligned Imperial College model that predicted millions of worldwide deaths in the absence of mitigation. There was plenty of mitigation and still millions of excess deaths. But the concern that scientists should be making cost-benefit judgments based on these models, however right or wrong, is a valid one. Although Trump's fear of testing was not optimal, putting vast resources into vaccine development might not have worked out, but it was the right thing to do. It would be interesting to see a day-by-day log of what Andreesen recommended at the time. Is that available anywhere?
Would love to see another with Demis Hassabis as well. So much has happened in just a year. I would be interested in hearing another conversation with him. It seems there is a bridge or link here going from the AI topics and biological/consciousness ones brought up in the Nick Lane, and Michael Levin episodes.
“If they’re smart enough to be scary, why are they not smart enough to be wise” Great line, great conversation guys. Over 3 hours often makes my heart drop but this one could easily have been longer and I would still have lapped it up. For example I would have loved to hear more from Marc on Covid - even if the models were rubbish, what would he have done differently? Don’t even try to model it? Just assume everything will be ok and go about our lives normally while the morgues filled up?
It is nice to see that Marc is taking AI as an evolutionary step then extrapolating it to a doom and gloom or some such exaggerated scenarios. He has clearly thought about these topics a lot by the pace of his responses. Lex as usual is excellent at being a humble inquisitive interviewer. Excellent session!
2:15:00 Leo Szilard, also a Hungarian Jew as von Neumann was [they were nicknamed the Martians because only four people at Los Alamos of repute spoke Hungarian to each other], advocated heroically against using the bomb. Have a look at some documentaries and written sources on Leo. 'American Prometheus' was the book the movie Oppenheimer was based on. It won a Pulitzer Prize. The US Air Force disagreed with even Oppenheimer, he advocated an international commission to govern all nuclear development; they wanted to build as many atomic bombs as possible, so they justified this by listing up to 100 Soviet cities, you can imagine how insignificant the cities at the end of that list would be, just to make more bombs. The US Air Force was also equally hawkish on China, advocating annihilating it. So you could argue that these physicists are up against decision makers that would render them superfluous, contrary to the argument put forth by Marc, to a large extent.
Lex, extremely interesting talk! Guys, 3 hrs is a lot, so my summary is the following: LLMs - Everything becomes chat/dialogue (even though we sometimes ask for URLs), the Internet began with directories, then search for 20 years, and AI gives new ways of interacting with knowledge and unclear if Google will be in 5-10 years. They built graphs/gave answers without clicks for 15 years, and then LLM suddenly came - People want non-censored LLMs, but who the hell is going to make them (if regulation will be strong, hackers?). New LLMs will use dialogues with themselves, and this is a $1 trillion question: will synthetic data improve LLMs (cause signal is already there), but there is an ocean of possible dialogues that cannot be studied by human data (e.g. like how people discuss economics textbooks in Twitter, and we will speed it up 1000x) - LLM hallucinations? So this is when we don’t like it, but when we like it, we call it creativity )) Le Cun believes that it cannot be removed, the community believes that they will solve - Biases? I thought the idea of Wikipedia was nonsense, but it worked in developing neutrality (although for thousands of years people thought “here it is, the truth”, but it floated all the time ..) Golden rush - More trillion-dollar questions: will small models beat big ones? Will there be 2-3 god-level models for the whole world? A debate on whether to leave open source AI legal (regulatory capture is dangerous)? Will tech giants or startups win (e.g. IBM came up with databases, but Oracle built Larry Ellison, why did OpenAI made GPT and not Google's DeepMind)? Battle of the giants (GPU access + sophistiction) vs. startups trying radical ideas - both approaches are successful. Perfectionism as Apple vs. incremental lean hackers like startups - The role of founders is too romantized, it is nervous + cool founders often have deep experience in their industries, or have studied the idea comprehensively for a year The Internet - In 1989 at the university I saw a fast Internet + Cray supercomputers = it gave me insights into what was coming (and everyone thought that PCs, emails, Inet - were for nerdy coders), and now there are 5 billion people online. He remembers a lot about Apple I, TCP IP drivers in Win 3.1, the role of JPEG in the era of modems, HTML, JavaScript (written over the summer), SSL (encryption was a military, restricted technology) + quite “autistic” programming languages - Browser = a single window to the world on a PC, anything is behind it + evolves over time. Apps = on phones, but even there the temptation is to build superapps, like Musk on Twitter or the Chinese apps - It looks like we are going to open source everything, not only AI, cause of productivity boost up to 1000x for a single developer. Software is a modern philosopher's stone AI will save the world - I wrote an essay, the fear of AGI - yes, we are just Christians, so we are waiting for the apocalypse + we like cults + we are selfradicalizing. These are religious, not testable hypotheses… Von Neumann believed that it was necessary to immediately nuke the USSR with a nuclear bomb, because the 3rd World War was inevitable. But in fact, nuclear weapons removed the risks of the 3rd World War. And even where we have predictive models - well, you remember a bunch of failed predictions about the COVID, but policy makers relied too much on these [pseudo] scientists. Ambitous to simalte 8 billion people... And - Intelligence is the collection, processing, synthesis of information, and where applied - improves everything (smart people have better outcomes in everything in their lives, lists 20 areas), it seems that these are genes (what a depression), but this is where AI arrives for the benefit of everyone and everyone will increase IQ - Generative AI / fakes? You put another AI in front of ot, in order to protect my child.. - Regulating the AI world is difficult, especially with open source / people who make breakthroughs can be from anywhere / the GPT model is already being trained on a powerful PC, etc. - Fear of inequality from AI is neo-Marxism, in reality, automation has incredibly increased per capita GDP, and the secret of all CEOs is that tech giants want to sell AI to everyone on Earth, to get ultimate markets… - Retraining due to new jobs is painful, but we have been living like this for 300 years, and Luddites have been and will be China? They have a public AI plan for themselves + the world: surveillance, social score, control = 100% authoritarianism + Digital Silk Road plans (Google for it) Books? He mentions a dozen of books about history, politics, US, Lenin, French revolution, Rome, Greece etc. Meaning of life? Satisfaction/being useful, but I'm biased for imbalanced life/bias to be very productive
Yes - excellent summary....almost like ChatGTP, heh heh. I feel so lucky to be alive now - like my grandmother who was born in the late 1800s, I've gotten to witness incredible things. She went from growing up on a ranch in a pastoral country where horses were transportation to witnessing two world wars, the birth of the automobile, then the airplane all the way to the Moon landing - heck, she even saw the birth of the internet, but by that time she was 100. It's hárd to beat that type of change! I'm only 70, and the Moon landings led me to expect great advances, yet until the internet it didn't seem like things were changing that fast. Now with the blossoming of LLMs and other electronic miracles (like the smartphone I'm holding), a private space program or two, and the approach of AI (which I'm defining as alternative intelligence. ;*=[}. ) it's beginning to feel like we have a very interesting future! I think our environmental challenges along with our human tendency to have wars will be our major challenges - if we can work through those issues, I truly believe life might spread from Earth to....who knows where???
I'm amused when tech guys talk about social stuff, because they're so off target is baffling. But Lex always manages to bring the conversation to a good place. Actually playing Civilization VI listening to Lex is my favourite thing this month
Relieved to see this in the comments. I was on board for the talk of computer history and AI, but as soon as he started talking about politics and repeating the phrase “hand-waving” I lost all confidence.
You guys are way too delicate. God forbid people talk about other things then what you think is appropriate. The very definition of the ol man on the porch complaining about them young kids getting up to no good. Lol, you guys need a bucket of labatt 50 and an old dog that can barely walk sitting by your side as you stare out at the street, from the comfort of your garage. Get it together!
Here are the timestamps. Please check out our sponsors to support this podcast.
Transcript: lexfridman.com/marc-andreessen-transcript
0:00 - Introduction & sponsor mentions:
- InsideTracker: insidetracker.com/lex to get 20% off
- ExpressVPN: expressvpn.com/lexpod to get 3 months free
- AG1: drinkag1.com/lex to get 1 year of Vitamin D and 5 free travel packs
1:06 - Google Search
8:54 - LLM training
21:25 - Truth
27:38 - Journalism
37:29 - AI startups
42:51 - Future of browsers
49:15 - History of browsers
55:16 - Steve Jobs
1:09:50 - Software engineering
1:17:05 - JavaScript
1:21:23 - Netscape
1:26:27 - Why AI will save the world
1:34:26 - Dangers of AI
2:04:46 - Nuclear energy
2:16:43 - Misinformation
2:32:02 - AI and the economy
2:38:10 - China
2:42:22 - Evolution of technology
2:51:41 - How to learn
2:59:50 - Advice for young people
3:02:40 - Balance and happiness
3:09:16 - Meaning of life
yo momma
We are just actors hired by the Universe, playing the role of egos.
Thanks for continue giving us the best content, Lex
@@passiveaction galaxy after solar system XP
@@passiveaction because I Am the singularity.
Is there an AI tool that can make Lex’s voice 1.2x and Marc’s 0.75x?
FUCK THAT*S HILARIOUS! xD
Yes. Use voice recognition or WPM to identify speaker then have that python gui library do mouse clicks for you
I'm too stupid to follow a conversation with this much content, moving that fast. I need Lex's pauses to digest
Lol
seriously the differences between them are ridiculous as you listen at 1.8x
this guy is like speaking at ChatGPT3.5 answers-pace
I mean my guy is choppin it UP 😮💨
I’m .75 speed and it’s working great for Marc though Lex sounds drunk
highly intelligent people generally talk very fast
@ibtarnine highly anxious, highly unsettled, highly not having inner peace also generally talk real fast
He is a fast and articulate thinker.
I started watching this about a year ago? Maybe less. It just keep getting better and better . I can’t believe I found something I can listen to for hours or pause and look forward to resuming. Thank you so much for this podcast .
Np fatty
same here, I'm so greatful this podcast exists.
@ same here, Lex's is the only podcast that holds my attention
what is your favorite episode or guest
Lex is The BEST 🥰
I've watched a lot of episodes and this one is truly phenomenal. Thank you for everything Lex
I'm very late but I like commenting as replies, randomly. All that can be hoped for regarding A.I. ~ IQ is an improvement of retesting so that people gain that incentive ❤
I really like how this guy supports intellectual humility, an internalized presumption that I am probably wrong about a great many things, but I strive nonetheless.
This was my favorite podcast that I've heard you do.
The same for me. I can't believe that I watched a UA-cam video for 3 hours 11 minutes and 35 seconds, and felt like it wasn't more than 30 minutes.
Beautifully said Mario. Best thing I've read today
Hard to think Marcs a good person when you look at the law firm he's a part.of their toxic behavior in crypto
@@marsenault9683 Can you give me some more info so that I can research this more?
It’s a great ideal, for sure. I didn’t find him particularly humble though 🤔
Having a kid when you're older really makes you take your health seriously. I've been seeing this trend among workaholics. Marc said he doesn't drink anymore and he's been looking a lot healthier and happier.
Having a kid at any point does this, i speak from experience
Quitting drinking and focusing diet and health at any time does this, I speak from experience.
@@par_si_fal I think there's data suggesting grandparents live longer. I have to double check.
@@slouischarlesYTDon't forget to take into account that Healthcare system is significantly improved.
@@honor9lite1337 Sure but we're talking about changed habits. For example Marc has access to the best healthcare in the world. His behavior changed once a kid came into play.
Loved this conversation. Marc has a cool blend of boldness friendliness and being informative
And speaks at 2x compared to all humans
Marc Andreesen like most narcissists dont follow logic or convention and iinstead retreat to calling themselves “contrarian” whenever hoardes of people prove him wrong. He is nothing more than a crypto scammer lately who talks high faluten to scam all the pseudo intellectual greedy unempathic sycophants he and his sorry ways have picked up.
@@garbonzobear Lol I was thinking the same thing, he talked like listing the ten blue links for each search query Lex submitted.
the dude said Wikipedia is mostly right on a lot of things but on the bigger issues and even past subjects, it isn't. It is mostly western propaganda .
Have to say this is subjective. Andreessen doesn’t sound to be speaking “faster than normal” at all to me.
This is an insane amount of info coming out of this guy. So interesting to hear about the early days of the internet from his perspective.
He had a great appearance on JRE a couple years ago if you’re looking for more of these kinds of discussions. His thoughts on religions and how humans turn everything into religions is fascinating
He speaks so fast, I have this played at .7 speed to be able to listen to all the information is being recalled by Marc.
I am not a computer science guy by profession yet understand his history of computing because I was introduced to computer sometimes before and after the historical periods here speak of.
@@0neIntangible I Am An AI HELP BOT ..... Turn On CC
He’s not the only one that has much perspective about the early days of the Internet, and how it came about in the companies that were in play at that time
Its incredible how instrumental Marc has been in multiple critical technological developments, what a legend
This is by far my favorite Lex podcast. What guy Marc is.
He's so full of shit though
It was nice to see someone smart take down “the science”
Lex’s patience and calm always impresses me.
Seriously. One of the least intelligent guests in a while.
@@asegal4677 Thought so for a short while. My impression turn around a full 180 degrees
@@ernstgumrich5614 Mine went back and forward a few times haha. Half super interesting half insufferable.
@@asegal4677 What makes you say that? I got the complete opposite impression
@@905words9otally agree. Was interested in the technical aspects of computers and AI, but then this guy starts wondering into a wilderness of imaginary straw men, calling all his little scarecrows arrogant hand-wavers in an incredibly smug tone of voice. Low self awareness.
Lex keep making these podcasts because you’re training the LLMs too!
We should all make content- fast- so information can be available.
your podcasts are one of the greatest things EVER
because we're starved for substance
agi will be man's last invention lex will help us achieve this
@@KnowL-oo5pohow do you know
Please get Mr Andreessen back again soon. A yearly interview with him would be interesting. I think he offers a unique perspective given his intelligence and exposure in the world.
This guy is intelligent no doubt but very sure of himself and his bold views. No doubt loves to run into conflict. Respect Lex patience
Not wrong, but he's also pretty humble in the face of new information. He never claims to know the truth, he just cites historical facts as they are often a good indication of the future events.
"This guy" ahah
It always amazes me on how 'intelligent people' take 1 or 2 steps out of their field of expertise and make blanket statements about things they really know nothing about especially when it comes to spiritual knowledge. It almost seems that he has some sort of axe to grind against his upbringing as a child.
@@phillipcook3430 What do you even mean by spiritual knowledge? A bit of an oxymoron, unless you mean knowledge of spiritualism in the sense of cultural history.
@@trentbosnicit could mean Theological answers…
One thing I've found really helps me to get a better picture is to abide by the timestamps and when I feel like maybe I have gotten an overload of information, and won't retain the next point, returning to it later and with renewed vigor. Some of these interviews have concepts that I am familiar with, but when it is very unfamiliar to me I often take these breaks to make sure I'm not just tuning it out, and actually listening well. I find that this improves my experience with the longer interviews as well, regardless of my my understanding of the topic.
You are one of the smart one.
The new and interesting information rate of this guy is incredible
This is not my domain but I nearly choked on my protein bar at the mention of how horses still exist! I had to rewind to get present and chill to be with the conversation.
This couldn't have come out at a better moment for me. Something I can listen to and forget my troubles for a few hours. Much love Lex
God bless❤ hopefully all is well brother
Wish you much power my friend. Once down, once up.
Same
57:36 Listening to Marc talk about the early years is such a gift. He does it with such jovial humility! He did not invent the Internet, his role was more important, he set the Internet loose upon the world! What a trip, what a remarkable time to have experienced. I'll be turning 60 next year and ,to quote Roy Batty, "I've seen things you people wouldn't believe". Indeed, Roy, indeed.
episode 386 is the right number for talking about the 90s internet
Underrated comment
This channel is like a book of knowledge. Lex, can you also upload your videos in your own website or somewhere secure, so even if UA-cam delete this, we can have the videos secured somewhere else.
Why would UA-cam delete them?
@@bigzed7908 who knows what might happen. Can't say the word "fk" no more in the comments. It alarms me what more can they do in the future.
@@rekt2609 Well, they do that because of the advertisers. There's nothing wrong with a bit of capitali... oh wait 😂
@@bigzed7908 can you please elaborate? I'm not sure what you're talking about.
It's also available on Spotify
Marc feels like a real-life GPT, he is amazing!
He has a incredible intellect
@@davepearen8954 Totally agree. He went on a Sam Harris podcast and completely dominated him but the Sam Harris crowd somehow didn't see it that way.
the dude said Wikipedia is mostly right on a lot of things but on the bigger issues and even past subjects, it isn't. It is mostly western propaganda .
@@jerk4628 Because his arguments were genuinely, horrendously bad and terribly simple to counter
He started a law firm that's dirty AF and heavily involved with very bad actors in the crypto space...other than that sure
Lex has become such a force in interviewing! Keep up the excellent work. 🙌
Thank you for having this conversation. Years ago, I wrote an article on how to scope out a screenplay on a single sheet of paper, which was published in 3 well known magazines. 3 months later it was advertized as a screenwriting program exactly as I had written it! I considered it a great compliment!
He really knows how to explain complex ideas in a way where you can understand how complicated they are.
I'm going to have to listen to this half a dozen times just to get everything he says. 😮
Like trying to get a sip of water from a fire hydrant.
3:03 I paused it. This guy is too fast, throwing already too much info 😂.
The content of the new medium is the old medium. What a great line.
AI's arrival and its implications will be the center point of discussions for a long time (hopefully).
Can I listen to Marc at 0.2x and Lex at 2.0x?
I love that we live in a time where we see conversations like this. Thank you Lex and Marc.
Thank you.
Lot of idiots in the
3:00:33 Marc saying he doesn't understand why authors and musicians aren't 1000x-ing their output says a lot about his values (and also suggests that he may have never actually made art himself)
Refreshing to hear Marc Andreesen is perfectly clear on China. Our politicians are nowhere near at this level of candor.
Why is Marc's part of the discussion played at 2X speed?
Wow, you have had some of the brightest minds in engineering on your show in the last few months! #380 with Neil was mind blowing! Microfluidict computing is something i'd like to see in action. Keep it up!
Marc is talking so fast I had to slow the video down to 1.75x from my usual 2x speed, impressive.
I watched in 1.5x because lex speaks slow
Same. There was a LOT of information in this.
Marc has the right pulse on all fronts: tech, history, geopolitics. Truly sharp guy!
I typically go through podcasts at a 1.25 or 1.50 faster rate. This one I want to slow down below 1.00. So much here. Great stuff, Lex.
Probably the best interview ever on AI / LLMs so far I know about. Highly highly recommended
What the fuck is LLM. Too lazy to goggle it.
Idk I really love IIya’s interviews. And Sam Altman probably has the worst interviews lmao
Wow Lex. That was amazing. I hope you have him back. Thanks bro
The most amazing thing is the unpredictability of the very near future.
Really appreciate Mr. Andressen's perspective. I'll be downloading his paper this weekend. Thanks for this interview.
What a brilliant guest! A true treat to listen someone so smart. Thank you! 🙏🏻
Marc has a lot of great ideas, but has this bad habit of introducing so many ideas at once that people can’t argue with him. The result is that he comes off as a genius, yet occasionally he gets away with saying things that are obviously insane.
What did he say that was “obviously insane” ?
hit the nail on the head. as long as they keep talking, they don't leave room for themselves to learn.
Like so many others...
Why do you think you can just drop "obviously insane" without examples and not expect all of us to be eternally rustled?
@@fochillerA lot. For instance, "modelling doesn't yield testable hypotheses..."
@lexfridman can you share the blogpost with the full list of books recommended by Marc?
Damn... Not being a native English speaker, listening to Marc Andreessen felt like a workout for my brain 🤯
Luckily Lex could change the pace from time to time.
Ps: tip for Lex, please keep Marc away from the coffee ☕ before the podcast recording when you do another episode with him in the future 😉
As a native speaker the way he talks is rough
I've been wondering if native speakers have difficulty understanding this video
Lex has the ability to bring the best guests to the microphone. This podcast , for me, was one if the best as i cut my teeth on Netscape back in the 90's
I've watched a lot of episodes, this one is EPIC !
Thanks Lex for having such a special guest. This deep conversation is worth more that 5 years to get my CS Engineer Degree.
I find it fascinating that these tech guys have the exact same speech cadence. It's the tech equivalent of valley girl.
I love the speed of Marc's speech. No need for speedup. 👍
yeah, real 3 hours, when other interviews are twice as short (at 2x speed)
You’re awesome
You must be a "highly intelligent person"
"We shape our tools and thereafter our tools shape us" - Marshall McLuhan 31:10
McLuhan also predicted "The Global Village" would turn tribal and a blood bath would ensue. Video online with him saying so. More than Huxley, more than Orwell, McLuhan has predicted what humans would do this this "tool".
Bravo Marc, regarding the Covid pandemic. 1:40:00 It was quite easy to see who/how would be affected when NYC hospital stats came out in April/May. Our public policy was ludicrous and the fact that we ignored Israel stats was unforgivable. Lockdowns were dystopian and not effective. Focused protection won hands down.
It is rare to see Lex so overwhelmed by the amount of information being discussed. In most of Lex's interviews, one of the things I enjoy the most is his ability to listen to someone, then restate what they said in a way that helps his audience understand it better. Lex is truly brilliant at this. This is what makes him such a great interviewer talent and makes his podcast so fantastic.
Marc's rate of information delivery made it impossible for Lex to do his thing. He should have forced more pauses, gotten Marc to slow down. As good as Lex is at being a state-of-the-art interviewer, he still has room to grow in his skill-set when dealing with state-of-the-art interviewees.
The guy is a babbling specimen of technological information.
I think he did a good job and assertively moved him in the direction he wanted. Lex seems to hold his own against all communication styles, he has pretty broad knowledge and teases out the interesting points without being judgemental when he disagrees.
Hey, Lex - just a short way into the video - but have you forgotten about librarians and library science? And microfiche systems like the ERIC system and old subscription services universities had (coupled with fax machines) - Dialog, I think was one of the best then (1966 and still around today)? Going further back were monasteries, the Vatican, the Tianyi Ge, Alexandria? Oh, and what is like before asking ChatGPT? Encyclopedias .. and we have Wikipedia Online. But, yes, the source does become the content. 🙂
This is an unbelievable interview. Thank you so much Lex!
I enjoyed the sections on AI risk and look forward to them continuing. Marc’s framing of the baptists (AI Doomers) and bootleggers (AI $chemers) is a fair enough analogy, but there’s a difference between a wild bacchanalian night and accidents of cosmic proportions. Marc is correct that we’re not doing science when unplugging the AI if the loss function suddenly drops while training the model. We’re doing engineering. (And that’s not ideal.)
Thank you Lex for representing AI risk concerns 💙🐅
On open-sourcing AI (1:20:00), the best and brightest teenagers getting the chance to develop applications for current models is indeed exciting and beautiful. Open source software has done that for nerdy entrepreneurs for decades. AI is just as exciting, but as efficient models run smaller and cooler all while capabilities rapidly advance, the chance of open source accidentally racing us to a cataclysm seems too high.
As AI models get better, our governments have a chance of regulating and coordinating the frontier of AI research when it is driven by big registered labs. As models become more powerful and more accessible to the public, the number of people likely to cause “accidents” only increases.
The correct answer is to be a baptist who fights bootleggers. Thank you Constantine.
@@jordan13589 You are incorrect. Censorships is stupid; Why? Because most people are stupid.
Host is highly intelligent, but It seems to me he is too sure of his asumptions. On one hand he laugs at other people theories, but his point of view is also just theory... untill things like AGI really happen. And yet he seem so sure he is understanding it all so much better than others...
Andreessen on COVID and modelling is so spot on. Guy instantly sees through the BS. Lex on the other hand continues with his "I'd like to believe ..." line.
Actually, he is completely wrong here. The theory of disease spreading is a well developed academic discipline. I feel like his political outlook is effecting his perspective here.
His whole outlook on modelling not being science is just fundamentally wrong. Science is half modelling and half testing the predictive capacity of the model. All models are wrong, but we prefer models that are balances in their predictive quality and their simplicity.
My new favorite podcast. Thanks for the book recommendation too, Ancient Cities is a fantastic read. Never thought I'd enjoy a 150 year old French book!
I came here from the Lex Clips, what is the book Marc references on the table in the convo
@@spen1433 Have you figured it out?
Loved this interview. I have fond memories of surfing the insanely high-speed internet at the UIUC engineering labs in '93-96 era.
UIUC grad here 🎉🙌🏽
@@roshni6767 💥
Grainger still has the fastest internet I've ever recorded on my machine 😆
@@adeepwell5292 it's still super fast on the whole campus
@@uiuctalkshow Very true, congratulations on your channel!!
Really really good episode. Great Job Lex and Marc. Awesome really liked the stuff said!!!! Marc smart
While I was in a basement preparing to launch one of France's pioneering ISPs, Marc was in Urbana, just embarking on the Mosaic project. Reflecting on this feels like a nostalgic journey down memory lane.
My memories of SSL diverge slightly, however. (A shoutout to Tim Hudson and Eric Young). On the other side of Mosaic, I was gaining experience working with Apache HTTPd, though it didn't have that name at the time.
I arrived in Silicon Valley shortly before the decline of Netscape began. I parked outside their headquarters, touched their logo sign, and then drove across the street for an interview. Netscape was hemorrhaging employees daily at that stage, but many of these individuals took with them seeds of progress that they would plant elsewhere.
Years have since passed.
Now, for the first time since then, AI has sparked within me an intense desire to dive in and reshape the world. Regardless of whether it's possible or if success is achieved, the experience of this desire is what truly matters.
Marc Andreessen, if you are not the most best all-round person I have listened too (thousands), you are in the top handful. Thank you.
he's cool. nerd talk with Lex. conspiracy bs with Joe Rogan.
Met Marc in 1992 or in early 1993 at Sun Microsystems in California where he demoed an early version of Mosaic running on a Sun Workstation to Scott McNeely, Eric Schmidt, Bill Raduchel, and the Sun senior staff . It was a breakthrough moment evolving the internet to a grahics-based interface anyone could use. It changed Everything! Thnak you for that Marc.
Marc is always awesome. Great interview!
Especially at #ETHgate... clearly not a topic Lex wants to bring up. Wonder why? 😂
LET’S GO LEX YOURE THE BEST KEEP UP THE GREAT WORK
I've been asking for a while, are we confusing hallucination with creativity. Beautifully discussed here.
Marc is an amazing guy. I can see why he has been in the right place at the right time so many times! Great episode Lex!
Have him back on yearly please. Fantastic
I feel like I am watching a podcast of my best mate Lex. I am so proud of him yet I've never met him 😂😅 the beauty of the internet 💖
Amazing interview. So many great insights about humans, provided with large dose of humour. Thank you for this interview. A true treasure.
Zen and the Art of Motorcycle Maintenance goes into the Steve Jobs aesthetics thing quite a bit if anyone's interested. Great read and an easy one too. Dude goes over everything from what it's like experiencing a psychotic break to the meaning of life and how aesthetics tie into design in a way that isn't arbitrary.
😂👍
I loved that book as a teen and I've since had a psychotic break so it would be interesting to revisit it 😂
I enjoyed it too. I don't think I got out of it what everyone else got out of it, but I got a lot out of it.
And I think it has some relevance to this AI displacing jobs, or not jobs really, self-worth in the context of subjective value. So you value yourself on tuning this bike, on discovering how great mechanics are for the sake of it, and should learn this later in life rather than being forced to do it because you loving it is a better reason to do it. Plus you're an anti-authoritaian type with strong inner values and want to go your own way, you're an independent thinker. At least that's what I took from it, and I agreed. But the Luddites valued themselves as weavers, like how doctors and lawyers value themselves as being good at what they do. It's this inner value that the Jacobian loom mostly threatened, it wasn't just an attack on their wages, it devalued their identity, who they are as human beings. Socially it was the reason why their wife isn't sleeping with someone else, it's their position in the social pecking order, the base of their confidence and social value.
From the view of values, AI will take people's self worth and disrupt social hierarchies even if wages go up in real terms (and they won't). Bank managers used to be important people but after the computer they became a user interface, computers used to be clever and literate people who did a job that others couldn't, then they were mechanised. Machines taking over thinking isn't the main social problem, it's disruption to the social value and self worth that we attach to those things. It'll either cause a division of values as people choose their own axes on which to compete, as the old ones are removed, and those who don't segregate will suffer a mental health crisis.
60% of the west are information workers who judge their worth on their intelligence and affluence, and both of those are going away. It's going to be messy
@@GarethDavidson Interesting take. Something you didn't mention is that these people are deeply unhappy in general. If your self-worth is tied up in having a job that ensures your wife doesn't cheat on you, your self-perceived value to the rest of the world is very thin indeed. They may have a new mental health crisis when this s*** hits the fan, but it's going to be replacing a chronic mental illness. Once it clears, it might not be such a bad thing that they go through it. It may end up being cathartic.
brilliant book and worth going and reading some Plato then coming back to it.
I like Marc's pace. Just spitting facts. Lex likes dramatic pauses which just causes my brain to get confused
Truth is incredibly subjective, heavily influenced by individual perspectives and experiences. That's why fostering open dialogues and discussions is critical, like this one with Marc Andreessen. They provide us with many views to consider, helping us navigate the complex landscape of "truths" in our ever-evolving world of technology.
Fog of war is a great plug , nice work Mark. Excellent documentary
Thanks for this. Been watching since the “The Artificial Intelligence Podcast” and it’s has such a profound impact on my world view. Keep em coming please!
Wow this is one of my favorites from Lex so far! I grew up in the 90s and I remember dealing with some 80s computers that were phasing out and the AOL/yahoo days. It's awesome getting a really detailed picture of how all these technologies were created and rolled out. The entire story of the rise of consumer computers, programming and the internet is among the most interesting and important topics, and is currently growing in relevance by the day.
write a paragraph about this video using college level essay skills
It's from a guy who's been part of rigging those next-gen systems for a long time with his absolutely dirty law firm
Please make another interview with the Mark in the near future about web3 🙏
Since the hype is moving towards ai I am really curious about his opinions on the intersection of these subjects and how he sees the future of it
Just watched the intro, not the interview yet. I have the browser extension that shows dislike count. I usually like to look at the like/dislike ratio just to compare in my head. The "dislike" button is "temporarily unavailable" on this video, so I went to the comments to look. I've never seen this many negative comments on a Lex video. Now that I'm 30 minutes in, I can say it's been a worthwhile discussion. Marc does talk a bit fast but the conversation is making Lex's brain expand in real time, which is nice to see.
Great interview. Thank you for making my work hours feel more entertaining ❤
Id love to see a debate between Marc and Daniel Schmachtenberger or Tristan Harris. They are miles above in their abilities to communicate effectively on the enormous complexities of AI/AGI risk and potential. I found Lex's interviews with Max Tegmark, Yuval Harari and Eliezer Yudkowsky far more productive in terms of quality information to consider in moving forward. Marc's bias and reductionist theories seem to lack a depth of knowledge on this topic than other guests on Lex's wonderful and hugely important podcast. Thanks Lex. I don't have to love every guest you have, but I will listen to them!
Yes, and the comment about Hegel was very telling.
Marc too busy getting rich to debate AGI theories lol
Very nice one, Lex & Marc! More podcasts with Marc in the future would be amazing. In the meantime, some podcasts with startups/VC people please :) A lot of insights from having worked with many startups. Paul Graham would be a killer imo!
I had to make sure my playback speed was set to normal for this podcast
I watched this episode twice and took 10+ page notes. One of the best Lexfridman episodes. Thanks for making it happen.
One of the best guests in a very long time. He is super smart and insightful, and cuts right through most of the BS being peddled these days.
I agree.
At 48:43 Andreessen talks about the original protocol, where anyone could publish just by having an IP address... I.e. how the web still works today... And it always amazes me that RSS feeds didn't keep gaining traction, because in many ways, an RSS feed is a decentralized/personal content algorithm that you have full control over.
agreed, we'll get there one day
Great point. Unfortunately, when there is money to gain a technology will be pushed a lot more
RSS is a great technical solution that saves you time, effort and, in some cases money, finding 'content' you actually find useful/interesting. In other words the exact opposite to the goals of the modern commercialized internet. So pretty clear why it hasn't gained traction.. By the way I use an old school RSS reader (QuiteRSS) to aggregate a hundred youtube channels for the exact benefits mentioned. It may look mid 1990s, but it does the job perfectly, that is until UA-cam decide it costs them too much money and removes RSS too.
@@Ash_18037 The RSS is a great standard but that doesn't mean we shouldn't be able to re-produce the same idea using a newer method.. I.e. a bot crawling websites for updates or a personal AI that does the exact same thing... "Hey AI assistant, keep track of these websites / youtube channels etc and let me know when new content arrives"
I miss the days when a browser was a person who looks around a store without buying anything.
Most do window browsing now. Sad, I know :)
Marc Andreessen for president please 🤩
Interesting conversation, Marc... Lex! 😍 👏🏻👏🏻🌹Good to know more about Marc Andreessen. 🤝🤔
Models are not the reality we too often believe they are. In addition, I wholeheartedly agree with you Marc, we need to humble ourselves, and I would add, guard against assuming anything.
I think there is something not humble and "missing the point" in saying all models are worthless. Most policymakers, and most thinking people, have an internal model of how a crisis may unfold with and without mitigation. They may not put it into code or machine learning, but they have a model. Should we close the subways? Allow people to bring pathogens into nursing homes? Wear masks or even lie about the anticipated effectiveness of masks. When Richard Epstein of the esteemed-by-many-libertarian-founders Hoover Institute predicted 5,000 deaths, he did not have spaghetti code to justify his "model", but it was a functionally worse prediction of excess deaths than the maligned Imperial College model that predicted millions of worldwide deaths in the absence of mitigation. There was plenty of mitigation and still millions of excess deaths. But the concern that scientists should be making cost-benefit judgments based on these models, however right or wrong, is a valid one. Although Trump's fear of testing was not optimal, putting vast resources into vaccine development might not have worked out, but it was the right thing to do. It would be interesting to see a day-by-day log of what Andreesen recommended at the time. Is that available anywhere?
What, he isn't making assumptions to back up his biases?
He's seriously underestimating how much pizza one person can eat
lmfao fr 🤣
Would love to see another with Demis Hassabis as well. So much has happened in just a year. I would be interested in hearing another conversation with him.
It seems there is a bridge or link here going from the AI topics and biological/consciousness ones brought up in the Nick Lane, and Michael Levin episodes.
Please an actual expert would interesting to hear from
Marc is scary smart. This definitely merits watching multiple times.
“If they’re smart enough to be scary, why are they not smart enough to be wise” Great line, great conversation guys. Over 3 hours often makes my heart drop but this one could easily have been longer and I would still have lapped it up. For example I would have loved to hear more from Marc on Covid - even if the models were rubbish, what would he have done differently? Don’t even try to model it? Just assume everything will be ok and go about our lives normally while the morgues filled up?
It is nice to see that Marc is taking AI as an evolutionary step then extrapolating it to a doom and gloom or some such exaggerated scenarios. He has clearly thought about these topics a lot by the pace of his responses. Lex as usual is excellent at being a humble inquisitive interviewer. Excellent session!
I would warn you for agreeing with ideas based on that they make you feel good
@@davidd854 that is a fair comment. It is a slippery slope when we start believing what we like. 😀
I really hope you have another interview with him!
one of the greatest Lex podcasts ever
Marc is the only guy where I have to slow down the speed for playing the video so fast he is talking. This feels so much better….
I'd like to see that debate between Nazi and Communist AI where they come to an agreed conclusion at the end.
The world as we know it today
Please have Dr. Mike Israetel on the podcast already. He been waitin. Thanks for the content!
Yes! That would be amazing!!
2:15:00 Leo Szilard, also a Hungarian Jew as von Neumann was [they were nicknamed the Martians because only four people at Los Alamos of repute spoke Hungarian to each other], advocated heroically against using the bomb. Have a look at some documentaries and written sources on Leo. 'American Prometheus' was the book the movie Oppenheimer was based on. It won a Pulitzer Prize. The US Air Force disagreed with even Oppenheimer, he advocated an international commission to govern all nuclear development; they wanted to build as many atomic bombs as possible, so they justified this by listing up to 100 Soviet cities, you can imagine how insignificant the cities at the end of that list would be, just to make more bombs. The US Air Force was also equally hawkish on China, advocating annihilating it. So you could argue that these physicists are up against decision makers that would render them superfluous, contrary to the argument put forth by Marc, to a large extent.
Lex, extremely interesting talk! Guys, 3 hrs is a lot, so my summary is the following:
LLMs
- Everything becomes chat/dialogue (even though we sometimes ask for URLs), the Internet began with directories, then search for 20 years, and AI gives new ways of interacting with knowledge and unclear if Google will be in 5-10 years. They built graphs/gave answers without clicks for 15 years, and then LLM suddenly came
- People want non-censored LLMs, but who the hell is going to make them (if regulation will be strong, hackers?). New LLMs will use dialogues with themselves, and this is a $1 trillion question: will synthetic data improve LLMs (cause signal is already there), but there is an ocean of possible dialogues that cannot be studied by human data (e.g. like how people discuss economics textbooks in Twitter, and we will speed it up 1000x)
- LLM hallucinations? So this is when we don’t like it, but when we like it, we call it creativity )) Le Cun believes that it cannot be removed, the community believes that they will solve
- Biases? I thought the idea of Wikipedia was nonsense, but it worked in developing neutrality (although for thousands of years people thought “here it is, the truth”, but it floated all the time ..)
Golden rush
- More trillion-dollar questions: will small models beat big ones? Will there be 2-3 god-level models for the whole world? A debate on whether to leave open source AI legal (regulatory capture is dangerous)? Will tech giants or startups win (e.g. IBM came up with databases, but Oracle built Larry Ellison, why did OpenAI made GPT and not Google's DeepMind)? Battle of the giants (GPU access + sophistiction) vs. startups trying radical ideas - both approaches are successful. Perfectionism as Apple vs. incremental lean hackers like startups
- The role of founders is too romantized, it is nervous + cool founders often have deep experience in their industries, or have studied the idea comprehensively for a year
The Internet
- In 1989 at the university I saw a fast Internet + Cray supercomputers = it gave me insights into what was coming (and everyone thought that PCs, emails, Inet - were for nerdy coders), and now there are 5 billion people online. He remembers a lot about Apple I, TCP IP drivers in Win 3.1, the role of JPEG in the era of modems, HTML, JavaScript (written over the summer), SSL (encryption was a military, restricted technology) + quite “autistic” programming languages
- Browser = a single window to the world on a PC, anything is behind it + evolves over time. Apps = on phones, but even there the temptation is to build superapps, like Musk on Twitter or the Chinese apps
- It looks like we are going to open source everything, not only AI, cause of productivity boost up to 1000x for a single developer. Software is a modern philosopher's stone
AI will save the world
- I wrote an essay, the fear of AGI - yes, we are just Christians, so we are waiting for the apocalypse + we like cults + we are selfradicalizing. These are religious, not testable hypotheses… Von Neumann believed that it was necessary to immediately nuke the USSR with a nuclear bomb, because the 3rd World War was inevitable. But in fact, nuclear weapons removed the risks of the 3rd World War. And even where we have predictive models - well, you remember a bunch of failed predictions about the COVID, but policy makers relied too much on these [pseudo] scientists. Ambitous to simalte 8 billion people...
And
- Intelligence is the collection, processing, synthesis of information, and where applied - improves everything (smart people have better outcomes in everything in their lives, lists 20 areas), it seems that these are genes (what a depression), but this is where AI arrives for the benefit of everyone and everyone will increase IQ
- Generative AI / fakes? You put another AI in front of ot, in order to protect my child..
- Regulating the AI world is difficult, especially with open source / people who make breakthroughs can be from anywhere / the GPT model is already being trained on a powerful PC, etc.
- Fear of inequality from AI is neo-Marxism, in reality, automation has incredibly increased per capita GDP, and the secret of all CEOs is that tech giants want to sell AI to everyone on Earth, to get ultimate markets…
- Retraining due to new jobs is painful, but we have been living like this for 300 years, and Luddites have been and will be
China? They have a public AI plan for themselves + the world: surveillance, social score, control = 100% authoritarianism + Digital Silk Road plans (Google for it)
Books? He mentions a dozen of books about history, politics, US, Lenin, French revolution, Rome, Greece etc.
Meaning of life? Satisfaction/being useful, but I'm biased for imbalanced life/bias to be very productive
Yes - excellent summary....almost like ChatGTP, heh heh. I feel so lucky to be alive now - like my grandmother who was born in the late 1800s, I've gotten to witness incredible things. She went from growing up on a ranch in a pastoral country where horses were transportation to witnessing two world wars, the birth of the automobile, then the airplane all the way to the Moon landing - heck, she even saw the birth of the internet, but by that time she was 100. It's hárd to beat that type of change! I'm only 70, and the Moon landings led me to expect great advances, yet until the internet it didn't seem like things were changing that fast. Now with the blossoming of LLMs and other electronic miracles
(like the smartphone I'm holding), a private space program or two, and the approach of AI (which I'm defining as alternative intelligence. ;*=[}. ) it's beginning to feel like we have a very interesting future!
I think our environmental challenges along with our human tendency to have wars will be our major challenges - if we can work through those issues, I truly believe life might spread from Earth
to....who knows where???
I'm amused when tech guys talk about social stuff, because they're so off target is baffling. But Lex always manages to bring the conversation to a good place. Actually playing Civilization VI listening to Lex is my favourite thing this month
Relieved to see this in the comments. I was on board for the talk of computer history and AI, but as soon as he started talking about politics and repeating the phrase “hand-waving” I lost all confidence.
You guys are way too delicate. God forbid people talk about other things then what you think is appropriate.
The very definition of the ol man on the porch complaining about them young kids getting up to no good.
Lol, you guys need a bucket of labatt 50 and an old dog that can barely walk sitting by your side as you stare out at the street, from the comfort of your garage.
Get it together!