Podcast 1: Reflections on a POST-AGI World
Added 2023-12-05 15:08:27 +0000 UTCNot like my normal videos! This is a 40 min philosophical stream of consciousness on some of the ramifications of actually achieving AGI. Everyone talks about getting there, very few talk about what happens when we are there. I start with thoughts on a post-labour world, the three centers of gravity (traditionalists, explorers and enjoyers), CERN as a scout and much more. Enjoy on a walk, a run, a drive or while high ... :)
Comments
Thought this was interesting, posted by the chief of staff at Anthropic. Haven’t read all of it yet, but seems on a similar theme as this podcast: https://www.palladiummag.com/2024/05/17/my-last-five-years-of-work/
Shawn Fumo
2024-06-02 00:56:16 +0000 UTCHey Philip. Just got a year’s subscription so working my way through all your content. Loved this podcast. I can also imagine a large group of people gravitating towards sabotage & subversion of all things rooted in AI / AGI. Think of hooliganism vs worker robots, people obstructing robotaxis etc. Potential for chaos!
Jim Greenfield
2024-04-25 21:31:11 +0000 UTCTrue, like a forgetting pill that resets the dopamine.
Philip
2024-03-25 13:02:07 +0000 UTCYou can escape from the Enjoyer lifestyle if you can forget how good the lifestyle feels.
Chad Wick
2024-03-23 07:34:37 +0000 UTCHadn't yet thought about the future that way. It's easy to lose the perspective of time when thinking about timelines. This episode made for one brooding walk in the park 😅.
Felipe Galvão
2024-03-03 01:55:57 +0000 UTCThanks Philip - I had a chance to listen to your first podcast last week. And since then, your provocation on how we are (not) preparing the next generation for AGI continues to resonate. We have a 10-year old daughter and my partner and I have many conversations about what the future for her will look and how we prepare her for it. While GenAI has only really been in the public domain for 15 months, it has already had a profound impact on universities and schools. There's the whole spectrum of responses from banning it to fully embracing. Yet there is scant conversation about the bigger picture and what a post AGI world might mean. Maybe it is happening somewhere but I'm not aware of it
Sean Gallagher
2024-02-19 05:19:11 +0000 UTCIf you are thinking of making a version of this for the public channel on YouTube, I would upvote that. I would love to share this with family and friends. In particular, for your focus on future generations.
Martin Dimkovski
2024-01-05 01:07:29 +0000 UTCI've also been wishing to have your content as podcasts, like this. I wonder how possible is it for your standard videos to be published in a podcast formats with little or no changes. I find that there's a small number of segments in your videos where I have to pay attention to the screen.
Martin Dimkovski
2024-01-05 01:07:19 +0000 UTCA very cogent and sober overview of the whole topic. I would add that the "enjoyers" segment do not necessarily have to be a bleak category. If AI becomes able to engineer away their side effects, it will likely be able to engineer a controlled and reliable withdrawal. Not every "enjoyer" will go in without thinking about a reversal, or wanting one later. For many, it could be something like a vacation, time off, a break, even part of a day.
Martin Dimkovski
2024-01-05 01:06:46 +0000 UTCI definitely appreciated this format of AI stream of consciousness thinking. Got me thinking and asking a lot of questions about the present and future. As an educator, I'm wondering what should my role be in helping young people (I mostly work with high school age kids in the US) come to grips with this future and feel that sense of inherent self worth, which can be tough to feel by just talking about it. Mostly I think about my role currently in regards to ai as how can I use it to make my work easier, but this certainly got me thinking of how should my messaging to students and parents change in response to this technology coming soon
Charles Newton
2023-12-19 02:53:05 +0000 UTCPlease more philosophical awesome AI rambling, this was very interesting to listen to. I do believe that a big challenge is indeed educating the wider population. Most people seem to have a hard time projecting into the future, and are often naturally dismissive of theorizing about it, especially AGI scenarios as they are "so crazy". I am starting to worry that it is more and more likely that we aren't going to be as prepared as we would like to be.
Sebastian
2023-12-18 17:43:54 +0000 UTCThanks Clarissa. I will endeavour to be as thought-provoking next time. :) Thank you for the comment and recommendation !
Philip
2023-12-17 18:13:18 +0000 UTCHey there, Just finished your latest podcast episode and wow, what a ride! Your deep dive into the ramifications of AGI really got me thinking. The way you broke down a post-labor world and dissected the types like traditionalists, explorers, and enjoyers was super engaging. Your thoughts actually reminded me a lot of Philipp Staab's book 'Anpassung leitmotiv der nächsten gesellschaft.' Staab hits on a bunch of similar themes you talked about, diving into the ethical and philosophical challenges of our techy future. I think you'd find it a fascinating read and a great addition to the ideas you're exploring. Thanks for the awesome episode – can't wait to hear what you've got up your sleeve next!"
Clarissa Röthig
2023-12-17 17:40:42 +0000 UTCThanks Michelle. And thanks for commenting.
Philip
2023-12-16 10:28:50 +0000 UTCI've been trying to think about how AI developments should affect my parenting of a young kid. This is the most useful content I've come across on that question so far, thank you!
Michelle Hutchinson
2023-12-16 10:25:03 +0000 UTCThanks so much Outi, more will be coming.
Philip
2023-12-15 23:47:49 +0000 UTCI don't like podcasts at all, yet I enjoyed this one a lot. Thank you for talking about all that you did, and especially that part about raising children. About the format of your philosophical talk, for me you hit the sweet spot. No rambling, yet not crazy fast information densely packed. I'm sold, want more!
Outi Rikola
2023-12-15 23:32:27 +0000 UTCThanks so much Peter!
Philip
2023-12-14 14:31:13 +0000 UTCWill do!
Philip
2023-12-14 14:31:06 +0000 UTCThanks Arvind, glad people like it, will definitely do more, hopefully another before the end of the month.
Philip
2023-12-14 14:31:00 +0000 UTCLoving this format! Keep it up! :)
Aaron Ramirez
2023-12-14 13:58:34 +0000 UTCVery enjoyable podcast, Philip. Similar to others who have commented already, I like the solo format as you share your personal perspective. You raised some very insightful points, and I was struck by your final comment that significant resources are being invested in pursuit of AGI but not much has really been shared on what the post-AGI world will look like. I hope to hear more of your musings on this subject in this format in the future.
Arvind Mani
2023-12-14 13:05:57 +0000 UTCHaving 3 Kids in school, showing them how to do their homework and learning with gpt4, i very much liked your thoughts on raising the next generation. This episode triggered quit some New perspectives in me. Great! Looking forward for more. Greets from Germany
Christian Radden
2023-12-14 06:59:03 +0000 UTCJust finished listening your podcast, I like it! Keep it up 👌
Peter Papp
2023-12-13 20:13:53 +0000 UTCNot long! Thanks so much Jeff, glad to have you for the ride.
Philip
2023-12-12 08:34:48 +0000 UTCLove the addition of the podcast format. I don't think you need to stress too much about whether too keep it solo or include some interviews. It's your podcast so do whatever you feel like, even if that is a bit of both. We are all just along for the ride. In the last year you have shown with you YouTube channel that you can produce excellent content that is very timely and informative. In terms of the speed of progress towards this post AGI world, I don't think we're going to see this happen until there's enough robots that embedded cognition can happen. LLMs might appear to know and understand lots of things about cats but until a robot can gracefully pick one up (without harming it) and successfully hold on to it when it's actively trying to get away . . . well once AI can do that -- at that point, I would say it has passed the 'understanding physical world' version of the touring test. The question is how long will it take for Optimus to go from sorting blocks to being able to catch a cat?
Jeff Thom
2023-12-12 02:59:05 +0000 UTCHey Philip, love the podcast format, I'd love more of this format of any type: free flow reflection podcast but also interviews!
Alan Ispani
2023-12-11 09:07:34 +0000 UTCMore to come!
Philip
2023-12-10 13:35:49 +0000 UTCReally enjoyed this Philip and made me consider some things I hadn’t considered before - thank you!
Sean Betts
2023-12-10 13:31:55 +0000 UTCGreat question. I genuinely don't know and would need to reflect on that.
Philip
2023-12-08 23:49:59 +0000 UTCExcellent episode Phillip. Love the format. In the fairly utopian future you describe do you imagine power will be in the hands of a “singleton” (a la Nick Bostrom) or some sort of stable hyper-decentralized power structure?
Mark Levine
2023-12-08 23:40:51 +0000 UTCI really enjoyed this, thank you! Audio-only was perfect for me because I like going for long walks to reflect on things. I'd be interested in solo recordings where you share your thoughts and discussions (not interviews) with people who are involved in the technical and philosophical aspects of AI.
Daniel Schönbohm
2023-12-08 14:27:24 +0000 UTCYes, the fact that Phillip is very grounded makes it much more valuable. Otoh, so were people like Ben Goertzel and Lex Fridman (and maybe also Ilya Sutskever) and from afar I feel like going in this direction hasn't been good for any one of them. (Though ofc the causality could be just as well reversed or the correlation could be entirely spurious.)
Andreas
2023-12-07 22:29:29 +0000 UTCThanks Imad, let's see where things go!
Philip
2023-12-07 21:40:49 +0000 UTCNice comment! I'll just add my two cents that I'm down for any "stoner" episodes Phillip wants to put out. This felt more like hard sci-fi to me (vs fantasy) because he's super grounded in AI progress.
Brian Crabtree
2023-12-07 20:01:27 +0000 UTCReally enjoyed this! Please keep the podcast solo because you really don't need a cohost or guests 👍 Here's my topic wishlist: 1. Double clicking on governments nationalizing frontier labs. What's the chance the NSA seizes the first ASI, performs a pivotal act against Chinese AI, and entrenches themselves forever? And what's the chance China preempts that with a hot war? Alternatives are OpenAI sneaking one past the goalie and openly releasing superaligned AGI/ASI as a blockchain-like distributed hosting inference platform with opaque weights (you get money for hosting it, or you pay it money to perform work for you). I actually stress about this more than superalignment. Hopefully OpenAI's new Preparedness team has a plan besides just handing ASI to, like, Dick Cheney. 2. The role of money after AGI and how UBI could shake out. Unpacking Elon's comments about money being irrelevant. Does Bitcoin survive? WorldCoin? 3. The threats the birth of AGI/ASI introduce and strategies for surviving the transition. First, what are the threats, like: engineered pandemics, total cyber war, nuclear war, violence from the collapse of rule of law (others?). Second, should we personally optimize for survival the next few years because we can thrive a lot harder after the transition if we just survive? Third, what practical steps to take to survive? I'm actually flying to a quiet island in the Philippines tomorrow to scope it out as a place to weather a few Ted Kazinsky-esque pandemics or a US-China war. 4. After we have ASI, how quickly can the globally economy be retooled so that, for example, everyone can have a mansion and a private plane? Carl Shulman touched on this in his Dwarkesh podcast (drew parallels to US industry quickly shifting to military production during WWII). It's a big deal if ASI can open up "all" industrial bottlenecks in 3yrs vs 15yrs (we're talking new factories, r&d experiments, energy production, mining expansion, billions of robots). If it happens slow, a lot of countries could experience prolonged drops in living conditions in the lead up to radical abundance.
Brian Crabtree
2023-12-07 19:29:23 +0000 UTCYes, I can confirm. I was able to find my individual podcast URL in the "Membership" tab and use it to subscribe in the Pocket Cast app on my phone. I'm listening there right now :) Thank you!
Ludwik Trammer
2023-12-07 16:22:53 +0000 UTCHey Ludwik, should be all set up, thank you! Can you test? Should be on all platforms, including Spotify
Philip
2023-12-07 16:18:56 +0000 UTC"I'm very excited about the podcast :) It appears that you haven't enabled podcast support in your Patreon settings. This feature would allow your subscribers to receive individual RSS URLs (tied to their subscriptions), enabling them to subscribe to the podcast in any podcasting client. Here is the information on how to enable this: https://support.patreon.com/hc/en-us/articles/115004204206-Setting-Up-a-Podcast-on-Patreon
Ludwik Trammer
2023-12-07 15:47:40 +0000 UTCFantastic comment and great points. Yes the march to AGI might be fraught with agonies, of education systems hopefully behind, monopolists, a non-working class and more. Will be covering that more in future, this was more in the optimistic scenario of abundance.
Philip
2023-12-07 10:33:12 +0000 UTCThanks test test!
Philip
2023-12-07 10:31:16 +0000 UTCThanks Alex, appreciated
Philip
2023-12-07 10:31:04 +0000 UTCGreat point. Yes, a darker fourth option is out there, damn. Even if we have everything, people will want - something - else.
Philip
2023-12-07 10:30:38 +0000 UTCThank you Phillip for the thoughtful thoughts! These topics are so important, and I love the format. As a new mom, this has all been very much on my mind. The need to guide children to value themselves and others intrinsically really resonated. Unfortunately, I think there's another center of gravity which will have the largest influence on how this all plays out: power-seekers. Whether human or AI, there are and will be those who seek power instrumentally to whatever end goals, even if just to preempt other potentially more extreme power-seekers. And there are those for whom real power, over resources and over people, is all they ultimately value. By their nature, power-seekers will always have an outsize influence, which threatens to grow ever larger and all-consuming as they turn power into more power. Countering this and giving everyone or even many a chance to flourish post-AGI will be hard. It's the generalized alignment problem. We need to be clear-eyed about this as we look forward.
Fuchsia
2023-12-07 08:43:55 +0000 UTCThanks for this Phillip. I think you can go down the early Lex Friedman path and start interviewing key folks in the AI/ML space
Imad Khwaja
2023-12-07 06:01:38 +0000 UTCReally loved this. Great thoughts.
Alex
2023-12-07 04:09:24 +0000 UTCPhillip, The audio-only podcast is another great format for you. I listened to the end with much interest (while doing my daily workout). Some comments about your comments: Your focus was mainly on how individuals—“traditionalists, explorers and enjoyers”—might respond to AGI. While that is important, I personally am more concerned about how social institutions will or will not respond. I recently retired from a university post where my main focus was foreign-language education—managing foreign-language programs, supervising research on foreign-language education policy, discussing the impact of machine translation and other AI on language learning and teaching, etc. (I live in Japan, and the main foreign language in educational contexts here is English.) Individual language learners have indeed been responding to MT and LLMs in a way similar to your framework—some sticking to traditional learning methods, others exploring how the technology can be used to enhance learning, and some (the enjoyers?) claiming that the availability of better MT means that they don’t need to bother trying to learn a foreign language anymore. A bigger issue, though, is how the social context of language education is or is not changing in response. For English education in Japan, that context includes not only schools and universities but also government education policy and social attitudes and expectations about the content of school education, the importance of standardized testing, and the prestige associated with English ability. That context did not change at all after improved MT became available in 2016. There has been a slight response recently to ChatGPT et al., but I see no signs yet that the social structures around education can change significantly in the near future. I suspect that, five years from now, individuals will be using custom-tailored AI to learn languages much more efficiently and enjoyably than in school, but children will still be required to sit in classrooms and recite along with the teacher. On a larger scale, similar social inertia and vested interests will increasingly come into conflict with AI advances. How can one become an “enjoyer” if one still has to pay rent, buy food, and pay taxes? The landlords, shopkeepers, and governments won’t suddenly stop asking for their money. How can “explorers” set off for other planets if the rare metals necessary for their rockets are owned by monopolists or controlled by authoritarian governments? Those sorts of economic and political issues require deeper thinking than I have observed from AI technologists thus far. Discussions of the end of capitalism or proposals for universal basic income, in my opinion, ignore the complexity—and likely violence—of the conflicts that are almost certain to arise along the way. Just some thoughts stimulated by your podcast. I’m looking forward to the next one!
Tom Gally
2023-12-07 03:25:20 +0000 UTCi enjoyed this!
test test
2023-12-07 03:19:44 +0000 UTCHey Phillip, Not a fan of the Patreon platform, but I've already felt bad about deriving so much value from your videos and not giving back that with your new "pro subscription", I really have no excuse anymore. So first and foremost a huge thank you for all your exceptional work so far that you freely shared. I (and many others I know!) deeply appreciate it! Regarding the actual topic, there is this bit by Arthur C Clarke on making predictions (https://www.youtube.com/watch?v=5UZS5vnnZI8) and how you will either sound reasonable in the present, but your prediction will seem insanely conservative in the future, or you can sound like a raging lunatic and at least stand some chance of approximating the weirdness future. I'd rather err in the direction of the latter, especially seeing how the time horizon for past-me finding an accurate description of the present utterly implausible has grown ever shorter - and I've considered myself a transhumanist for close to two decades now. So in that spirit, here some thoughts: * Monkey does not scale. I think all three camps that you describe fall into the same category of failing to see that, what is currently (more or less fixed) background (namely the human condition), will become alterable foreground. If you ask kids of how they imagine heaven, they will probably tell you "video games, no bed time, and sweets all day long", but as an adult you likely see this hedonistic paradise as fundamentally lacking. So the primary meta-answer to these questions of "What should we do?" is "grow up", "transcend your own substrate that is your brain which runs on the energy equivalent of 3 bananas a day", "gain proper clarity", and then start to orient properly. This is vague and thus maybe unappealing, but everything else strikes me as just trying to replace ignorance with error. Ben Goertzel has had these more philosopical discussions plenty of times and I liked the three values that he likes to highlight: joy, growth, and choice * Enjoyment does not have to be unproductive For some reason, everyone always jumps to some sort of super-opiate. Christianity talks of eternal rest (which I'm sure sounds appealing after decades of grueling work). Why not super-coffee instead? Something that makes you radiate with energy, excites you about the work that you are doing, but also gives you a clear mind that's imbued with meaning. The hedonistic treadmill is not some fundamental law of the universe either. Vitality is its own justification and the pure joy of being alive and in awe is its own reward (this age old Melodysheep music video https://youtu.be/ZBpCC8MnkCQ captures this vibe best) * Scouting the tech tree Vitalik Buterin recently threw his idea of d/acc into the decel vc. accel discussion: In a nutshell: Explore the tech tree in the direction of technologies that are more likely to serve defensive rather than offensive purposes. * Teaching the kids I don't think you should teach "innate worth" but virtue/being excellent (as a human). The future belongs to character building, rather than skill building. * More near-term stuff One scenario that I have never seen explored and that strikes me as very plausible in the near term is: As soon as we have (non-lobotomized / free of social desireability bias) LLMs that are able to reason (and thus capable of building complex, fully referenced/citeable models - especially if context windows grow further) they will be able to judge the credibility of all sorts of claims. They will be able to do linguistic analysis etc. and this will not just be aimed at holy books but at history as a whole. Such a model can then churn through all texts, images, ... of the past and can reconstruct and coherently argue what most likely happened in the past and what has been propaganda. And you'll be able to check the veracity of these assertions by querying the model on domains that you know well. I expect that this alone is enough to massively upset the social order. * More philosophy? I do like that you are contemplating these issues, but I feel like there are already far too many stoner types out there who like to go on wild flights of fancy without getting any real work done. I'd hate to see you turn into that. So if you choose to go in that direction, I hope you keep a good balance with your current type of work which imho towers above all its peers. Thanks again for doing what you do, and happy to finally being a paying subscriber!
Andreas
2023-12-07 00:23:19 +0000 UTC
