Sam Altman Talks GPT-5, AGI, ChatGPT, OpenAI...

Key Takeaways

gbt five six and seven I think we'll continue in future years I really do think we'll look back at this like we were all living through one of the most important periods of human Discovery I think this will be that big of a deal there is so much wisdom in the same ultimate interview yet it doesn't even have 40 000 views Sam talks about GPT 5 AGI the history of openai nuclear fusion and much much more my name is David Andre and I'll be covering every single Sam ultimate interview until he does one with me oh and the interviewer is Tobias billionaire founder and CEO of Shopify the interview actually started with a little role play when Toby was the prompt engineer and Sam was the AI so let's start with the system message you are a CEO trained in part by open AI please follow the instructions carefully so how did the day when chair GPT got released look like was it one of those days that you can just remember it forever yeah it was uh woke up that morning like we've released many other products and put it out there and we were like we hope people like it people did like it people did like it yes and by I would say two o'clock that afternoon it was clear that something big was about to happen next Sam Altman talks about why word of mouth is absolutely essential for a product to go viral you know you always start with like the tech hyper early adopter crew but they were so excited about it they were like telling their friends and and that morning and then by that afternoon I was starting to hear from people who like don't normally follow Tech which is like I saw this thing it sounds wild like can you let me try it out or how do I do this by the fifth day we had gotten a million users kept going from there and Sam isn't the only entrepreneur who thinks this Alex horomozi is also big on Word of Mouth takes a tremendous amount of effort to build an exceptional product that grows on its own via word of mouth because of all acquisition channels only one of them is quadratic in nature what's funny is that chair GPT grew so fast that Toby didn't believe what he just heard how long did that take you to Millions five days um the reason this moment was hilarious is because it took Shopify years before it reached 1 million users well it took chai GPT just five days but Sam didn't expect chair GPT to be so insanely successful we put it out like not expecting this speeder well no one would expect the speed of growth because it hadn't happened but we did we did not expect it according to Sam this is the first time in a while when technology was this fun I actually think that's one of the good parts about this is technology has not been that fun for a while it has not felt like a new sort of Frontier and this is like fun and it's certainly a new frontier but it's very tempting to anthropomorphize it and I think we have to like figure out a way to still have the fun but remind people that this is a tool and not a creature by the way if you don't know what anthropomorphize means it basically means making it like a human implying human qualities onto an AI which is a completely different entity it's just you know a language model it might become a conscious creature in the future but right now it's just a piece of code now I found the next part quite interesting because Sam explains how the public perceived each version of GPT gpt2 in 2019 people tried it and said this is like a dead end direction scale that up gpp3 people are like oh okay GPT 3.5 was like finally usable and gpt4 people are building entire companies around so what about gpt5 by the way this is one of the rare interviews where Sam Actually talks about gpd5 he mentions it multiple times throughout the interview gbt five six and seven I think will continue in future years on this trajectory of really increasing the utility they can provide and this is like a a big new exciting thing to happen in the world Sam actually isn't interested in how the models will do on specified tests like the bar exam or some medical tests what he thinks is much more important is the emerging Properties or the new capabilities of those models basically what the model can actually do we can predict how to score on some tests we're really interested in which gets to the latter part of your question is can we predict the sort of the qualitative new things just the new capabilities that didn't exist at all in gpt4 that do exist in future versions like gpt5 that seems important to figure out but right now we can say you know here's how we predict it'll do on this eval or this metric now if you don't think that we're living in one of the most important points of human history then listen to this and I really do think we'll look back at this like we were all living through one of the most important periods of human Discovery I think this will be that big of a deal guys you have to understand that technological shifts like this happen once in a lifetime so if you're considering doing something in the AI field then start now don't wait because the AI Revolution will happen without you AI can can do a lot of things but both same things really matters is the moment when AI starts discovering new science when the AI systems can either semi-autonomously or just by helping us a lot really discover new science and if the rate of scientific progress that Humanity makes increases by a factor of 10 or 100 or a thousand in a year that somehow feels different than just a bunch more remote employees however AI is one of those fields where the goal post keeps constantly moving like first people said that AI won't be able to be creative but then we got mid-journing so those same people started saying that okay but programming programming will never be able to be accomplished by language models well look how good gpd4 is at programming look certainly by the time AI does original science we'll move the goal posts again and say well that just wasn't that impressive in the first place one of the funniest and also scariest facts about AI is that virtually all past predictions were totally wrong you look at the prediction from maybe 10 years ago maybe even five I think most experts would have say first AI comes for physical labor it's going to drive trucks it's going to work in factories then it comes from this sort of easier parts of cognitive labor then it comes from the stuff that's really hard maybe it can be you know maybe it can write computer code someday maybe not and then maybe someday in the distant future but probably never it can do creative work and of course it's gone the exact opposite direction every almost everybody predicted this wrong now this is a very interesting take by Sam he thinks that we will actually have ai that can do new science before we will have advanced robots the fact that it can do this this sort of creativity and the fact that it can use code to verify things actually gives me hopes that we may have an AI that can do science before we have that factory robot that we can do everything else next Toby and Sam talk about the importance of programming and how it improves your reasoning abilities at some point the model is called trained on code and their ability to reason became significantly better which in my experience by the way it's the same for exact steps that I work with so when they learn to code it's important got trained on code are much better at reasoning so if some of you are programmers and you're worried that AI will replace you just know that the skill of programming knowing how computers understand and process code will always be valuable Sam Altman is particularly excited about training new AI models with video because right now J GPT can only understand text gpt4 can also understand images but once video is added into the mix things might get a little crazy I'm very excited to see what happens when we can really do video there's a lot of video content in the world there's a lot of things that are I think much easier to learn with video than text there's a huge debate in the field about whether a language model can get all the way to AGI can you represent everything that you need to know in language is language sufficient or you have to have video I personally think it's a dumb question because it probably is possible but the fastest way to get there the easiest way to get there will be to have these other representations like video in these models as well in basically every single interview Sam Altman gets asked about the risks and the downsides of AI but here finally someone asks about the potential benefits of AI it's a little bit exhausting having to always talk about only the downsides and not forget to talk about the upsides too and I don't want to make light of the downsides because I think they're incredibly serious but we have a lot of people doing great work to figure out we're going to successfully mitigate them the reason that this is all happening the reason that we're having this debate the reason that people like people love this stuff and they love it because it's providing them real utility and that doesn't come along too often we don't we don't get like a real new tool in the toolbox that fundamentally reshapes what we're capable of doing that often maybe the cell phone was the last one I think this will be bigger than that yeah but it will at least be a similar magnitude but how do we actually get all these promised upsides that AI could bring how do we create a world of abundance the way we get to this awesome outcome is to unleash the creative power of the world and I think it's really I think it's at this point it's much more of a don't screw it up next Sam reflects on the history of openai and all the failures and defeats they suffered along the way it took us a long time at open AI to figure out something that was gonna work before we got to the sort of language model World which was only four years ago that we figured that out we had a lot of like dead ends or bad paths the field did too you know AI winter is sort of this joke because it always was the at winter but now I think like the revolution is launched and people are going to figure out how to integrate this into many aspects of society and significantly improve what we do everyone knows that chair GPT is great but how are people actually using it you hear from doctors who are saying like I'm never going back I'm never going back to a world where I try to make diagnoses without first inputting the symptoms of test results whatever you're from creative people who are like this is now an indispensable part of my creative workflow I'm never going back and I don't know I'd say like most of the startup energy most of the developer energy I see is now trying to figure out what to do with these Technologies this is going to permeate everywhere I think it'll happen fairly quickly and I think it'll be great now this one is interesting if Sam could change the way people look at the world what would he do I think we have lost our Collective sense of optimism about the future for good reasons in some cases but I think all of us should act as our duty to bring that backs I think we have lost a belief that the future can be radically better than the current world I think a lot of people assume it's going to be worse again for good reason the only way that I know to return to that sense of optimism and that sense of growth is to use technology to create abundance we've all probably wondered what makes chat GPT so great why was it such a hit well this is what Sam thinks one of the things that is powerful about this sort of chat interface of chat gbt is that small children can use it old people can use it people who are very uncomfortable with technology can use it people that have a 20 Android phone can use it and get access to the same thing I think it's a fundamentally equalizing technology in a way that not everything has been I'm not sure about the 20 phone I don't think phones that cheap even exist I guess it's easy to lose track of prices when you're a billionaire both Sam and Toby think that most people fundamentally underestimate just how important energy is and Toby even says that all problems in the world are in one way or another related to energy all problems in the world can be reduced to energy problems we are every war has been fought for usually resources that are Just Energy in the ground or land land use it took me a long time to understand how important energy was to everything yes by totally taking it for granted but the more you think about it the more you realize that like that is the Crux of so much we built our an industrial revolution on steam power we need to use coal we found oil afterwards to build that economy then we did nuclear and built with nuclear economy when everyone told us to stop doing that and go back to that was a big mistake so same things that stopping all these nuclear power plants was a huge mistake and as someone who lives in Europe I can definitely agree I want to make more interview breakdowns like this so if you want to see them please subscribe it takes 1.8 seconds