This video discusses Sam Altman's predictions for the future of AI, specifically focusing on advancements expected in 2025, 2026, and 2027. The video analyzes Altman's insights on the evolution of AI, highlighting key developments in areas such as data handling, agent technology, scientific breakthroughs, and robotics.
so Sam Elman recently reflected on OpenAI's evolution from a 14 person research lab to a dominant AI platform and in his talk at Sequoia Capital he actually gives us an insight to what the future might hold namely the next three years which are of course pivotal years for AI development take a look at some of the changes coming in 2025 2026 and 2027 so one of the things that Samman actually talks about here is the future of how we're going to carry our data around on the internet one of the things that we'll be able to do is potentially sign in with chat dbt as AI becomes more and more available on many different applications it's quite likely that this feature is going to ensure that we can you know carry our preferences around when we use different applications online answer to Alfred's question about where you guys want to go is focus mostly around consumer and being the core subscription and also most of your revenue comes from consumer subscriptions why keep the API in 10 years i really hope that all of this merges into one thing like you should be able to sign in with OpenAI to other services other services should have an incredible SDK to like take over the chat GBT um UI at some point but like to the degree that you are going to have a personalized AI that knows you that has your information that knows what you want to share later and you know has all this context on you you'll want to be able to use that in a lot of places now I agree that the current version of the API is very far off that vision but I think we can get there so now one of the things that Samman actually talks about here is the fact that there actually might still be some incredible breakthroughs left a lot of people think that AI is plateauing and I do think that we are in an S-curve paradigm but Samman clearly says that there are quite a lot of you know breakthroughs that are probably still going to occur in the near future and in the long-term future which means that AI is still going to improve a ton far from what we know now um so you have conviction in the road map about smarter models awesome i have this mental model there's some ingredients like more data bigger data centers a transformer architecture test time compute what's like an underrated ingredient or something that's going to be part of that mix that like maybe isn't in the mental model of most of us i mean that's kind of the each of those things are really hard and you know obviously like the highest leverage thing is still big algorithmic breakthroughs and I think there still probably are some 10 xs or 100 x's left not very many but even one or two is a big deal um but you know yeah it's kind of like algorithms data compute those are sort of the big ingredients uh hi so now this is where we get Samman talking about what is for the rest of the year and of course for 2026 now firstly he talks about the fact that this year is mainly going to be the year of agents which is by far you know the the truth like we've seen agents are super super popular they recently just released Codeex and it's super interesting to see what happens where do you think most of the value creation will come from in the next 12 months would it be maybe advanced memory capabilities or maybe security or protocols that allow agents to do more stuff and interact with the real world i mean in some sense the value will continue to come from really three things like building out more infrastructure smarter models and building the kind of scaffolding to integrate this stuff into society and if you push on those I think the rest will sort itself out um at at a higher level of detail I kind of think 2025 will be a year of sort of agents doing work coding in particular I would expect to be a dominant category i think there'll be a few others too um next year is a year where I would expect more like uh sort of AI discovering new stuff and maybe we have AIS make some very large scientific discoveries or assist humans in doing that and you know I'm I am kind of a believer that most of the sort of real sustainable economic growth in human history comes from once you've like kind of spread out and colonized the earth most of it comes from just better scientific knowledge and and then implementing that for the world and that's crazy because I'm not sure if you're all familiar with the recent innovation that Google had which is of course the alpha evolve paper which essentially made some incredible really incredible maths breakthroughs and it was you know super incredible because for the longest time many people had doubted LLM by saying they cannot create new knowledge they can only combine existing things and it's clear that this barrier has been broken and I think that is so important because if we now know that LLMs can generate new knowledge then I think it just goes to show that there is a lot that we don't know about these systems that we probably ought to investigate and I think it just goes to show as well that there's literally so much more to explore with these systems because it was only recently that we got these breakthroughs and I can only imagine what happens when you know with systems like Alpha Evolve once we get even more capable AI systems running more experiments concurrently and then 27 I I would guess is the year where like that all moves from the sort of intellectual realm to the physical world and robots go from a curiosity to like a serious economic creator of value but that was like an off the top of my head kind of guess right now so that is actually a startling prediction because when we look at that timeline that's 2 years from now 2027 seems like such a long time away but 2 years is a fraction of the timeline when it comes to AI i mean when we actually look at robots being a real creator of value can you imagine humanoid robots doing what humans do now due to many different breakthroughs AGI and super intelligence i mean there are so many companies now building humanoid robots and not just that a lot of these robots are actually functional i know often times we get the odd statement that these robots are only able to do fancy demos but there are real breakthroughs currently being made i've been looking at robotics quite a lot this past week and I've honestly seen many different companies make decent breakthroughs when it comes to having these robots do real tasks we even had Pi Zero recently have their breakthrough with their new foundation model that enables robots to go into homes and you know work in environments that they haven't seen before and we also had another company called Foundation Robotics do something similar where they had their robots you know adapt a new system that basically allowed them to understand real world physics so overall the robotics part is just one key stepping stone and I think once that milestone is truly truly realized and achieved and we get you know those real world tasks becoming more effective and I mean in 2 years that's going to be you know an incredible amount of development because AI and robotics move so quickly as a field i think it's going to be really surprising to see just how far we've come ai has already changed the digital world but to live among us it needs to move beyond the internet data into the chaos of reality so what's missing the ability to reason through our worlds just like we do with an intuitive understanding of physics we learn that from birth by interacting with the world but AI doesn't methods like reinforcement learning and behavior cloning can learn specific tasks but they don't handle new situations well they don't understand how the world works they just copy behavior or learned by trial and error enter latent space models they simplify this messy real world data into abstract maps think a bit of it as AI building its own understanding of reality just like our brain does deep variational base filters or DVBFs take this one step further learning the laws of motion without being spoonfed each and every example dvbf encode sensory inputs like a robot's camera or touch sensors into a latent space and using bijian inference they update their beliefs about a world as new data comes in then encode predictions to act it's a bit like giving AI a sense of imagination unlike behavior cloning or traditional neural networks DVBFs don't just generalize from data they understand the why behind actions they need much less data adapt on the fly and can predict what's next for me that's like AI finally grasping the rules of the game and not just following the playbook and then this is where we have Samman actually saying that by the end of 2025 it's quite likely we will have AI systems that are able to do cognitive tasks to where we are pretty much surprised and overwhelmed by those capabilities but I I will say um by the end of next year end of 25 I expect we will have systems that can do truly astonishing cognitive tasks like where you'll use it and be like hm that thing is smarter than me at a lot of at a lot of hard problems