OpenAI CEO 炸裂发言：请把电脑赶出幼儿园！AI 时代的 10 条生存法则

首发于微信公众号「田野与浪潮」 · 阅读原文 ↗

这是 OpenAI CEO 山姆・奥特曼（Sam Altman）在 2026 年 1 月与开发者和创业者进行的 Town Hall（全员会） 问答环节的文字实录。笔者侧重教育和学习进行了总结。

1. “学编程”不再是唯一的免死金牌 (The End of "Code as Destiny")

过去十年，教育界最大的共识是“让孩子学编程”。但 Altman 明确表示，这不再是绝对正确的路径。

• 原文观点： “曾经‘去学编程’显然是绝对正确的选择，但现在情况变了。”
• 解读： 并不是说编程毫无用处，而是单纯掌握“写代码”这项硬技能，已经无法保证未来的饭碗。AI 将把代码变成一种廉价的基础设施。

2. 未来最重要的四项“软实力” (The New Core Competencies)

如果代码由 AI 来写，人类需要拼什么？Altman 列出了四个具体的替代技能：

• 原文观点：高能动性 (High Agency)：主动寻找方案、解决问题的意愿。产生好想法的能力 (Generating Ideas)：当执行成本降为零，想法的质量就是一切。韧性 (Resilience)：面对失败和挑战的复原力。适应力 (Adaptability)：适应这个快速变化世界的能力。引用来源：
• 解读： 教育的重心必须从“技能培训”转向“性格与思维培养”。这些都是可习得的技能，甚至可以在三个月内通过高强度的训练营获得显著提升 。

3. 幼儿园不需要 AI，甚至不需要屏幕 (Kindergarten Should Be Analog)

对于低龄教育，这位 AI 巨头反而持极其保守的态度。

• 原文观点： “我支持把电脑赶出幼儿园。幼儿园应该是通过在户外奔跑、玩实体玩具来学习如何与人互动的。”
• 解读： 在我们甚至还没完全搞懂社交媒体对青少年的负面影响前，不要急着让幼儿接触大量 AI 。过早的数字化可能并不是优势，真实世界的社交互动才是地基。

4. 对 AI 建设者来说，大学可能在“浪费时间” (University is Optional for Builders)

对于那些极具野心、想要在这个特殊时期有所建树的年轻人，按部就班读完大学可能意味着错失良机。

• 原文观点： “如果你是一个 AI 建设者（AI Builder），现在待在大学里可能并不是时间利用的最优解。”
• 解读： 这是一个特殊的历史窗口期 。对于高能动性的人才，实战优于学历。当然，Altman 也补充说，如果你后悔了，随时可以回去读书，但这波浪潮不等人 。

5. “好想法”将成为最稀缺的资源 (Ideas > Execution)

在生产端，AI 解决了执行效率，人类的瓶颈变成了“创意的贫瘠”。

• 原文观点： “想出好的新点子非常难……而在生产端，对于所有的建设者来说，限制因素就是想法的质量。”
• 解读： 未来的教育不应奖励“按部就班完成任务”，而应奖励“提出惊人的点子”。因为在 AI 的加持下，一旦有了好点子，尝试和验证的反馈循环将极度缩短 。

6. 写作是为了学会思考，而不是为了输出文本 (Writing is for Thinking)

即便 ChatGPT 能帮你写论文，你依然要学写作。为什么？

• 原文观点： “你仍然需要学习思考，而写作、学习写作或写作的练习，对于学习如何思考至关重要。”
• 解读： 不要把写作看作一种“生产文本”的工作（这部分 AI 可以代劳），要把它看作一种“整理思维”的大脑体操。教育需要改变考核方式，但不能废除写作训练 。

7. 越是 AI 时代，人类的连接越值钱 (The Human Premium)

当 AI 生成的内容泛滥时，“人味”成了奢侈品。

• 原文观点： “在未来的几十年里，这将是一个深刻的趋势：我们非常在乎其他人，而对机器却不怎么在乎。”
• 解读： 哪怕 AI 创作的画作更精美，当人们知道那是 AI 画的时，欣赏度会立刻下降 。教育应鼓励孩子去讲故事、去建立个人品牌、去展现脆弱和真实，因为“你是谁”比“你做了什么”更重要。

8. 学会与 AI 协作：做“无限研究员”的导师 (The Unlimited Grad Student)

未来的科研和学习模式，是人类提供直觉，AI 负责“广度优先搜索”。

• 原文观点： 科学家正在将 AI 视为“无限的研究生”或“无限的博士后” 。人类负责直觉判断（Intuition）和方向指引，AI 负责繁重的验证工作 。
• 解读： 未来的学霸，不是那个算题最快的人，而是那个最懂得如何指挥一千个 AI 助手同时探索二十个方向的人 。

9. 适应“个人定制化软件”的时代 (Hyper-Personalized Tools)

未来的孩子将不会使用统一的“Word”或“Excel”，软件将是为每个人量身定制的。

• 原文观点： “我不再认为软件是静态的……如果我有个小问题，我期望计算机立刻写一段代码帮我解决。”
• 解读： 学习使用标准化工具的能力将贬值，理解问题本质并让 AI 生成专属工具解决问题的能力将升值。

10. 安全教育：从“封堵”转向“韧性” (Resilience over Blocking)

在 AI 安全和生物安全教育上，试图屏蔽风险是徒劳的。

• 原文观点： 我们需要像对待火灾一样对待 AI 风险。不是禁止用火，而是建立“防火规范”和增强社会的“抗火韧性” 。
• 解读： 教育孩子如何识别 AI 的幻觉、如何防范 Deepfake、如何在信息过载中保持清醒，比单纯禁止他们使用 AI 更重要。

Sam Altman 的这场分享撕开了未来的一角：

在一个执行力（Coding/Execution）变得极度廉价的时代，教育的溢价将重新回归到人性的深处——你的愿力（Agency）、你的创意（Ideas）、你的思考质量（Thinking）以及你与他人建立连接的能力（Connection）。

不要再逼孩子成为一个更好的“工具人”，因为机器已经赢了。教孩子成为一个更鲜活、更具领导力、更懂得提问的“人”。

附上原文分析

OpenAI Town Hall with Sam Altman

Summary

Sam Altman addresses questions from developers and entrepreneurs about the future of AI and its impact on various industries. He predicts AI will be massively deflationary, empowering individuals and equalizing economic opportunities, while also acknowledging the risk of concentrated power. Altman envisions AI transforming software engineering, leading to customized software experiences and ubiquitous AI agents. He stresses the importance of focusing on general-purpose models and the need for advancements in coding intelligence. Other topics include the role of AI in scientific research, biosecurity, education, and the balance between human creativity and AI-assisted creation. He also shares insights on OpenAI's development roadmap, emphasizing memory, personalization, and the integration of AI in collaborative environments.

Takeaways

• The increasing power and convenience of AI tools may lead to widespread adoption without sufficient consideration of potential catastrophic failures and security vulnerabilities.
• AI's deflationary pressure has the potential to equalize economic opportunities, empowering individuals regardless of their background, provided that appropriate policies are in place to prevent wealth concentration.
• The most valuable skills in the age of AI are adaptability, resilience, and the ability to generate ideas, all of which are learnable and crucial for navigating a rapidly changing world.
• The key to building a durable startup in the age of AI is to create something that benefits from model improvements rather than being threatened by them, focusing on solving the go-to-market problem and building a competitive advantage.
• The future of computing involves a shift towards customized, constantly evolving software tailored to individual user needs and preferences, moving away from static applications.
• Human connection and collaboration will become more valuable in a world of abundant AI, with AI-enabled tools facilitating more productive group interactions and problem-solving.
• While AI can assist in scientific research, human creativity, intuition, and judgment remain essential for making breakthroughs and guiding the research process.
• The world needs to shift from blocking AI-enabled threats to building resilience, similar to how society manages fire risks through safety codes and flame-resistant materials.
• People's appreciation for art is significantly higher when they believe it was created by a human, highlighting the importance of the artist's story and personal connection in the creative process.
• OpenAI aims to provide GPT 5.2-level intelligence at a hundredfold cost reduction by the end of 2027, while also focusing on delivering faster output speeds to meet user demands for real-time AI assistance.

Q&A

Q: How can AI be used to solve economic gaps that have existed for decades, considering women leave about a million dollars on the table due to the wage gap?

A: AI is going to be massively deflationary, which is mostly good news. With progress in work done in front of a computer and robotics, we're going to have massively deflationary pressure in the economy. Things will get radically cheaper, empowering individual people, regardless of whether society has structured in a way where they've naturally had all the advantages. By the end of this year, for a hundred or a thousand dollars of inference and a good idea, you'll be able to create a piece of software that would have taken teams of people a year to do. This economic change should be a very empowering thing for people, leading to more abundance, access, and decreased costs to create new things, new companies, and discover new science. It should be an equalizing force, giving those not treated fairly a really good shot, as long as we don't screw up the policy around it. I am worried that AI could concentrate power and wealth, and policy needs to prevent that.

Q: Given that startup features can be replaced by model updates, how should builders think about durability, and what is the one layer of the stack you promise not to eat?

A: It's tempting to assume the laws of physics for business have totally changed, but they haven't yet. What's changed is you can do work and create new software much faster. All the other rules of building a successful startup still apply: figure out a way to get users, solve the go-to-market problem, provide something sticky, and have some sort of moat, network effect, or competitive advantage. None of that has changed for us either. There have been many startups that have done things that maybe in a perfect world we would have done sooner, but it was too late, and people built up a real durable advantage, and that will keep happening. The general framework I give people is, will your company be happy or sad if GPT-6 is a wildly impressive update? I encourage people to try to build things where you are hoping the model gets wildly better.

Q: What tools can you build to improve the quality of ideas that people are coming up with, given that human attention is the rate-limiting factor on the consumption side and the quality of ideas on the production side?

A: It's very hard to come up with good new ideas, and I am increasingly a believer that we think at the limits of our tools. I think we should try to build tools that help people come up with good ideas. As the cost of creation keeps plummeting, we'll be able to have such a tight feedback loop on trying them, we'll find the good ones faster. As AI can discover new science in addition to writing very complex code bases, I feel confident there will be just a whole new possibility space. The experience of sitting down in front of an AI and just not being sure what to ask for next is something that a lot of people report. If we can build tools to help you come up with good ideas, I believe we could look at all your past work and all your past code and try to figure out what might be useful or interesting to you and can just continuously suggest things. If we can help provide a really great brainstorming partner, even if most of them are bad, something like that is going to be a very significant contribution to the amount of good stuff that gets built in the world, and the models feel like they ought to be capable of that.

Q: How do you foresee education changing and being shaped by AI during those really formative years, like kindergarten and middle school?

A: Generally speaking, I'm a fan of keeping computers out of kindergarten. Kindergarten should be like running around outside and playing with physical things and trying to learn how to interact with each other. So not only would I not have AI in most kindergartens, most of the time, I wouldn't put computers either. Developmentally, we still don't understand all of the impacts of technology. There's been a lot written about the impact of social media on teenagers, and that seems like it's been pretty bad, but I have a sense that unfortunately a bunch of technology on young children has been even much worse, and it's still talked about relatively little. I think until we understand that better, like probably we don't need kindergartners using a ton of AI.

Q: As agents start moving and operating more production systems at scale, what do you think is the most underestimated failure mode, and where do you think the really hard work is being under-invested in right now?

A: The general worry I have is that the power and convenience of these are so high, and the failure rates are maybe catastrophic. Failures when they happen are maybe catastrophic, but the rates are so low that we are going to kind of slide into this like, you know what, YOLO, and hopefully it'll be OK. As the models get more capable and harder to understand everything they're doing, if there's a misalignment in the model, if there's some sort of complex problem that emerges over weeks or months of usage, and you kind of put some security vulnerability into something you're making, the pressure to adopt these tools to use them is going to be so great that people get pulled along into sort of not thinking enough about the complexity of how they're running these things. The general worry I have is that capability is going to rise very steeply. We're going to get used to how the models work at a certain level and decide we trust them, and without building very good big picture security infrastructure around it, we will sleepwalk into something.

Q: What do you think is the most important skill that people should learn in the age of AI?

A: These are all kind of like soft skills. None of them are like as go like learn to program was so obviously the right thing, you know, over recent period of time, and now it's not, but skills like become high agency. Get good at generating ideas, be very resilient, be very adaptable to a rapidly changing world. I think these are going to matter more than any specific and I think these are all learnable. This is one of the surprises to me of having been a startup investor is the degree that which you can like take people and in a three-month sort of boot camp style thing make them extremely formidable. And do the things on all those axes I was just talking about is very surprising. It was a big update to me. And, and so I think these are the skills that may matter the most. And they're they're like quite learnable.

Q: Realistically, what is the timeline for agents that can run long workflows autonomously without constant human intervention, given that even simple on-chain tasks often break after five to ten steps?

A: It really depends on the kind of task. There's a number of tasks today where just inside OpenAI, we see people who are prompting codecs in a very special way. Maybe they're using the SDK. So it's like a custom harness that keeps prompting it to continue, but they basically having it running, you know, forever. And so I think this isn't a question of when, but it's a question of like broadening of the horizon. So if you have a very specific task that you understand very well, try doing it today. You know, if you're starting to think like, okay, I want to get to the point where I can prompt the model to build a startup, like that's a much more open-ended problem with like a much harder verification loop. So I would encourage you to like figure out, okay, how can I break that down into a different problem where an agent can like verify itself? And then when I can verify its final output at the end of it, and then over time, we can let the agent do more and more and broader tasks.

Q: How should builders think about durability when start-up features can be replaced by model updates? What is the one layer of the stack you promise not to eat?

A: It's so tempting to assume that the laws of physics for business have totally changed, and they haven't yet. They may continue to change over time, but right now, what's changed is that you can do work faster and you can kind of create new software much, much faster. But all the other rules of building a successful startup, you know, you got to figure out a way to get users, you got to figure out a way to like solve the GTM problem, you got to figure out a way to provide something sticky, have some sort of mode network effect, competitive advantage, whatever you want to call it. None of that has changed. And the good news is like, It hasn't changed for us either.

Q: As you look into the future, how do you think about models being specialized versus generalized?

A: I think we just screwed that up. We will make future versions of GPT 5.x hopefully much better at writing than 4.5 was. We did decide, and I think for good reason, to put most of our effort in 5.2 into making it super good at intelligence, reasoning, coding, engineering, that kind of thing. And we have limited bandwidth here and sometimes we focus on one thing and neglect another. But I believe that the future is mostly going to be about very good general purpose models. You know, even if you're trying to make a model that's really great at coding, it'd be nice if it writes well too. Like if you're trying to have it be able to generate a full application for you, you'd like good writing in there. When it's interacting with you, you'd like it to have a sort of thoughtful, incisive personality and communicate clearly. Like good writing in the sense of clear thought, not like beautiful prose. So my hope is that we just push to get future models really good in all of these dimensions and I think we will do that.

Q: How do you think about small models, costs, and dramatic cost reductions for developers over the next months and years?

A: I think we should be able to deliver sort of GPT 5.2 X high level intelligence by the end of 2027 for at least a hundred X less. But there's another dimension which we haven't thought about as much historically and now as these model outputs get so complex, more people are pushing us on the speed we can deliver it at than the cost. And that is, we are really good at writing down the cost curve. You can look at the progress we've made even from like the first O1 preview until now. We have not Thought as much about how we deliver the output, the same output and maybe at a, at a much higher price, but in one 100th of the time. And I think for a lot of things you're talking about, people are going to really want that. Uh, we have to figure out how we're going to balance between prioritizing for those two things.

Q: What innovations in customized agentic interfaces could further accelerate the trend towards microapps?

A: I have noticed in my own use of codecs recently is I no longer think of software as this static thing. If I have a little problem, I expect the computer to like write some code right away and get it solved for me. And I think this trend is going to go much further. I suspect that the whole way we use computers and operating systems is going to change. I don't think it'll be like, oh, every time you need to edit a document, a new version of a word processor is going to be written for you right on the spot. Because, you know, we get, like, very used to our interfaces and it's very important that, like, that button is in the same place it was last time. But For a lot of other things that we do, I think we will find that we expect software to be written just for us and when maybe I want to use the same word processor every time. But I do kind of have a bunch of repeated quirks of how I use it. And I would like the software to be increasingly customized, you know, a static or a slowly evolving piece of software, but written for me.

Q: What are your thoughts on like more granularity, like, for example, grouping memory, like, this is my work identity, this is my personal identity, so that way, as you're doing different prompts, you can sort of be more selective about what you want to be included on it?

A: We're gonna push super hard on memory and personalization. Clearly people want it and it delivers a way better way to use these tools. I I have gone through my own evolution here, but at this point I am ready for ChatGPT to just look at my whole computer and my whole internet and just know everything. The value from it is so high and I don't feel uncomfortable about it in the way that I used to. I really hope, you know, all AI companies take security and privacy super seriously and I hope that society as a whole does too because The utility is so great, like AI is going to know about my whole life. I'm not going to get in the way of that. I don't yet feel ready to like wear the glasses recording everything. I think that's still uncomfortable for a bunch of reasons. But I do feel ready to say like, hey, you can just have access to my computer and figure out what's going on and be useful to me and understand everything and like have a perfect representation of my digital life.

Q: How do you see kind of this relationship between human creative identity and also AI assisted creation is going especially as these models continue to advance?

A: The place that we can study and I think learn the most right now is ImageGen. It's been around the longest. The creative community has used it and disliked it and liked it the most. There's many interesting observations there, but one of them is that consumers of images report dramatically higher appreciation, satisfaction, whatever, if they are told a person made it versus an AI. And I think this is going to be a deep trend in the coming decades is we care a lot about other people and we care very little about the machines.

Q: Do you think university could be, in its current state, limiting sometimes?

A: Everybody's got to make their own decision, but I think you do need to make your own decision and not just do what society tells you to do. And I don't personally think this is a time where from If you are an AI builder, it is probably not the best use of your time to be in university right now. If you're just like a sort of ambitious high agency driven person, this is an unusual period of time. And, you know, you can always go back later and I think just tell your parents that you're not, like, it doesn't mean that it's not the right thing for many people. It doesn't mean that it won't be the right thing for you sometime. But like right now you got to do this thing and I think they'll understand eventually.

Q: How did the interview have changed in past months or years?

A: We're going to keep hiring software developers, but we are for the first time, and I know every other company, every other startup is thinking about this too. We are planning to dramatically slow down how quickly we grow, because we think we'll be able to do so much more with fewer people. And I think at this point, a lot of the impediments that we face or that other companies face is just like the internal policies that have built up at most companies did not contemplate a majority AI co-workers. And that's going to take a while. But what I think we shouldn't do, and what I hope other companies won't do either, is hire super aggressively, then realize all of a sudden, AI can do a lot of stuff and you need fewer people and have to have some sort of very uncomfortable conversation. So I think the right approach for us will be to hire more slowly, but keep hiring and trust that I'm not a believer that eventually, well, maybe someday far in the future OpenAI has zero employees, but for a long time, I think we'll just have a lot. We'll have a gradually increasing number of people doing much more stuff. And this is kind of what I expect the shape of the economy to look like more generally.

Q: I'm curious, as to what you think about that bottleneck of GTM, right? That I can build these things, but how do I find people to gain value from what I'm building?

A: Before this, I used to run Y Combinator. And the consistent thing you'd hear from startup founders is, I thought the hard part of this was going to be building a product, and the hard part is getting anyone to care or to use it or to connect people. So I think this has always been extremely hard, but now it's gotten so much easier to build that you feel the delta even more. I don't have an easy answer for this. I think always it's been hard to build a business and to figure out a way to build differentiated value, to get the go-to-market function working. I think all of the old rules still apply there. The fact that AI can kind of make it far easier to create software doesn't mean any of the rest of this gets easier. Except, I think you're starting to see in the same way that AI has transformed software engineering, people are now using it to automate sales, automate marketing, and there will be some success there. But I think it's always going to be difficult because even in a world of incredible abundance, human attention remains like this very limited thing.

Q: I'm wondering as a builder building on the Codex SDK, Am I safe? Like, do you see room for a lot of different types of UIs for multi or multi agent orchestration? Or do you see like OpenAI owns that stay away?

A: No, I think we don't know what the right interface for all of this is going to be. We don't know how people are going to want to use it. We have seen people build incredible multi-agent setups. We have seen people build just a really great single interactive thread. We're not gonna figure this out on our own and also not everybody is gonna want the same thing. You know, there will be people that'll be like in one of the, they're in one of those like old movies with 30 computer screens and they're like looking at this crazy thing here and doing this and you know moving stuff around. And I think there will be people who want to be in a very calm conversation in voice mode where they're saying something to their computer once per hour. The computer's figuring out a lot of things and they're not, you know, they're trying to think really hard about what they say, but they don't want like the constant supervision of tons of agents.

Q: What do you need? Do you want people to like be able to bring their own token budget or do you want them to like bring their ChatGPT memories or all?

A: We do want to figure out how to do this. It's very scary because ChatGPT does know so much about you. And if you tell a person that you're very close to a bunch of secrets, you can be relatively confident they'll know the exact social nuances and when they share what with who and when something overrules something else. Our models are not quite there, although they're getting pretty good at it. I would I think feel uncomfortable if I connected my ChatGPT account to a bunch of sites and said, just use your judgment about like when to share what you know about me from all of my chat history and everything I've connected. But when we can get there, it will clearly be a cool thing to offer. And in the meantime, I think doing something just with, you know, token budgets, and if I pay for the pro model, then I can use it on other services, that seems like a cool thing to do. So I think we will at least do that. And we'll try to figure out a way to get the information sharing right. But like, we really don't want to, we really don't want to screw that up.

Q: So I want to go back on this like human attention and this question about GTM. I was thinking like the other side, like human attention is the rate limiting factor on the consumption side. On the production side for all the builders, it's the quality of ideas.

A: There's a lot of It's popular to like call AI output slop, but there's a lot of like human generated slop in the world, too It's very hard to come up with good new ideas, and I am Increasingly a believer that we think at the limits of our tools I Think we should try to build a tools that help people come up with good ideas. I believe there are a lot and I believe as the cost of creation keeps to plummet, we'll be able to have such a tight feedback loop on trying them, we'll find the good ones faster.

Q: Do you think we'll be able to steer the models enough to get them to use new things or are we just done improving the technologies we build on now?

A: I think we really will be very good at getting the models to use new things. Fundamentally, if we're using these models correctly, they're like a general purpose reasoning engine. The way we have things architected right now, they also have, you know, a huge amount of world knowledge built into them. But I think we are moving in the right direction. And I hope that updates and using new things and learning new skills even faster than humans do, is like a, you know, next couple of years thing. The A milestone that we will be very proud of is when the model can be presented with something totally new, new environment, new tools, new technology, whatever. And you can explain it once or you don't need to explain the model can explore it once and then just super reliably use that and get it right. And that doesn't feel that doesn't feel very far away.

Keywords

Highlights

<aside> 💡 (00:06:32) The overhang of what these models are capable of relative to what most people can figure out how to get out of them is huge and growing.

</aside>

<aside> 💡 (00:09:08) Massively more abundance and access and massively decreased costs to be able to create new things, new companies, discover new science, whatever.

</aside>

<aside> 💡 (00:11:55) I think intelligence is a surprisingly fungible thing and we can get really good at all of these things in a single model.

</aside>

<aside> 💡 (00:16:39) It's so tempting to assume that the laws of physics for business have totally changed and they haven't yet.

</aside>

<aside> 💡 (00:31:05) The shift that I think the world needs to make for AI security generally and AI biosecurity in particular is to move from one of blocking to one of resilience.

</aside>

<aside> 💡 (00:36:04) You still need to learn to think, and writing, learning to write, or the practice of writing is very important to learn how to think.

</aside>

<aside> 💡 (00:39:29) The general worry I have is that the power and convenience of these are so high and the failure rates are maybe catastrophic, failures when they happen are maybe catastrophic, but the rates are so low that we are going to kind of slide into this like, you know what, YOLO, and hopefully it'll be OK.

</aside>

<aside> 💡 (00:52:56) I think this is going to be a deep trend in the coming decades is we care a lot about other people and we care very little about the machines.

</aside>

<aside> 💡 (00:58:13) Get good at generating ideas, be very resilient, be very adaptable to a rapidly changing world.

</aside>

Mindmap

• Part 1: The Future of Software and EngineeringThe Jevons Paradox and the Future of Software Engineering with AIAI's impact on software engineeringJevons paradox in software engineeringChanging role of software engineersIncreased software usage and customizationFuture GDP impactThe Bottleneck of Go-to-Market Strategies in the Age of AIGo-to-market (GTM) is a significant bottleneckBuilding a business and GTM has always been difficultAI is being used to automate sales and marketingHuman attention remains a limited resourceCreative ideas and great products are still essentialNavigating the Landscape of Multi-Agent Orchestration Tools and OpenAI's VisionOpenAI doesn't know the ideal interface for multi-agent systemsEncourages experimentation and diverse approachesBuilding tools for productivity with capable models is valuableOpenAI welcomes user feedback on desired features
• Part 2: Economic Impact and Model StrategyAI as a Deflationary Force and Equalizer in Addressing Economic GapsAI as a deflationary forceEmpowerment of individualsRisk of concentrated power/wealthSpecialized vs. Generalized Models: OpenAI's Focus on General Purpose AIGPT-5's writing ability is perceived as worse than GPT-4.5OpenAI prioritized intelligence, reasoning, and coding in GPT-5.2Future models will aim for excellence in all dimensions, including writingGeneral-purpose models are the future, even for specialized tasks like codingIntelligence is fungible, enabling a single model to excel in multiple areasCost vs. Speed: Balancing Priorities for AI Model DeliveryCost reduction for AI modelsImportance of speedBalancing cost and speedFocus on cost reduction
• Part 3: Product Evolution and Startup DurabilityThe Rise of Microapps and Customized Agentic InterfacesSoftware is evolving from static to dynamic and personalized.Expectation for computers to solve problems by writing code on demand.Trend towards microapps and customized software experiences.Tools will constantly evolve and converge to individual user needs.Customization is already happening internally at OpenAI with Codecs.Durability in the Age of AI: Building Advantages That LastStartup features can be replaced by model updates, but fundamental business laws remainBuild companies that benefit from improved AI models (GPT-6)Autonomous agents with long workflows still face challengesRealistic Timelines for Autonomous Agents and Long WorkflowsTask specificity is keyComplex tasks require breaking downGradual expansion of AI agent roles
• Part 4: Enhancing Human Intellect and ResearchImproving the Quality of Ideas with AI ToolsHuman attention is the rate-limiting factor on the consumption sideQuality of ideas is the rate-limiting factor on the production sideAI can help improve the quality of ideasTools should help users come up with good ideasAI as a brainstorming partnerScientific progress of AI models is no longer trivialEnsuring Models Adapt to New Technologies and UpdatesFear of technological stagnationDifficulty in adopting new technologiesQuestion of model adaptabilityThe Future of Scientific Research: Human Collaboration with AIAI models as general-purpose reasoning engines with extensive world knowledge.Near-future milestone: AI reliably using new tools/tech after single explanation/exploration.AI's role in research: speeding up idea pursuit, potential transition to AI-led research.Transition to AI-led research: may require new algorithms/world models.AI as an Unlimited Research Assistant: Automating Scientific DiscoveryAI for fully autonomous research is still distantCollaboration between AI and researchers is currently optimalAI's understanding of complex, multi-step processes will surpass human capabilitiesCurrent AI models lack creativity, intuition, and judgmentAI opens up new problems, requiring breadth-first search strategiesAutomating physical science: leveraging existing scientific community and data contributions
• Part 5: Safety, Security, and SocietyBiosecurity in the Age of AI: Moving from Blocking to ResilienceAI's dual role in biosecurityCurrent strategy: restricting access to AI modelsShift towards resilience is crucialAI as a solution to bioterrorism and cybersecurityAI helpful in dealing with novel pathogensBiosecurity as a potential area for significant AI-related failuresThe Role of Human Collaboration in the Age of AIAI's impact on learningHistorical parallels with the advent of GoogleAdapting education for the age of AIThe importance of writing and thinkingThe Value of Human Connection in an AI-Driven WorldAI and extreme autodidactsAI and collaborative learningFuture hardware and AIUnderestimated Failure Modes and Security Infrastructure for AI AgentsOverconfidence in AI agent safetyModel misalignment and complexityNeed for robust security infrastructure
• Part 6: Education, Career, and PersonalizationAI in Early Education: Balancing Technology and DevelopmentClaire's PerspectiveSam Altman's ViewThe Future of 3D Reasoning in Drug DesignGenAI benefits biopharma, especially in clinical trial document writingExploring GenAI for drug design, specifically compound developmentCurrent limitation: 3D reasoning capabilitiesFuture development: 3D reasoning will be addressedThe Value of University in the Age of AIParental pressure is expected after dropping outMake independent decisions, especially in a rapidly changing worldCurrent climate may not favor university for AI buildersAngel investing: Sam Altman is not currently angel investingSign In with ChatGPT: Balancing Convenience and SecurityFeature request: Sign in with ChatGPT accountOpenAI's plans: Implementing sign-in with ChatGPTConcerns about data sharing:Future possibilities: Sharing ChatGPT memory and project dataThe Changing Landscape of Software Development and HiringHiring slowdownChanging interview processFuture company structures
• Part 7: Identity and Essential SkillsHuman Creative Identity and AI-Assisted CreationConsumer appreciation is higher for human-made images versus AI-generated ones.People care about the creator's story and connection to the art.Entirely AI-generated art may not be as desirable as human-directed art.Human editing, curation, and personal story will continue to matter in art.The Future of Personalization and Memory in AIFocus on memory and personalization in AI developmentSam Altman's perspective on AI access to personal dataImportance of security and privacyUser laziness and AI understandingThe Most Important Skills to Learn in the Age of AIImportant soft skills for the AI ageThese skills are learnableOpenAI seeks user input for future AI models

Transcript

The Jevons Paradox and the Future of Software Engineering with AI

<aside> 📝 (00:00:04) Sam Altman addresses the Jevons paradox in software engineering, questioning whether AI's ability to create code faster and cheaper will reduce demand for software engineers. He believes the role of an engineer will significantly change, with more people creating and capturing value by using computers to fulfill their needs and others' needs. The demand for software is not slowing down, and Altman predicts a future where many people customize their own software, leading to a greater percentage of the world's GDP being created and consumed in this manner.

</aside>

Sam Altman: Okay. All right. Thank you all for coming. As we, as we start to think about the next generation of tools for builders, and how we're going to use these incredibly powerful models that are starting to come online. We wanted to just hear from you all about what you'd like, what's on your mind, answer any questions. But I hope we come out of today with a much clearer sense of what to build for you all and how to make these incredibly powerful models useful. I figured I would start with a question from Twitter. Where do you fall on the Jevons paradox of software engineering? If AI makes code dramatically faster and cheaper to create,

does that reduce demand for software engineers or does cheaper custom software massively increase demand and keep engineers employed for decades? I think what it means to be an engineer is going to super change. There will be probably far more people and creating far more value and capturing more value that are getting computers to do what they want. Getting computers to do what other people want. Figuring out ways to make these useful experiences for others. The shape of that job and the amount of time you spend typing code or debugging code or a bunch of other things is going to very much change. This has happened many other times in engineering, and each time it's happened, so far,

at least, more people have been able to join and be productive, and the world has gotten much more software. Demand for software seems to not be slowing down at all. And my guess on the future is a lot of us use software that was written for one person or a very small number of people. And we're constantly customizing our own software. So I think many more people will get computers to do the things they want to do, and it will be a very different way than we do it today. So if you count that as software engineering, then I think we'll see much more of it, and I think a greater percentage of the world's GDP will be created that way and consumed that way too. Any questions live?

Otherwise, I have a long list. Go ahead.

---

The Bottleneck of Go-to-Market Strategies in the Age of AI

<aside> 📝 (00:02:17) An attendee raises the issue of go-to-market (GTM) challenges, noting that while AI makes building products easier, finding users remains difficult. Altman acknowledges this has always been a challenge for startups. He suggests that while AI is transforming software engineering, the traditional rules of building a business and achieving GTM success still apply. AI is starting to automate sales and marketing, but human attention remains a limited commodity, making competition intense. Altman anticipates a future where human attention is the primary commodity, emphasizing the need for creative ideas and great products.

</aside>

Unknown Speaker: Well, first, I want to thank you for giving us this opportunity to be here and ask you these questions. From a consumer standpoint, I use ChatGPT heavily and I'm always on Reddit seeing everyone's building, whether it's using Codex, Lovable or Cursor. But it seems like the new problem is GTM, right? That I can build these things, but how do I find people to gain value from what I'm building? And I see it to be a bottleneck. I'm curious as to what you think about that.

Sam Altman: Before this, I used to run Y Combinator. And the consistent thing you'd hear from startup founders is, I thought the hard part of this was going to be building a product, and the hard part is getting anyone to care or to use it or to connect people. So I think this has always been extremely hard, but now it's gotten so much easier to build that you feel the delta even more. I don't have an easy answer for this. I think always it's been hard to build a business and to figure out a way to build differentiated value, to get the go-to-market function working. I think all of the old rules still apply there.

The fact that AI can kind of make it far easier to create software doesn't mean any of the rest of this gets easier. Except, I think you're starting to see in the same way that AI has transformed software engineering, people are now using it to automate sales, automate marketing, and there will be some success there. But I think it's always going to be difficult because even in a world of incredible abundance, human attention remains like this very limited thing. And so you're always going to be competing with other people trying to build their own go-to-market muscle and figure out how to get the distribution and every potential customer is busy and everything else. You know,

I could tell a version of the future where all of the radical abundance comes true and human attention really is like the remaining commodity. And so I just expect this to be hard and you got to come up with creative ideas and build great things.

---

Navigating the Landscape of Multi-Agent Orchestration Tools and OpenAI's Vision

<aside> 📝 (00:04:34) A solo developer asks about the future of OpenAI's agent builder tool and whether there's room for different UIs for multi-agent orchestration. Altman responds that OpenAI doesn't know the ideal interface and encourages experimentation. He notes that people have diverse preferences, from complex setups to calm voice-mode interactions. Altman believes building tools to enhance productivity with capable models is a good idea, as the potential of these models exceeds most users' ability to leverage them. He invites feedback on desired features.

</aside>

Unknown Speaker: I'm George. I'm a solo developer and I'm building like on top of the Codex SDK, I'm building a way to orchestrate multiple agents. And I have a question about your agent builder tool and sort of your product vision for where that agent builder tool goes. At the moment, it's just kind of workflows and chaining prompts together. But I'm wondering as a builder building on the Codex SDK, Am I safe? Like, do you see room for a lot of different types of UIs for multi or multi agent orchestration? Or do you see like OpenAI owns that stay away?

Sam Altman: No, I think we don't know what the right interface for all of this is going to be. We don't know how people are going to want to use it. We have seen people build incredible multi-agent setups. We have seen people build just a really great single interactive thread. We're not gonna figure this out on our own and also not everybody is gonna want the same thing. You know, there will be people that'll be like in one of the, they're in one of those like old movies with 30 computer screens and they're like looking at this crazy thing here and doing this and you know moving stuff around.

And I think there will be people who want to be in a very calm conversation in voice mode where they're saying something to their computer once per hour. The computer's figuring out a lot of things and they're not, you know, they're trying to think really hard about what they say, but they don't want like the constant supervision of tons of agents. Like many other things, People just have to try different approaches and see what they like and the world will probably converge on a few but we won't figure out all of them. I think building tools to help people get be productive with extremely capable models is a very good idea. That's totally missing right now.

The overhang of what these models are capable of relative to what most people can figure out how to get out of them is huge and growing. And someone is going to build a tool to really help you do that, and no one's gotten it right yet. We will also try our own versions of that, but this is a place where it seems like there's a lot of room and people are going to have different preferences. If any of you have things you'd like us to build here, let us know and we can try.

---

AI as a Deflationary Force and Equalizer in Addressing Economic Gaps

<aside> 📝 (00:07:02) An attendee asks how AI can address economic gaps, particularly the wage gap affecting women. Altman believes AI will be massively deflationary, leading to radical abundance and empowerment for individuals, regardless of their background. He anticipates that by the end of the year, a small investment in inference and a good idea could create software that would have previously taken teams a year to develop. This should equalize society and provide opportunities for those who have been unfairly treated, provided policies don't hinder this progress. Altman also expresses concern that AI could concentrate power and wealth, emphasizing the need for policy to prevent this.

</aside>

Unknown Speaker: Hey Sam, I'm Valerie Chapman and I'm building Rooth on OpenAI. I would love to know your thoughts on this because currently women actually leave about a million dollars on the table due to the wage gap and I'm curious how you think AI can be used to solve economic gaps that have existed for decades.

Sam Altman: I think the good news, well there's like a lot of complex news, but one of the things I think is mostly good news is that AI is going to be massively deflationary. I've gone back and forth on this a little bit because you can imagine some weird things happening with all of the money in the world going into self-replicating data centers or whatever. It looks like, on the whole, given certainly progress with work you can do in front of a computer but also what looks like it will soon happen with robotics and a bunch of other things, we're going to have massively deflationary pressure in the economy, and I said mostly good because there will be some complicated things to navigate through there,

but things getting radically cheaper other than the areas where the sort of social or governmental policy prevents that, like, you know, building more houses in San Francisco or something. I expect that to be pretty strong and pretty quick. The empowerment of individual people, whether or not society has structured in a way where they've naturally had all the advantages, looks like it's going to go up and up. I I still find it hard to wrap my head around that, you know, I'd say by the end of this year for a hundred or a thousand dollars of inference, you will be able to, and a good idea, you'll be able to create a piece of software that would have taken teams of people,

you know, a year to do. This is like very hard to wrap at least my head around this sort of magnitude of this economic change and That should be a very empowering thing for people. Massively more abundance and access and massively decreased costs to be able to create new things, new companies, discover new science, whatever. I think that should be an equalizing force in society and a way that people who have not gotten treated that fairly get a really good shot, as long as we don't screw up the policy around it in a big way, which could happen. I am worried that you can imagine worlds in which AI really concentrates power and wealth,

and that feels like it needs to be one of the main goals of policy for that not to happen.

---

Specialized vs. Generalized Models: OpenAI's Focus on General Purpose AI

<aside> 📝 (00:09:53) An attendee questions whether models are becoming too specialized, citing GPT-5's writing abilities as less impressive than GPT-4.5, despite its superior agent and coding capabilities. Altman admits they "screwed that up" and will improve writing in future versions. He states that OpenAI decided to focus on intelligence, reasoning, and coding in the current version due to limited bandwidth. Altman believes the future lies in very good general-purpose models, as even specialized models benefit from broader capabilities like good writing. He hopes to improve future models in all dimensions, viewing intelligence as a fungible trait.

</aside>

Unknown Speaker: Hey, my name is Ben Hylak. I'm the CTO of a company called Raindrop. I'm curious as you look into the future, how you think about And I'm here to talk about models being specialized versus generalized. So one example of this is like GPT 4.5 was the first model that I thought was really good at writing. I remember seeing the outputs. I was like, all right, this is really good writing. There's been a lot of discourse on Twitter and X recently about GPT 5's writing in ChatGPT and being a little unwieldy, hard to read. Obviously, GPT 5 is a much better agent model, really good tool use, intermediate reasoning, whatever.

It feels like models are a little bit spiky or they've gotten even spikier where some spikes like coding got super high, some spikes like, or it's very unspiky around writing. So I'm just kind of curious how OpenAI thinks about that feature.

Sam Altman: I think we just screwed that up. We will make future versions of GPT 5.x hopefully much better at writing than 4.5 was. We did decide, and I think for good reason, to put most of our effort in 5.2 into making it super good at intelligence, reasoning, coding, engineering, that kind of thing. And we have limited bandwidth here and sometimes we focus on one thing and neglect another. But I believe that the future is mostly going to be about very good general purpose models. You know, even if you're trying to make a model that's really great at coding, it'd be nice if it writes well too. Like if you're trying to have it be able to generate a full application for you, you'd like good writing in there.

When it's interacting with you, you'd like it to have a sort of thoughtful, incisive personality and communicate clearly. Like good writing in the sense of clear thought, not like beautiful prose. So my hope is that we just push to get future models really good in all of these dimensions and I think we will do that. I think intelligence is a surprisingly fungible thing and we can get really good at all of these things in a single model. It does seem like this is a particularly important time to push on kind of, let's call it coding intelligence, but we will try to excel and catch up on everything else quickly. I'll do a couple from Twitter after this. Go ahead.

---

Cost vs. Speed: Balancing Priorities for AI Model Delivery

<aside> 📝 (00:12:24) An attendee asks about cost reductions for developers, particularly for always-on AI applications. Altman estimates that GPT 5.2-level intelligence could be 100x cheaper by the end of 2027. However, he notes that speed is becoming increasingly important as model outputs become more complex. OpenAI is good at reducing costs but hasn't focused as much on delivering the same output in a fraction of the time, even at a higher price. He acknowledges the need to balance these priorities, as they present different challenges.

</aside>

Unknown Speaker: I'm CTO at this company called Unify. To your point earlier, we're doing go-to-market automation. One thing that we think a lot about and spend a lot of time on is kind of always-on AI, sort of like this ubiquitous AI. Something that you've said that really resonated with me is intelligence too cheap to meter. The limiting factor for us to run millions, tens of millions, hundreds of millions of agents for our customers is cost. How do you think about small models, costs, sort of dramatic cost reductions for developers over the next months and years?

Sam Altman: I think we should be able to deliver sort of GPT 5.2 X high level intelligence by the end of 2027 for, do you want to give a better guess? I can give them otherwise. Anyone want to give a guess? I would say at least a hundred X less. But there's another dimension which we haven't thought about as much historically and now as these model outputs get so complex, more people are pushing us on the speed we can deliver it at than the cost. And that is, we are really good at writing down the cost curve. You can look at the progress we've made even from like the first O1 preview until now. We have not Thought as much about how we deliver the output, the same output and maybe at a,

---

The Rise of Microapps and Customized Agentic Interfaces

<aside> 📝 (00:14:31) Altman discusses the trend towards microapps and customized agentic interfaces, noting that he now expects computers to write code to solve problems immediately. He believes the way we use computers and operating systems will change, with software being written specifically for individual users. While some interfaces will remain static, many tools will evolve and converge to meet individual needs. Internally at OpenAI, people are adopting codecs for their workflows and creating custom solutions, indicating a guaranteed shift towards personalized software.

</aside>

Sam Altman: at a much higher price, but in one 100th of the time. And I think for a lot of things you're talking about, people are going to really want that. Uh, we have to figure out how we're going to balance between prioritizing for those two things. Uh, they unfortunately are very different problems, but assuming we go push on costs and assuming that's kind of like what you all in the market wants, we can go very far down that curve. Yeah, let me do a couple of other ones. Current interfaces weren't built with agents in mind but were seen arise in apps I built for me. Why innovations in customized agentic interfaces could further accelerate the trend towards microapps? Yeah,

so this is one that I have noticed in my own use of codecs recently is I no longer think of software as this static thing. If I have a little problem, I expect the computer to like write some code right away and get it solved for me. And I think this trend is going to go much further. I suspect that the whole way we use computers and operating systems is going to change. I don't think it'll be like, oh, every time you need to edit a document, a new version of a word processor is going to be written for you right on the spot. Because, you know, we get, like, very used to our interfaces and it's very important that, like, that button is in the same place it was last time.

But For a lot of other things that we do, I think we will find that we expect software to be written just for us and when maybe I want to use the same word processor every time. But I do kind of have a bunch of repeated quirks of how I use it. And I would like the software to be increasingly customized, you know, a static or a slowly evolving piece of software, but written for me. And the way I use it is different than the way you use it. And that idea that our kind of tools are constantly evolving and converging just for us, that seems like it's going to happen. And certainly, internally at OpenAI, where people have like, very much adopted codecs for their workflows right now.

---

Durability in the Age of AI: Building Advantages That Last

<aside> 📝 (00:16:17) Altman addresses concerns about the durability of startups when their features can be replaced by model updates. He emphasizes that the fundamental laws of business haven't changed, and startups still need to solve the GTM problem and create something sticky with a competitive advantage. He advises building companies that benefit from model improvements rather than relying on temporary patches.

</aside>

Sam Altman: Everybody has their own little custom things and use things super differently. So that one seems like it's guaranteed. And I think that's like a very good direction to go build in and figuring out what that's going to look like and how people are going to do that seems seems great. How should builders think about durability when start-up features can be replaced by model updates? What is the one layer of the stack you promise not to eat? We talked about this a little bit earlier. It's so tempting to assume that the laws of physics for business have totally changed and they haven't yet. They may continue to change over time,

but right now what's changed is that you can do work faster and you can kind of create new software much, much faster. But all the other rules of building a successful startup, you know, you got to figure out a way to get users, you got to figure out a way to like solve the GTM problem, you got to figure out a way to provide something sticky, have some sort of mode network effect, competitive advantage, whatever you want to call it. None of that has changed. And the good news is like, It hasn't changed for us either. So there have been many startups that have done things that maybe in a perfect world we would have done sooner but it was too late and people built up,

you know, a real durable advantage and that will keep happening. The general framework I always give people when they ask some version of this question is, will your company be happy or sad if GPT-6 is like a wildly impressive update? And I encourage people because we continue to hope to make a lot of progress. I encourage people to try to build things where you are. So hoping the model gets wildly better. And there's so many things to build that way. And then the things where it's like a little sort of patch around the edge, that actually can still work if you build up enough of an advantage before the models get upgraded, but it is a harder and more stressful path.

---

Realistic Timelines for Autonomous Agents and Long Workflows

<aside> 📝 (00:18:23) An attendee asks about the timeline for agents that can run long workflows autonomously. An OpenAI speaker responds that it depends on the task. They note that some users are already running Codex in custom harnesses for extended periods. The speaker advises breaking down complex problems into smaller, verifiable tasks that agents can handle, gradually expanding their capabilities over time.

</aside>

Sam Altman: Let me do one more and we'll go back to the room. Realistic timeline for agents that can run long workflows autonomously without constant human intervention given that even simple on-chain tasks often break after five to ten steps. Anybody from OpenAI want to give an opinion on that? Go ahead.

Unknown Speaker: I think it really depends on the kind of task. So there's a number of tasks today where just inside OpenAI, we see people who are like prompting codecs in a very special way. Maybe they're using the SDK. So it's like a custom harness that keeps prompting it to continue, but they basically having it running, you know, forever. And so I think this isn't a question of when, but it's a question of like broadening of the horizon. So if you have a very specific task that you understand very well, try doing it today. You know, if you're starting to think like, okay, I want to get to the point where I can prompt the model to build a startup,

like that's a much more open ended problem with like a much harder verification loop. So I would encourage you to like figure out, okay, how can I break that down into a different problem where an agent can like verify itself? And then when I can verify its final output at the end of it, and then over time, we can let the agent do more and more and broader tasks.

---

Improving the Quality of Ideas with AI Tools

<aside> 📝 (00:19:31) An attendee raises the issue of human attention as a limiting factor on the consumption side and the quality of ideas as a limiting factor on the production side. Altman acknowledges the difficulty of generating good ideas and suggests building tools to help people brainstorm. He believes that as the cost of creation decreases, tighter feedback loops will help identify good ideas faster. Altman envisions AI as a brainstorming partner, suggesting ideas based on past work and code, similar to a "Paul Graham bot."

</aside>

Sam Altman: Other questions there? Thank you.

Unknown Speaker: Hi, Sam. So I want to go back on this like human attention and this question about GTM. I was thinking like the other side, like human attention is the rate limiting factor on the consumption side. On the production side for all the builders, it's the quality of ideas. And I just wanted to tee this up for you on, well, I spent all my time helping AI companies with their GTM and a lot of times the products actually just aren't worth So I guess how do people, what tools can you build to improve the quality of ideas that people are coming up with?

Sam Altman: There's a lot of It's popular to like call AI output slop, but there's a lot of like human generated slop in the world, too It's very hard to come up with good new ideas, and I am Increasingly a believer that we think at the limits of our tools I Think we should try to build a tools that help people come up with good ideas. I believe there are a lot and I believe as the cost of creation keeps to plummet, we'll be able to have such a tight feedback loop on trying them, we'll find the good ones faster. And as AI can discover new science in addition to writing very complex code bases, I feel confident there will be just a whole new possibility space.

But The experience of sitting down in front of an AI, you know, like a gentic code writer and just not being sure what to ask for next is something that a lot of people report. And if we can build tools to help You come up with good ideas, you know, and I believe we could do that. I believe we could like look at all your past work and all your past code and try to figure out what might be useful to you or interesting to you and can just continuously suggest things. If we can help provide a really great brainstorming partner. There have been like three or four people in my life I have a lot of ideas. They're people who are really good at asking questions or giving you seeds to build on.

And like Paul Graham is off the charts amazing at this. If we can build like a Paul Graham bot that you can have the same kind of interaction with to help generate new ideas, even if most of them are bad, even if, you know, you kind of say absolutely not to 95 out of 100 of them. I think something like that is going to be a very significant contribution to the amount of good stuff that gets built in the world, and the models feel like they ought to be capable of that. With 5.2, like a special version of 5.2 we use internally, we're now for the first time hearing from scientists that the scientific progress of these models is no longer super trivial,

and I just can't believe that a model that can come up with new Scientific insights is not also capable, you know, with a different harness and trained a little bit differently of coming up with new insights about products to build and stuff like that.

---

Ensuring Models Adapt to New Technologies and Updates

<aside> 📝 (00:23:02) An attendee expresses concern that as models improve, we might get stuck with current technologies, making it difficult to adopt new ones. Altman believes models will be able to use new technologies effectively, acting as general-purpose reasoning engines. He hopes that models will learn new skills faster than humans.

</aside>

Unknown Speaker: Howdy, I'm Theo, dev, YouTuber, YC founder, also really want the Paul Graham bot. I want to ask about something a little bit different though, more on the technical side. I really love when technologies like the building blocks that we use evolve. And I've been there for some crazy revolutions in the web world, like moving to TypeScript and Tailwind and all of these things. One of the fears I have as the models and the tools we use to build with get better is that we might get stuck with the way we have things working now. Like the power grid in the U.S. is built a certain way that makes things worse and we can't really change it. Do you see this potential here?

Are we making foundations out of the technologies as they exist right now that are going to be harder to swap in the future? Because even trying to get the current models to use the update to a technology that happened two years ago can feel like you're pulling teeth. Do you think we'll be able to steer the models enough to get them to use new things or are we just done improving the technologies we build on now?

---

The Future of Scientific Research: Human Collaboration with AI

<aside> 📝 (00:24:00) An attendee asks if models will eventually take over the entire research enterprise. Altman believes it's a long way off for most areas, though mathematics might be an exception. He notes that mathematicians are currently making the most progress by collaborating with models, using their intuition to guide the process.

</aside>

Sam Altman: I think we really will be very good at getting the models to use new things. Fundamentally, if we're using these models correctly, they're like a general purpose reasoning engine. The way we have things architected right now, they also have, you know, a huge amount of world knowledge built into them. But I think we are moving in the right direction. And I hope that updates and using new things and learning new skills even faster than humans do, is like a, you know, next couple of years thing. The A milestone that we will be very proud of is when the model can be presented with something totally new, new environment, new tools, new technology, whatever.

And you can explain it once or you don't need to explain the model can explore it once and then just super reliably use that and get it right. And that doesn't feel that doesn't feel very far away.

Unknown Speaker: I have a question that I think you touched on already, but as a scientist and kind of on the older end, when you do a scientific project, it tends to generate multiple ideas for further research. So it's exponentially growing while a given scientist only has a linearly decreasing amount of time to carry that out. So it's incredible how these tools are speeding it up. But we're all greedy for more. And you touched on this earlier, but do you think that the possibility that, you know, in addition to helping us pursue all those interesting ideas in shorter time, that there will be a transition where the models, you know, will take over the whole research enterprise? And if so,

do you see that as coming from the existing algorithms or requiring new ideas or world models or that sort of thing?

---

AI as an Unlimited Research Assistant: Automating Scientific Discovery

<aside> 📝 (00:26:06) Altman compares the current state of AI in research to the period in chess history when humans plus AI were better than AI alone. He suspects a similar transition will occur in research, with AI eventually surpassing human capabilities in understanding complex, multi-step processes. However, creativity, intuition, and judgment remain areas where current models fall short. He highlights the new skill of using AI to explore multiple research problems simultaneously, acting as an "unlimited grad student" or "unlimited postdoc."

</aside>

Sam Altman: I think it's a still a long or reasonably long way away from the models doing truly completely closed loop autonomous research in most areas. We could look something like mathematics. And say, okay, that one doesn't need a wet lab or physical input at all. You know, maybe you can just think really hard and make a ton of progress if you kind of continually updating the model. Eventually, even there for now, the mathematicians who are making the most progress with the models are very heavily involved in looking at intermediate progress and saying, nah, this just doesn't feel right. You know, I have an intuition that there's something different on this other path.

But I've gotten to meet a few mathematicians who now say their entire day is collaborating with the latest models. They're making rapid progress, but they do something very different than the model. To be honest, it feels to me like that period in chess history when Deep Blue beat Kasparov, then there was this period of time where, okay, AI is better than humans, but a human plus an AI, where the human is choosing the best of ten moves from the AI is better than that, and then very quickly after that, the AI was again better, and the human was just making it worse. I sort of suspect for something like many kinds of research,

something like that should happen over time and things are going to get so complex that the AI can understand sort of multi-step things better than most people can or all people can. But there seems to be something about creativity, intuition, Judgment that we are not close to with the current generation of models. I can't come up with any principled reason why we won't get there, so I assume we will. But today, I don't think just sort of saying like, hey, GPT 5 point, GPT 6, go solve math. That is certainly not going to outperform a few very good people doing math with it and saying, OK, this is a good direction. This is, you know, even though we can verify, even though we can say,

hey, you did a great proof, put that back into the training set. There's something else happening. In terms of the workflows that are working, though, you touched on something, which is you solve one problem and it opens up many new problems. That's where it's been very cool to talk to the scientists that are really using this aggressively. I mean, they burn a lot of GPUs in the process, but I think there is a new skill of being able to say, here's the 20 new problems, and I'm going to, like, I'm going to do a breadth-first search on them. I'm not going to go deep on anyone, and I'm going to use the AI to be an unlimited grad student, is how someone described it.

I actually recently upgraded them to unlimited postdocs. In terms of automating physical science, we go back and forth a lot about whether we should be building Automated wet labs for every field, which we're open to doing. Or if the world as a whole will figure out great experiments and has a lot of equipment and will happily contribute data back in. It sort of seems like just watching the scientific community embrace 5.2 and how much they've been willing to help, that that might work. And that would clearly be a easier, better, more distributed, more smart people, more different kinds of equipment world.

---

Biosecurity in the Age of AI: Moving from Blocking to Resilience

<aside> 📝 (00:29:50) An attendee raises concerns about the harms of AI-enabled biological redesign and asks about the role of security in OpenAI's roadmap. Altman acknowledges the risks of AI in bio and notes that the current strategy of restricting access and using classifiers is not sustainable. He advocates for a shift from blocking to resilience, drawing an analogy to fire safety.

</aside>

Unknown Speaker: Hi, Sam. My name is Emmy. I'm a Stanford student and I run a biosecurity startup. So to your conversation about Scientific experiments and really just like, for example, like cloud labs and where all this is going. My team spends a lot of time thinking about how we can prevent the harms of AI-enabled biological redesign, but also how can we use AI to uplift security infrastructure. So I guess my question is, where does security fall in this 126 roadmap? And broadly, how do you think about some of these issues?

Sam Altman: Security broadly or biosecurity specifically? Either. Preferably biosecurity. There are many ways AI can go wrong in 2026. Certainly one of them. We are quite nervous about is bio. The models are quite good at bio and right now most of our and by like our not just OpenAI, the world strategy is to try to restrict who gets access to them and you know put a bunch of classifiers to not help people make novel pathogens. I don't think that's going to work for much longer and The shift that I think the world needs to make for AI security generally and AI biosecurity in particular is to move from one of like blocking to one of resilience.

My co-founder Vojtech uses the analogy I really like about fire safety. Fire did all these wonderful things for society. Then it started burning down cities. We tried to do all of these things to restrict fire. I just actually learned this weekend that curfew comes from like when you weren't allowed to have fires anymore because they were burning down cities. And then we got better at like resilience to fire and we came up with fire code and Flame resistant materials and a bunch of other things. And now we're pretty good at that as a society. I think we need to think about AI the same way. AI is going to be a real problem for bioterrorism. AI is going to be a real problem for cybersecurity.

AI is also a solution to those things. It's a solution to a lot of other problems as well. And I think we need like a society-wide effort not to sort of provide the infrastructure for this resilience, not labs that we trust to sort of always block what they're supposed to block. You know, there will be many good models in the world. We've been talking to a lot of bio researchers companies about what it takes to be able to deal with novel pathogens. I think there are a lot of people interested in the problem and a lot of people reporting that AI actually seems helpful at this, but it won't be a technological, it won't be an entirely technological solution.

You will need the world to think about these things differently than we have been. So I am I'm very nervous about where things are, but I don't see a path other than the sort of resilience-based approach. And it does seem like AI can really help us do that fast. But if something goes really wrong, like visibly really wrong for AI this year, I think bio would be a reasonable bet for what that could be. And then as we get into next year and the following year, you can imagine lots of other things going really wrong too.

---

The Role of Human Collaboration in the Age of AI

<aside> 📝 (00:33:16) An attendee questions the role of human collaboration when AI can provide answers instantly. Altman compares this to the introduction of Google, where teachers feared students would no longer need to learn. He believes AI tools are a problem for the current education system, suggesting that teaching methods need to change.

</aside>

Unknown Speaker: Hi, my name is Meghna. My question is kind of related to human collaboration in the sense that when we're talking about AI models improving and stuff like that, they become really good, I think, for learning topics or learning subjects really quickly on your own. And that's something that we've explored in the ChatGPT and education lab and which I really valued and I appreciate a lot. But one thing that I What I kind of often find myself coming back to is the role of other humans and human collaboration, where it's just like, if you can get an answer at your fingertips, then why would you necessarily take the time and maybe even the friction to ask another human about it?

And so that's something I've been thinking about a lot. Also related to, I think, something that you mentioned before in this session, where you're just like, All these AI coding tools that can perform the work of human teams at a much faster pace. So when I'm thinking about like cooperation and collaboration and kind of collective intelligence output, I know that human plus AI is a very powerful kind of avenue for that. But what about maybe humans plus AI? Yeah, if that makes sense.

Sam Altman: Totally. Many things in there. I'm much older than most of you, but I was still like, I was in middle school when Google came out. And the teachers tried to make the kids promise not to use it. Because there was this feeling that if you could look up anything at your fingertips, then why come to history class? Why memorize anything, you know? And It seemed to me totally insane. And I was like, actually, I'm going to be way smarter, and I'm going to learn way more, I'm gonna be able to do way more. And this is the tool that I'm going to like live with as an adult. And if I don't learn, it would be crazy to make me learn something that assume it felt like it would be like,

you know, Many decades earlier, if you like made me learn to use an abacus, even though you knew there were calculators, because that was just like a important thing to learn or slide. I don't even know what came before the calculator at this point, but it was not a valuable skill to learn. And I feel the same way about AI tools. I understand that in the current way we teach kids, AI tools are a problem. But I think that suggests that we need to change the way we teach people. Not that, like, we don't want, you know, the fact that you can, like, have ChatGPT write something for you. That's the way the world's going to be. You still need to learn to think, and writing, learning to write,

---

The Value of Human Connection in an AI-Driven World

<aside> 📝 (00:36:21) Altman believes human connection will become more valuable in a world of AI. He notes that OpenAI is exploring interfaces to facilitate collaboration and multiplayer experiences with AI. He envisions a future where AI is a standard part of group brainstorming and problem-solving sessions.

</aside>

Sam Altman: or the practice of writing is very important to learn how to think. But probably the way we should teach you to think and the way we should evaluate your thinking ability has changed, and we shouldn't pretend otherwise. So and I feel totally. Like, this is going to be fine. We will, you know, the 10% of learners that are like extreme autodidacts are already doing amazing. We will figure out new ways to teach the curriculum and bring the other students along. And then there's another thing you said, which is how do we make this be a collaborative thing and not just your learning and performing and doing amazing things, but you're just with your computer by yourself.

We haven't seen evidence of this yet, and it is something that we try to measure. I suspect that human connection is going to be more valuable in a world of lots of AI, not less, and that people are going to value getting together with other people and working with other people more. But we have started to see people explore interfaces to make that easier. As we think about making our own hardware, our own devices, we have thought a lot about, I mean, we've even thought first about what a collaborative sort of multiplayer plus an AI experience looks like. And my sense is that although no one has cracked it quite yet,

---

Underestimated Failure Modes and Security Infrastructure for AI Agents

<aside> 📝 (00:38:08) An attendee asks about the most underestimated failure modes of AI agents operating in production systems. Altman expresses concern that the power and convenience of AI will lead people to overlook potential catastrophic failures. He worries that the pressure to adopt these tools will cause people to neglect the complexity of running them securely.

</aside>

Sam Altman: we will be surprised at how much This is enabled by AI in a way that no other technology has enabled. So you can have, you know, five people sitting around the table and a little, you know, kind of robot or something also there. And you will be able to be way more productive as a group. And you'll just be used to this all the time. Like every, every group brainstorm, every time you try to solve a problem, there'll just be a AI as part of it, and it'll help the group do better. Super happy to talk about anything. As a reminder, any requests, like if you tell us something, we'll probably just build it.

Unknown Speaker: Oops, thanks. I'm curious about, like, as agents start moving and operating more production systems, especially at scale, like, where do you think or what would you think is the most underestimated failure mode, like security, cost, reliability? And related to that, where do you think the really hard work is being under-invested in right now?

Sam Altman: Tons of problems everywhere you mentioned. One of the things that surprised me personally, and I think it's in two ways and has surprised many of us here, is when I first started using codecs, I said, look, I don't know how this is going to go, but for sure I'm not going to give this thing like complete unsupervised access to my computer. I was so confident in that. And I lasted about like two hours. And then I was like, you know what, it seems very reasonable. The agent seems to really do reasonable things. I hate having to approve these commands every time. I'm just going to turn it on for like a little bit and see what happens.

And I never turned it, I never turned like, you know, full access off. And I think other people have had a similar thing. So The general worry I have is that the power and convenience of these are so high and the failure rates are maybe catastrophic. Failures when they happen are maybe catastrophic, but the rates are so low that we are going to kind of slide into this like, you know what, YOLO, and hopefully it'll be OK. And then as the models get more capable, And harder to understand everything they're doing.

If there's a misalignment in the model if there's some sort of complex problem that emerges over weeks or months of usage and you kind of put some security vulnerability into something you're making and you know you can have various opinions on this curve of like how crazy sci-fi you want to get with the AI being misaligned or whatever, but I think what's going to happen is the pressure to adopt these tools to use them, not just the pressure, but like the light and the power of them is going to be so great that people aren't that people get pulled along into sort of not thinking enough about the complexity of how they're running these things,

how they're sure about, you know, this, whatever sandbox they've set up. And the general worry I have is that capability is going to rise very steeply. We're going to get used to how the models work at a certain level and decide we trust them. And without building very good I'll call it big picture security infrastructure around it. We will sleepwalk into something. I think that would be a great kind of company to build.

---

AI in Early Education: Balancing Technology and Development

<aside> 📝 (00:41:11) An attendee asks about the role of AI in kindergarten and middle school. Altman is generally against computers in kindergarten, advocating for outdoor play and social interaction. He believes the impact of technology on young children is not fully understood and should be approached cautiously.

</aside>

Unknown Speaker: Hi, I just wanted to go back to the conversation about education. My name is Claire. I'm a sophomore at Berkeley studying cognitive science and design. And I was in high school when I saw my peers using ChatGPT to generate essays and homework and stuff like that. And now I'm in college and we're having these discussions about AI policy and coursework and CS and humanities and everywhere. And I wanted to go back to this idea of kindergarten and middle school and what it might look like for AI to be in classrooms during those periods of really, you know, it's the time when you figure out how to problem solve and how to write and how to think. And, you know, as someone who is a father now,

how do you foresee education changing and being shaped by AI during those really formative years?

Sam Altman: I mean, generally speaking, I'm a fan of keeping computers out of kindergarten. And I think kindergarten should be like running around outside and playing with physical things and trying to learn how to interact with each other. So not only would I not have AI in most kindergartens, most of the time, I wouldn't put computers either. I think developmentally we still don't understand all of the impacts of technology on, you know, there's been like a lot written about the impact of social media on teenagers and that seems like it's been pretty bad,

but I have a sense that unfortunately a bunch of technology on young children has been even much worse and it's still talked about relatively little and I think until we understand that better, like probably we don't need kindergartners using a ton of AI.

---

The Future of 3D Reasoning in Drug Design

<aside> 📝 (00:42:54) An attendee asks about the progress of 3D reasoning for drug design. Altman says that OpenAI will solve this problem, although he doesn't know if it will be a 2026 thing. He notes that it is a common request.

</aside>

Unknown Speaker: Hi, my name is Alan and I'm in biopharma. So GenAI has been really helpful for clinical trial document writing, accelerated a lot of things, been amazing. We're also trying to use it for drug design, particularly for compounds. And one of the things we run into is 3D reasoning. I was wondering if there's going to be an inflection point or if there's something you see down the line on that.

Sam Altman: We're going to get that solved. I don't know if that's a 2026 thing, but that is a super common request. And I think we know how to do it. We just have a lot of other urgent areas to push on, but we will get there. Thank you.

---

The Value of University in the Age of AI

<aside> 📝 (00:43:43) A recent university dropout asks if university is limiting and whether Altman angel invests. Altman, who also dropped out of university, says that parents will always pressure their children to finish their education. He says that it is probably not the best use of time to be in university if you are an AI builder.

</aside>

Unknown Speaker: Hi, Sam. I'm Dan. I just dropped out of university in London to join the W26 Y Combinator badge. And I got two quick questions. First one is, my parents are still kind of pressuring me to finish university and do you think university could be, in its current state, limiting sometimes? And second is, do you angel invest?

Sam Altman: I dropped out of university and it took my parents It's been 10 years to stop asking when I was going to go back. So I think like parents are just going to do that and they love you and they're trying to give you advice they think is best and you just sort of keep explaining to them that you can always go back if you want but the world is in a different place now and going to keep being a different place. Everybody's got to make their own decision, but I think you do need to make your own decision and not just do what society tells you to do. And I don't personally think this is a time where from If you are an AI builder,

---

Sign In with ChatGPT: Balancing Convenience and Security

<aside> 📝 (00:45:30) An attendee requests the ability to sign in to other services with their ChatGPT account. Altman says that OpenAI is going to do that. He says that it is very scary because ChatGPT knows so much about you.

</aside>

Sam Altman: it is probably not the best use of your time to be in university right now. If you're just like a sort of ambitious high agency driven person, this is an unusual period of time. And, you know, you can always go back later and I think just tell your parents that you're not, like, it doesn't mean that it's not the right thing for many people. It doesn't mean that it won't be the right thing for you sometime. But like right now you got to do this thing and I think they'll understand eventually. And then on the second thing, I respect the hustle, but not anymore. I miss it.

I just got really busy with OpenAI and it kind of gets strange because if I end up investing in companies that are like big OpenAI customers, I've decided it's easier not to. Hey, Sam.

Unknown Speaker: I'm Michael from WorkOS. We do a lot of stuff with authentication and identity and signing in. So I have a feature request for you, which is sign in with my ChatGPT account. I think a lot of people would like that.

Sam Altman: We are going to do that. People ask me for it all the time. What do you need? Do you want people to like be able to bring their own token budget or do you want them to like bring their ChatGPT memories or all?

Unknown Speaker: Yeah, so that's my question. I mean, definitely token budget. People be able to bring their accounts and You know, what models they have access to. But I think there's all this other stuff too, like what, you know, MCP servers does my company have access to or what memory does ChatGPT have about me or what projects am I working on? I'm curious how you're thinking about that. Like ChatGPT knows so much about me from a work perspective, but also a very personal perspective.

Sam Altman: Yeah.

Unknown Speaker: And how developers can leverage that.

Sam Altman: So we do want to figure out how to do this. It's very scary because ChatGPT does know so much about you. And if you tell a person that you're very close to a bunch of secrets, you can be relatively confident they'll know the exact social nuances and when they share what with who and when something overrules something else. Our models are not quite there, although they're getting pretty good at it. I would I think feel uncomfortable if I connected my ChatGPT account to a bunch of sites and said, just use your judgment about like when to share what you know about me from all of my chat history and everything I've connected. But when we can get there, it will clearly be a cool thing to offer.

And in the meantime, I think doing something just with, you know, token budgets, and if I pay for the pro model, then I can use it on other services, that seems like a cool thing to do. So I think we will at least do that. And we'll try to figure out a way to get the information sharing right. But like, we really don't want to, we really don't want to screw that up.

---

The Changing Landscape of Software Development and Hiring

<aside> 📝 (00:47:56) An attendee asks how software development interviews have changed. Altman says that OpenAI is planning to dramatically slow down how quickly they grow because they think they will be able to do so much more with fewer people. He says that the right approach will be to hire more slowly, but keep hiring.

</aside>

Unknown Speaker: Hey, Sam. My name is Oleg. And I guess we all agree here that software development as a craft has changed dramatically recently. But at the same time, LinkedIn still has opening a job of offerings for software developer. And I'm curious, how did the interview have changed in past months or years?

Sam Altman: We're going to keep hiring software developers, but we are for the first time, and I know every other company, every other startup is thinking about this too. We are planning to dramatically slow down how quickly we grow, because we think we'll be able to do so much more with fewer people. And I think at this point, a lot of the impediments that we face or that other companies face is just like the internal policies that have built up at most companies did not contemplate a majority AI co-workers. And that's going to take a while. But what I think we shouldn't do, and what I hope other companies won't do either, is hire super aggressively, then realize all of a sudden,

AI can do a lot of stuff and you need fewer people and have to have some sort of very uncomfortable conversation. So I think the right approach for us will be to hire more slowly, but keep hiring and trust that I'm not a believer that eventually, well, maybe someday far in the future OpenAI has zero employees, but for a long time, I think we'll just have a lot. We'll have a gradually increasing number of people doing much more stuff. And this is kind of what I expect the shape of the economy to look like more generally. In terms of what the interview looks like, it has not yet changed as much as it should, but I was in a meeting today with people talking about how we want it to change.

We basically would like to sit you down with something that would have been impossible for one person to do in two weeks. You know, this time last year, and watch them do it in 10 minutes or 20 minutes or whatever. There are Yeah, I think that's like the high order bit is you just you want to interview you want to see that people are going to be able to work in this new way very effectively I think software engineering interviews have been bad for a long time and maybe not that relevant, but now they're even less relevant So that's one thing there's like a more general thing that a few of these questions have hinted at Which is is the future going to be?

You know, companies don't hire many people and have a lot of AI co-workers or, you know, some version of that. Or is it going to be that companies that win in the future are entirely AI? You know, like it's a rack full of GPUs and no people. I really hope it's the former. There are a bunch of reasons why it seems like it could be something closer to the latter. But if companies don't adopt AI aggressively, if companies don't figure out how to hire people that are going to use the tools really effectively, they will eventually just be outcompeted by a fully AI company that doesn't have to have the sort of silly policies that prevent big companies from using AI or whatever.

And that feels like it would be a very destabilizing thing for society. We've been trying to figure out how to talk about this, but I think it's very, because it sounds self-serving for us to say, but I think it's very important that companies adopt AI in a big way very quickly.

---

Human Creative Identity and AI-Assisted Creation

<aside> 📝 (00:51:41) An attendee asks where Altman sees the relationship between human creative identity and AI-assisted creation going. Altman says that consumers of images report dramatically higher appreciation if they are told a person made it versus an AI. He says that the person and their life story and their editing or curation is going to matter a lot.

</aside>

Unknown Speaker: Hi, Sam. I'm Cole. I'm a creator and cinematographer. I think we've seen, especially in the past year, AI has completely changed the way we tell stories, and as a result, the way we view ourselves. There have been so many interesting plays in like the creative space like for Sora it was such an interesting use of like the self as a canvas and allowing you to be able to use AI and put yourself in all of these fantastical scenarios. I'm really curious where you see kind of this relationship between human creative identity and also AI assisted creation is going especially as these models continue to advance.

Sam Altman: The place that we can study and I think learn the most right now is ImageGen. It's been around the longest. The creative community has used it and disliked it and liked it the most. There's many interesting observations there, but one of them is that consumers of images report dramatically higher appreciation, satisfaction, whatever, if they are told a person made it versus an AI. And I think this is going to be a deep trend in the coming decades is we care a lot about other people and we care very little about the machines. Of all of the slurs for AI, Clanker is my favorite one. I think it just it like so evokes people's emotional reaction and You can see these incredible,

beautiful, creative, to me at least, you know, clinker made images. And as soon as you're told that, many people's subjective appreciation goes way down. There was a thing I saw on the internet last year of they would go to people who said they really hated AI generated art, like still images. And The people would all say, and I can tell for sure what the AI generated images are, because they're terrible. And they'd show them 10 images and say, rank your favorite ones. Half would be done entirely by human, half entirely by AI. And like fairly consistently, they would rank the AI ones at the top. And then as soon as they were told that they would say, actually, I don't like it.

And you know, this is not the one I want. And that is kind of the test is what you like. When I finish reading a book that I love, the first thing I want to do is like look up the author and understand their life and, you know, kind of how it led them to do that, because I felt this connection to this person that I don't know, and I want to understand them. And I think if I read a great novel, and at the end, I learned it was written by an AI, I would sort of be kind of sad and crestfallen. And I think this is going to be a deep and durable trend. However, If the art is even a little bit human directed, and how little, maybe we'll have to figure out how people feel over time,

people don't seem to have that same strong emotional reaction. And this is versions of this been going on for a long time, you know, if digital artists used Photoshop, people still love their art. So my expectation, given the behavior that we're seeing now from creators and consumers, is that The person and their life story and their editing or curation or whatever goes into that process is going to matter a lot. And we're not going to want the entirely AI generated art, broadly speaking, at least from what we can learn from images.

---

The Future of Personalization and Memory in AI

<aside> 📝 (00:55:28) An attendee asks how Altman sees personalization and memories evolving over time. Altman says that OpenAI is going to push super hard on memory and personalization. He says that he is ready for ChatGPT to just look at his whole computer and his whole internet and just know everything.

</aside>

Unknown Speaker: We have time for two more questions. Hey, Sam, my name is Keith Curry, recent graduate of San Francisco State. And my question revolves a bit around like personalization and memories. So first part is kind of how do you see that evolving over time? And then also, what are your thoughts on like more granularity, like, for example, grouping memory. So like, this is my work identity, this is my personal identity. So that way, as you're doing different prompts, you can sort of be more selective about what you want to be included on it.

Sam Altman: Yeah, so we're gonna push super hard on memory and personalization. Clearly people want it and it delivers a way better way to use these tools. I I have gone through my own evolution here, but at this point I am ready for ChatGPT to just look at my whole computer and my whole internet and just know everything. The value from it is so high and I don't feel uncomfortable about it in the way that I used to. I really hope, you know, all AI companies take security and privacy super seriously and I hope that society as a whole does too because The utility is so great, like AI is going to know about my whole life. I'm not going to get in the way of that.

I don't yet feel ready to like wear the glasses recording everything. I think that's still uncomfortable for a bunch of reasons. But I do feel ready to say like, hey, you can just have access to my computer and figure out what's going on and be useful to me and understand everything and like have a perfect representation of my digital life. I am lazy. I think most users are lazy too, though. So it's like a reasonable representation. And I don't want to sit there and have to group like, this is a work memory, this is a personal memory, this is something that what I want, and what I believe is possible, we talked about this a little bit earlier,

is for AI to have a such a deep understanding of the complex.

---

The Most Important Skills to Learn in the Age of AI

<aside> 📝 (00:57:39) An attendee asks what Altman thinks is the most important skill that people should learn in the age of AI. Altman says that skills like become high agency, get good at generating ideas, be very resilient, be very adaptable to a rapidly changing world are going to matter more than any specific skill.

</aside>

Unknown Speaker: Hi, Sam. My name is Luan. I'm an international student from Vietnam. And my question is, what do you think is the most important skill that people should learn in the age of AI?

Sam Altman: These are all kind of like soft skills. None of them are like as go like learn to program was so obviously the right thing, you know, over recent period of time, and now it's not, but skills like become high agency. Get good at generating ideas, be very resilient, be very adaptable to a rapidly changing world. I think these are going to matter more than any specific and I think these are all learnable. This is one of the surprises to me of having been a startup investor is the degree that which you can like take people and in a three-month sort of boot camp style thing make them extremely formidable. And do the things on all those axes I was just talking about is very surprising.

It was a big update to me. And, and so I think these are the skills that may matter the most. And they're they're like quite learnable. Are we out of time? Okay. Um, thank you all very much for coming and talking. We really do want input on what you'd like us to build, like assume we will have a model that is 100 times more capable than the current model with 100 times the context length that 100x the speed of 100x reduced cost, perfect tool calling extreme like coherence over like, we're gonna get there. Tell us what you'd like us to build. We're going to hang around and if you're like, hey,

I just need this API or I just need this kind of primitive or I just need this sort of runtime or whatever it is, we're building it for you and we'd like to get it right. But thank you all for coming.

首发于微信公众号「田野与浪潮」 · 阅读原文 ↗