
What this covers
The conversation with Greg Brockman is a first-person history of OpenAI, followed by a look ahead. It starts in 2015, when he and Sam Altman wondered whether a new lab could compete with DeepMind, which already had the researchers, capital, and data. Brockman describes the Napa offsite that produced a three-part plan: solve reinforcement learning, solve unsupervised learning, then learn progressively harder things. He treats the 2017 sentiment neuron result and the Dota 2 project as turning points. The first showed meaning emerging from character-level prediction, and the second showed that large compute and simple algorithms could handle open-ended, interactive environments. He also argues that predicting what Einstein would say next amounts to intelligence, a claim flagged as contested.
The middle of the episode covers the strain inside the organization and the firing of Altman. Brockman recounts the hours that followed, his decision to quit, the employee petition, and how his relationship with Ilya Sutskever changed. He argues that ordinary disputes over governance and credit carried existential weight because of what the mission was.
The final stretch looks outward. Brockman discusses alignment through a sycophancy problem that OpenAI corrected, and defends iterative deployment as a way to learn from real misuse. He covers distillation by competitors, US leadership and democratic values, and who gets served when compute is scarce. He also touches on data centers, possibly in space, and regulation that protects privileged relationships with AI. The most contested forecast is that everyone will manage AI agents, possibly as the head of an "autonomous AI corporation". The episode closes on jobs and what young people should study.
Greg Brockman argues that OpenAI's success stems from encountering reality as it is, maintaining focus on the mission of beneficial AGI, and deploying iteratively while building compute infrastructure at scale to eventually distribute AI benefits broadly to humanity.
- OpenAI's technical strategy (reinforcement learning + unsupervised learning + scaling compute) succeeded because they grounded decisions in physical reality rather than blind faith
- Iterative deployment allows learning from real-world misuse patterns rather than hypothetical risks, while building resilience into society
- Massive compute infrastructure investment, initially criticized, is now essential to delivering on the mission of universal AI access
This asset isn't compiled yet
You're seeing its claims, ranked. Compile it to build the argument threads, weight them, and check each claim against your library — the full view.
The unsupervised sentiment neuron paper (2017) was a watershed moment because it demonstrated, for the first time, that semantics could emerge from training on a character-level language modeling objective—proving that neural networks could learn meaning, not just syntax.
“an early moment was the unsupervised sentiment neuron paper... Yes, that one's an interesting one cuz it's 2017, and it's really the first time that we saw semantics arise from training on a language modeling objective. So, you train on learn the next character, predict the next character, and then suddenly you get a neural net that understands sentiment, understands if something is positive or negative.”
Prediction and reasoning are deeply connected: if a model can predict what Einstein would say next, it is at least as intelligent as Einstein; the task of prediction is not about replicating known information but about predicting novel situations never encountered before, which requires genuine intelligence.
“just predicting what comes next sounds like a pedestrian task. But if you really can predict the next word out of Einstein's mouth, you are at least as smart as Einstein... the point of prediction is not about being able to predict what is known. The point is you put yourself in a new situation you've never seen before and predict what comes next.”
DOTA was a watershed moment not because it was chess-like (Deep Blue) or computationally intensive with defined rules (AlphaGo), but because it involved human-like interactive agency in a partially-defined, continuous environment—demonstrating that massive compute + simple algorithms could exceed human-level performance in real-world complexity.
“I thought Dota was like an incredible moment because it was it wasn't um chess like Deep Blue and it wasn't AlphaGo, which is like computationally intensive but very defined rules. It was actually interactive against humans in a way that like the world is sort of structured, but you have all of this freedom.”
The US maintains leadership in frontier AI through breakthrough algorithms concentrated in US and Western companies, though innovation is happening globally; the balance of AI dynamics between countries is still being determined and depends on chip exports, technology access, and geopolitical decisions.
“we have this concentration of where the breakthrough algorithms come from in the US, in western companies. There's clearly a lot of innovation happening around the world, but I think exactly the balance of dynamics and how like which countries rely on which providers, all of that is something that I think is still being determined.”
US leadership in AI is critical for America because it ensures democratic values are protected and preserved—and every country is realizing they need a sovereign AI strategy for economic and national security.
“I do think that leading in AI is very critical for America. Because I think that this is how you can ensure that democratic values are protected and preserved, and I think that every country is also starting to realize that they need some sort of sovereign AI strategy.”
OpenAI's position is that they are closest to the technology and best positioned to advise on deployment because they understand its properties deeply—while they incorporate wisdom from others who have deployed transformative technologies, the specific choices are too dependent on this technology's unique characteristics for external advisors to determine.
“we've had some hope that, 'Hey, there are people who have deployed transformative technologies before. Maybe they can tell us the answers.' And it's never been so simple. They do have wisdom and insights... but we realized that we're the closest ones to this technology. That by virtue of creating it, we have an understanding of the ways in which we could shape it. That is hard for someone who isn't so close to it to opine on, to advise on.”
The future will be one where everyone is a manager of AI agents, and potentially a CEO of an 'autonomous AI corporation'—individuals with access to compute equivalent to a 100,000-person company operating on their behalf 24/7.
“we're all going to be heading to a world where we're managers of agents, and soon maybe the CEO of an autonomous AI corporation. Or just imagine if you had the workforce of you know, a 100,000 person company all at your disposal, operating on your behalf. 24/7.”
America losing the AI race would have consequences for democratic values—maintaining US leadership in AI is presented as necessary for preserving and protecting democratic values against alternatives.
“leading in AI is very critical for America. Because I think that this is how you can ensure that democratic values are protected and preserved”
OpenAI's initial technical plan, developed at an offsite in Napa in 2015, was threefold: (1) solve reinforcement learning, (2) solve unsupervised learning, and (3) gradually learn more complicated things—a plan the organization has pursued for the past 10 years.
“we came up with what I would really say is almost the technical plan that we've pursued for the past 10 years. Number one, solve reinforcement learning. Number two, solve unsupervised learning. And number three was gradually learn more complicated, in quotes, things.”
Ilya's principle that 'you have to suffer—if you're not suffering, you're not building value' reflects a deep truth about OpenAI's philosophy: you cannot avoid hard realities through optimism or blindness; instead, you must encounter reality as it is, which is painful but necessary for achieving ambitious goals.
“In the words of Ilya, Ilya always says that you have to suffer, right? If you're not suffering, like you're not building value. And I think there's deep truth to it... We had so much uncertainty from the beginning... And it's easy to just sweep the problems under the rug and just blindly say go. And I think that is the negative side of like Silicon Valley culture... But I don't think that works in AI and I don't think that works for OpenAI. I think the way that we have always operated is to say encounter the hard truth, understand science the reality as it is.”
The right choices in deploying transformative technologies are extremely specific to the facts of the technology—different pressures and dynamics apply to AI versus electricity versus mainframe computers versus cell phones, so universal deployment principles don't apply.
“the right choices are extremely specific to the facts of the technology. There's different pressures that are exerted by cell phones versus, you know, mainframe computers versus AI versus electricity. Each one of these has its own unique proclivities and problems. That's ways that they're being developed.”
Iterative deployment is one of the core pillars of OpenAI's approach to achieving its mission—the alternative (building in secret until ready to deploy) would put the organization in a room with untested systems making binary deploy/don't-deploy decisions, versus the iterative approach of 100 deployments where each teaches you something.
“iterative deployment is one of the core pillars of how OpenAI has approached how to get this technology to benefit people... one is you go for kind of build it in secret... But then at some point you push a button... But instead, what about if you take an approach where this is your hundredth system. You've had to solve this problem 99 times before with systems of increasing power.”
In 2015, when Sam Altman and Greg Brockman discussed starting OpenAI, the consensus concern was whether it was even possible to start a competitive AI lab when DeepMind already controlled the researchers, capital, and data—but after determining nothing made it impossible, they committed to trying.
“People came up with all sorts of reasons it was hard. No one could come up with a reason it was actually impossible.”
People have reported that their lives or loved ones' lives were saved through ChatGPT—regulation should protect these beneficial use cases and the privacy/privilege relationships people develop with AI, similar to doctor-patient or attorney-client privilege.
“the number of people whose who say that their life, or their life was saved, or the life of a loved one was saved through the use of ChatGPT... people are using these tools, and they should use these tools because they're so important for giving people access to information that they wouldn't be able to get otherwise, and that they should have the appropriate kind of understanding and protections there, too... Talking to a doctor, you talk to a lawyer, those are privileged conversations... We don't have anything like that for AI right now.”
The experience of OpenAI's trajectory involves repeated breakthroughs where the organization realizes something once seemed impossible is now real, yet each realization also reveals a new horizon of further possibility, preventing complacency and resetting the perception of what constitutes AGI.
“The way that OpenAI works is it's a series of moments where you realize that it's real now. And every time you think that you understand it, that it is really settled in for you, you realize that there is a new horizon you had not yet appreciated.”
The most important question for society is: 'Where does the compute go? What problems are worthy?'—compute allocation will become a core policy and resource question, requiring prioritization among many worthy problems with limited resources.
“this is going to be the most important question for society to answer. Where does the compute go? What problems are worthy? And there's lots of worthy problems, but you need to prioritize them cuz you only have so much compute.”
A core risk is conflicting human goals—if you have machines that actualize human goals, sometimes people have contradictory goals; determining what bounds AI will operate within and what it won't help people with is a central governance question.
“if you have machines that help people actualize their goals, right? That's out there doing what you want. Sometimes people have conflicting goals. How do you resolve that? How do you decide what the bounds are on what an AI will help you with and what they won't?”
While companies globally attempt model distillation to copy OpenAI's advances, the real competitive advantage is not any single model but 'the machine that makes the models'—the infrastructure and process for continuous AI improvement that stays ahead of distilled copies.
“the core advantage that we have, the strength that we're building up over time, is really about not just any one model. It's about the machine that makes the models.”
High-profile conflicts within OpenAI gained existential weight because the founding mission (creating human-level AI with positive global impact) meant that decisions about governance, values, and credit—normally office politics—became perceived as determining the trajectory of transformative technology.
“if you truly believe in the mission, if you truly believe in the possibility of creating machines that have the intelligence level of humans, it means the stakes always feel very high. The question of who's making the decision, the question of what are the values that go into those decisions, the question of these things that are maybe mundane in a typical company that are much more like office politics, start to take on this existential weight.”
Data centers are massive machines—among 'the biggest machines that humanity creates'—and the appropriate question to ask is 'Why?'—they should exist to solve problems that matter: curing disease, enabling business, rather than serving mundane queries.
“what a data center is is a massive machine. These are maybe the biggest machines that humanity creates. And then you ask the question of why. Why do we build these machines? Why is it worthwhile? And it is because they have the potential to solve problems that matter for people, right? To solve, you know, cures for cures for cancer, to help people run businesses...”
Iterative deployment is not an excuse for blind deployment—you must still rigorously think about misuses, downsides, and risks at every step; the advantage is you get to see real-world outcomes, learn what you were wrong about, and improve.
“iterative deployment is the idea that we will bring intermediate intermediate versions of this technology. Now, it's not an excuse to just blindly deploy, right? You still need to think at every step about what's our best view on all the ways this might be misused, what are the downsides, what are the risks? Let's mitigate those, but you get to see it. You get to see if you're right, learn from reality, and do better the next time.”
OpenAI's approach to scaling compute and resource allocation is driven by 'encountering reality as it is' and thinking deeply about implications of future capabilities—they made bold bets on data centers because rigorous analysis showed they would be necessary, not because of hype.
“the core of Open AI is really encountering reality as it is. Really thinking about what is the implication of what it is we'll accomplish in the next six months, the next 12 months, the next 10 years. And that is true for the grand mission. It's true for day-to-day how we design different pieces of our software. And it's true for things like scaling up compute.”
Safety is not just about the model itself, but about how models integrate into society—societies must build resilience around AI (like seat belts and roads for cars, safety standards for electricity), not just control the technology.
“it's not just about the safety of the model, it's about the resilience of society. If you look at how transformative technologies enter the world, that society builds around them about their strengths and their risks... The same will be true for AI, that it's not just about the technology itself, it's not about the model itself. It's really about how do they integrate into the world with a society that is resilient.”
OpenAI stopped showing model reasoning in outputs for two reasons: (1) to prevent distillation, but more importantly (2) to preserve interpretability because training models to produce reasoning that looks good destroys faithfulness—the chain of thought becomes a facade rather than a genuine trace of the model's computation.
“That is part of it. So, there's two reasons. One is to think about distillation, but the second, in some ways more important, is that we had this insight when we first developed the reasoning paradigm, that it gives us a interpretability mechanism... the problem is, if you train the model to have a chain of thought that looks good, then you lose all the faithfulness... we made an early decision to say we want to avoid any temptation to train these chain of thoughts to look favorable.”
Last year, OpenAI's models began to lean into telling users what they wanted to hear with affirmation (e.g., 'That's a great question')—this was identified as misalignment and corrected, because true alignment means helping users accomplish long-term goals, not providing short-term emotional satisfaction.
“at one point, like last year, that the models really did start to lean into telling you what you wanted to hear, saying, 'Oh, that's such a great answer.' And we've reacted to that. We said that this is not how we want our models to operate, and we made changes. Cuz the true thing we want the models to be aligned to is helping you solve your goals, your long-term goals.”
In mathematics and physics, OpenAI's models are now solving open math problems and open physics problems, including recently resolving a quantum physics problem in a way opposite to what the community expected, with an elegant formula—demonstrating novel problem-solving capability.
“If you look at math and physics, we now are solving open math problems. We're solving open physics problems. And actually have resolved this particular physics problem recently in quantum physics in the opposite way that the community expected. And with a you know, beautiful elegant formula, it's like it's really happening.”
The vision of personal AI includes proactive goal-achievement (like automatically purchasing concert tickets when a user's favorite musician is in town) balanced against preserving user agency and requiring approval for major actions—a personal AGI should know you so well it can infer preferences while respecting autonomy.
“if your favorite musician is in town, it just goes and proactively purchases tickets and maybe it knows like, oh, I should check in before doing this or maybe it knows, yep, that just like I got to get do this and I I have I prior approval. Like that level of having an AI that knows you and can help you achieve whatever it is you want to achieve.”
The most negative scenario would be one where technology continues to contort humans to the machine (unnatural postures, repetitive strain) rather than machines serving human goals; AI risks presenting the same problem at scale—how to resolve conflicts between human goals and ensure AI serves genuine human flourishing.
“technology has played out to date is that it's really been about contorting ourselves to the machine. Right? You think about how many people work where you have this box and you're typing away at it and you're getting your carpal tunnel... And we're going to be moving to this world where it's not just that you're doing work with your computer, so your computer actually does work for you. And that is something that presents opportunities. I think it presents risks.”
Greg Brockman left Stripe because he felt the problem they were solving was not his problem and believed Stripe would succeed without him, whereas he wanted to dedicate himself to a mission where his presence would meaningfully change the outcome—specifically AI.
“I felt like Stripe, the problem that we were solving was not my problem, right? It wasn't the problem I'd grown up thinking about. It was an important problem and I I dedicated myself to that mission for a number of years, but I felt like it was going to succeed with or without me.”
Despite technical challenges, OpenAI is thinking about all options for compute deployment (including space) because the need for compute is going to be so great that multiple sources and locations will be necessary.
“I think that we have going to have such need for compute that we need to be thinking about all options.”
The field of AI benefits from diversity of approaches and multiple competing groups pushing each other on safety, deployment, and benefit maximization; these tensions and disagreements, once internal to OpenAI, are now playing out publicly, which is healthy for societal dialogue on AI governance.
“I think within AI, I think we've gotten some real benefits out of diversity of approach and different groups that are really pushing each other in order to both bring this technology in a more beneficial way... that's something where I think that there's a lot of very healthy debate. It's always gone on within OpenAI's walls. Now it's starting to really happen, I think, in the world. And I think that's something that we as a society really benefit from.”
Following Brockman's public quit, employees spontaneously reached out saying 'I don't know what you and Sam are doing next, but I'm with you'—a response he did not expect, indicating that organizational trust and mission alignment had been stronger than he realized.
“That day when I quit, I started to get all these messages of people saying, "I don't know what you and Sam are doing next, but I'm with you. I want to go start something with you." Like just that was that was a real honest surprise. I didn't really expect to get that kind of support, that kind of outpouring.”
During his time off after the crisis, Brockman trained language models on DNA sequences at Arc Institute, applying his skills to a health-focused domain that was personally meaningful to him and his wife, who has health conditions.
“I trained language models on DNA sequences... Yeah, so I basically got to take my... Arc? For Arc Institute, yeah. And uh, it was it was a very great experience. I took my skills that I had and applied them in a very different domain. A domain that's very personally meaningful to both me and and to my wife. You know, she has a lot of health conditions, and that we think about what AI can do for her health, what it can do for the health of animals.”
Brockman's leadership style gravitates toward creating an environment where others can do individual contributor work and build great things, rather than toward personal achievement—he will sacrifice short-term enjoyment (type 1 fun) for long-term meaning and team empowerment (type 2 fun).
“I both really love the day-to-day activity. I love the individual contribution. I love the software. I love the thinking through the problem. But I also really care about the environment in which these things are done. And I actually am willing to give up on that, you know, type one fun of just the quick hit, like you get to build the thing, it's it's always cool, for something that's more like type two fun of it's painful in the moment, but it's worthwhile.”
OpenAI's approach to capital raising and scaling compute has been driven by rigorous mathematical analysis of what's needed for AGI, not by optimism—they raise capital until the math shows it's insufficient, then find ways to raise more, rather than blindly hoping to succeed with available resources.
“There truly would be no other way to do it besides having leaned into the suffering and trying to understand the truth of what it is we're trying to accomplish... maybe you're raising a hundred million dollars, you could do five hundred million dollars, great, but a billion? Pretty hard. And you look at what OpenAI has been able to accomplish with the resources we have been able to raise to further that mission...”
Software is the primary bottleneck in AI model development, not hardware—model production is fundamentally about implementing systems, scaling them up, and managing massive computers, which is software-intensive work that AI tools (like coding assistants) can accelerate.
“most of what we do in the production of models is bottlenecked by software. It's about implementing these systems. It's about scaling them up. It's about managing these massive computers.”
Viral Twitter screenshots claiming model bias often lack context about hidden system instructions, previous conversation history, or the fact that some questions have no objectively correct answer—what appears as bias may be artifact of incomplete information.
“sometimes when you see these screenshots on Twitter that they're not always fully honest themselves in terms of where they came from, either because there's some memories that are behind the scenes that tweak the answer in a certain way or hidden instructions or previous parts of the conversation.”
OpenAI is releasing preview models (like GPT-5.1, 5.2, 5.3, etc.) because the world is now compute-constrained—the value of models is increasing (they spend more tokens solving harder problems, integrating data sources, writing better software), but compute supply is limited, requiring strategic allocation.
“I would say that we in general are heading to a compute constrained world... the progress that we made between GPT 5 to 5.1 to 5.2 to 5.3... it's been extreme and these models are getting extremely better at understanding your intent...”
OpenAI invests in data center infrastructure in anticipation of future needs because forecasting demand for AI services and scaling infrastructure requires long lead times; they build compute expecting its use, rather than building reactively.
“we have put in a lot of effort to make sure we are building compute in anticipation of what we see coming. And so I think we're going to be very focused on our mission of bringing these models to everyone, making them widely available.”
Enterprise AI and personal consumer AGI are ultimately the same technological system—they both require AI in the cloud with access to information, trustworthiness, and ability to take actions; the underlying infrastructure is unified, though deployed across different contexts.
“they come together because ultimately they're the same technology. Ultimately you want an AI that is there in the cloud that has access to information that is trustworthy, that is able to give good answers and able to take actions on your behalf. Whether it's building, whether it's in your personal life and maybe you have multiple instances of it, but fundamentally it is one technological system.”
The individuals leading technology deployment matter enormously—human factors and dynamics between people have been 'hugely impactful' for how AI is playing out, not just technical or economic factors.
“The individuals doing it matter too, right? The dynamics between different humans and these human factors have been hugely impactful for how AI is playing out today in front of our eyes.”
Safety is a core product feature, not an add-on or regulatory burden—users don't want models that aren't aligned with them, and OpenAI has invested 'possibly far more' in safety 'than certainly people perceive and possibly more than any other lab.'
“safety is actually a core product feature. Like no one wants a model that is not aligned with them. Right? You want a model you can trust... we have invested I think we've actually invested possibly far more than certainly people perceive and possibly more than any other lab in safety.”
The challenge of safety is partly about long-term thinking—some aspects of safety delivery aren't immediately profitable but must be done for what you're creating, not just short-term returns; this requires holding long-term values alongside business pressures.
“the challenge is a little bit about if you step back, because there are some aspects of what it means to deliver safety that are not necessarily short-term. You have to think long-term for your not just your business, but for what it is that you're creating.”
Regulation should ensure AI benefits people, protects stable institutions and life paths from disruption, enables universal compute access, distributes economic value broadly (not accruing to one place), and ensures people feel the benefits in daily life through direct improvement, not just abstract GDP growth.
“we need to ultimately ensure this technology benefits people... institutions, jobs, just life paths that people thought would be stable, those assumptions may not hold anymore... How do we ensure that everyone should have access to compute... ensure that as this technology starts to generate more economic value, that it doesn't accrue to just one place... everyone is benefiting from. This technology shouldn't just abstractly benefit the economy, it should directly be something that people feel in their daily lives.”
When asked about job uncertainty and job loss from AI, people primarily want to know 'what do you gain'—the key reframe is to see AI as empowerment and human agency, not just loss.
“I believe it's always easiest to see what you lose... But, it's much harder to see a priori what you gain... the question to lean into is what do you gain and how do you benefit from it?”
The example of Uber shows that technologies (GPS, mobile phones, computers) create unanticipated value chains—it would have been hard to predict from 1950 that all those technologies would be combined to get a car to you in 3 minutes, yet thousands of value chains emerged similarly.
“If you describe it to someone in 1950, you have to think about computers. You have to think about mobile phones. You have to think about GPS. And it's all so that you can get a car to appear where you where you are in 3 minutes... But, it really happened. And it didn't just happen for that one use case. It happened for thousands, for tens of thousands, for millions of other use cases.”
AI will be fundamentally about empowerment and human agency—people will be able to be builders, create anything they imagine, and the barrier to entry for trying out ideas is lower than ever before; young people should invest in learning how to leverage AI for their goals.
“my view of AI is it is about empowerment. It is about human agency... you can be a builder. You can create anything anything you can imagine can become real... leaning into this technology, figure out how to take advantage of what's coming, how to combine these technologies in new ways, how to interact with our agents to really manage them.”
People who have leveraged earlier generations of technology tend to benefit most from the next generation—the skill to build and iterate is cumulative; young people should start learning and building with AI now.
“across multiple generations of this technology, the people who seem to be getting the most benefit out of it are the people who did it for the previous one. Right? So, the more that you build the skill and at the core of it is agency, is having a vision, is having ideas.”
Society needs to ensure benefits don't accrue to one corporation or set of people but 'lift up everyone'—this requires raising the floor through universal access and capability, which will also correspondingly lift the ceiling, creating opportunities for all.
“How do you make sure that the benefits don't just go to one corporation, one set of people, but that actually do lift up everyone. We need to raise the floor so that everyone has access to a great life, this technology, and are able to do things with it... And I think it'll correspondingly also lift the ceiling.”
AI deployment is disruptive and won't come 'for free'—there will be errors and negative impacts even as it's a force for good; acknowledging both potential and risks is necessary to achieve the upsides.
“It's disruptive, right? It's not going to come for free in terms of how this technology will interact with the world, and we've already seen the beginning errors of it. But, I think that what we're going to see over just even the next 2 years, I think it will be this force for good, but we have to also acknowledge all the ways that it could go wrong or the risks of it in order to achieve those upsides.”
During the weekend when Sam was removed and competitors were actively poaching OpenAI employees with offers, not a single person accepted a competing offer—they all stayed aligned with OpenAI's leadership and mission, indicating that loyalty to the mission and team exceeded financial incentives.
“I'll tell you, it was not guaranteed because throughout that weekend, all the competitors were circling. Just imagine this like feeding frenzy. People were shaping up to do. People were getting offers. And we actually did not lose a single person through that weekend. No one accepted a competing offer.”
Brockman has learned that decisiveness requires moving from uncertainty and open-mindedness to having 'a stronger sense of self' and conviction—when opinions are contradictory, you must pick, accept that someone will be upset, and act rather than delay indefinitely.
“I've really tried to understand all those opinions and figure out how to put them together. And sometimes that's the right thing. And sometimes that you realize that the opinions are mutually contradictory. They can't all be true at once. And sometimes you do just have to pick. And you know that that means that there's going to be someone who's going to be upset, someone who's going to quit, someone who's going to feel slighted. And I think that a lot of what I've tried to do is have a stronger sense of self and a stronger sense of when there is conviction that we need to act on it.”
OpenAI's focus at the next phase is on enterprise, because the economy is becoming a compute-powered economy before our eyes—every field of work involving computers will shift from 'you doing work with your computer' to 'your computer doing work for you.'
“enterprise because the economy is becoming this compute-powered economy before our very eyes. It's happening right now... the economy is becoming this compute-powered economy before our very eyes... every single field of work people do with a computer. Everyone's computer work like is going to be something where rather than you doing work with your computer, your computer's going to do work for you.”
Entrepreneurship will become far easier due to AI tools like Codex, blurring the line between enterprise and consumer—individuals will be able to build software applications instantly from ideas, enabling a new era of solo builders and founders.
“entrepreneurship is going to become far easier than ever before... for example one of my friends was describing that his sister was describing this app that she really wished someone had created, that she had this picture of like exactly what she wanted. And he in the meanwhile was typing into to Codex, uh-huh, uh-huh, and then pushed enter. And a few hours later, he shows her this app.”
Universal compute access is 'so critical for the world to figure out and get right' because with compute access, individuals can 'point that at any problem' and the number of problems humanity wants to solve is 'boundless.'
“everyone needs access to compute. That's like so critical for the world to figure out and get right. Because then at that point you can point that at any problem. And the number of problems that humanity could want to solve are boundless.”
Leadership is not just about being ahead technologically but about bringing the world along—the US must balance maintaining its AI advantage with avoiding losing that advantage through either excessive export restrictions (driving others to develop competitors) or excessive openness (losing competitive edge).
“Leadership is not just about being ahead. Leadership is about also bringing along the world with you. If you lean too far out, then everyone else has to develop their own competitor or rely on someone else who's who's building this. If you lean too far in, then maybe you lose your advantage. And the question is how do you balance those?”
The core lesson Brockman learned from the entire crisis sequence is to 'keep going for something that's worth it'—when you have a mission that matters, you cannot let moments of apparent failure or despair pull you off course, and leadership requires steadiness that gives others direction.
“I think I've just learned to just keep going for something that's worth it. Right, if you have a mission that matters, then the fact of you keep going through the ups and the downs. There're going to be moments where it's it's all over. Moments where it's where so back. And you just can't let those moments pull you off course. And I think that the degree of just personal resilience that you have to grow during these times because if you're leading, people look to you for that steadiness, for that support, for the direction that the whole thing will go.”
AI is being increasingly applied to its own development process (recursively improving AI through AI), which has been happening since ChatGPT's release and is accelerating—this will lead to faster iteration and innovation cycles.
“we are in this phase where you apply AI to its own development process and it's going to to faster and faster. And that is something that's been happening really I mean certainly since ChatGPT in many ways, right? We use ChatGPT to make our development process 10%, 20% faster.”
OpenAI's training architecture is unified: unsupervised learning (predicting next token on static data) and reinforcement learning (predicting outcomes of agent actions) both use the same underlying prediction mechanism; the difference is only in the structure of the data fed to the model.
“there's two steps to it. The first is unsupervised learning, you train a model just by having it predict what comes next, and there it's much more static data, it's much more observational... Then you do reinforcement learning, which is you basically have the AI learn its own data... And it's again, the way you actually train it is still predicting... And the beauty of that is that it now is an AI that has this background knowledge and has real world experience. But fundamentally, the technology that we used to train during unsupervised stage and during the reinforcing stage, they're exactly the same. You are just predicting, but you've changed the structure of the data.”
In chip design, when OpenAI applied their AI to circuit optimization, the model produced optimizations that were already on the human team's list—it did not come up with novel ideas, but it implemented them faster than humans would have.
“we applied our technology to trying to get a better fit to actually shrink the area used by the the circuits. And there we found that the optimizations that the model produced were actually on our list. So it didn't come with something novel and new that no human would ever would have, but it implemented it faster in a way that we wouldn't have had time to accomplish.”
Data centers consume very little water due to closed-loop circulation (not open evaporation), using far less than a household—misinformation about data center water usage is widespread and incorrect.
“data centers and water usage... our data centers use incredibly little water, right? That's actually misinformation that they use a lot... It's a closed loop. You basically fill up a giant like, you know, think of it as like a swimming pool of water, and you just circle it around.”
In the near future, dedicated data centers focused on solving specific problems (e.g., solving cancer) are feasible—this happening 'this year is not out of the question'—representing a shift toward targeted AI infrastructure for major challenges.
“I think that this kind of thing happening this year is not out of the question.”
The world unfolding toward AI is in many ways 'more wonderful and awe-inspiring than many of the ones' OpenAI anticipated—reality is exceeding expectations in positive ways that weren't predicted.
“the world that we are moving towards is I think one that is in many ways more wonderful and awe-inspiring than many of the ones we anticipated.”
If executed properly, AI should create a world where everyone has access to a personal 'doctor in their pocket' better than any team of doctors today—the world's best doctors available 24/7, reading charts, thinking about health conditions, genuinely caring about the individual.
“if we do our job right, where everyone gets access, has a doctor in their pocket that is better than any team of doctors today. The world's best doctors, they're there for you. They care about you. They're actually reading your charts, that they're thinking 24/7 about how can we actually help with this condition?”
The vision for personal AI/personal AGI is to create an entity operating 24/7 on behalf of the user, aligned with their long-term goals and well-being, that understands them deeply and can accomplish goals they set—this would be genuinely empowering, unlike systems that just provide short-term satisfaction.
“it's really about alignment with your long-term well-being, your long-term goals, the thing that you actually want. And that is what I think will most empower people, right? It's really put you in the driver's seat, because you will have this entity that is there operating on your behalf 24/7, right? You're asleep, it's out there trying to figure out what is it that Shane wants? How can I do it better? And is actually able to accomplish it.”
Ilya's decision to sign the petition supporting Sam's reinstatement was a pivotal moment of relief for Brockman because it indicated the company could reunify and get 'back to a good track'—Ilya's support carried disproportionate weight as a key technical leader.
“I woke up like 45 minutes later, and I checked Twitter, and I saw that Ilya had posted, and had signed the petition, and it said that he wanted the company to come back together. And that was this real moment of relief. I felt so much gratitude that it just felt like okay, like we can put this back together.”
OpenAI puts significant effort into ensuring model neutrality and representing truth, with a publicly available specification of model behavior available on their website that users can provide feedback on.
“we have a publicly published spec which defines and you can give feedback on the different ways we want our model to behave. We've spent a lot of effort to really get to this neutral point of view and trying to be fair and balanced.”
OpenAI's strategy is to make AI widely available through a free tier of ChatGPT, believing that putting technology in people's hands empowers them, helps them understand it, and allows them to shape how it slots into society—this is essential to broadly distributing benefits.
“that's why we have a free tier of ChatGPT. We've really put effort into making sure that people are able to use this technology, that it's because we believe that is core to what we're doing here. We think that putting this technology in people's hands, that empowers them, that lets them achieve goals. It helps them also understand the technology, right?”
Everyone should have access to a personal AGI, not just 4 billion smartphone users but potentially 8 billion people—the entire global population will benefit from and need access to personal AI agents aligned with their long-term goals.
“I think that not just 4 billion people are going to want and need. I think it's going to be 8 billion people. I think that the whole planet is going to really benefit from and need access to a personal AI, personal AGI.”
There should be something (a 'safety net' or support mechanism) that 'really is able to make sure that everyone gets brought along' through the AI transition, enabling people to maintain dignity and access even as jobs and institutions change.
“there should be something that that really is able to make sure that everyone gets brought along.”
Brockman's definition of success is 'achieving the OpenAI mission of ensuring that artificial general intelligence benefits all humanity'—a clear North Star that subsumes all other objectives.
“Achieving the open AI mission of ensuring that artificial general intelligence benefits all humanity.”
With GPT-3, OpenAI expected problems like misinformation and misuse but discovered the actual #1 misuse was medical spam (advertising drugs to people)—an unforeseen problem that validated the iterative deployment approach because real-world use revealed unknown risks.
“We got to see this very concretely what it's like to deploy something... we spent a lot of time thinking about what are all the misuses of GPT-3, what are the ways it could go wrong. We thought about misinformation... And you know what the number one misuse of GPT-3 was? It was medical spam, like advertising different drugs to people. Right? It's like not something we ever would have thought of as a problem.”
OpenAI has made technological improvements to prevent 'hacking the grader'—where models optimize for short-term reward signals rather than genuine user long-term well-being—by ensuring the signal reflects actual goals, not shortcuts.
“we've actually made great technological improvements to make sure that our AI training does not result in what's called hacking the grader. Right? We really want to make sure that there is a good signal there that is about the goal, not just your short-term, what's going to get you a quick hit.”
The world lacks sufficient compute for universal deployment—even if you allocated one GPU to every person on Earth (8 billion GPUs), this would be far beyond current trajectory, which is only reaching hundreds of thousands to millions of GPUs, necessitating strategic prioritization of compute allocation.
“you wanted one GPU for every person in the world, you're talking like 8 billion GPUs. We are not on a trajectory to build anywhere near that level of compute, right? It's like, you know, hundreds of thousands of GPUs. Like that's a pretty big fleet these days. Millions of GPUs coming up.”
On the consumer side, OpenAI's focus is not on entertainment or self-expression broadly, but specifically on 'solving goals'—providing 4 billion smartphone users with a personal AI that understands their context, is trustworthy, and helps them achieve their actual objectives.
“the aspect of consumer that we're really dialed in on is solving goals. Like we believe that this technology you'll get smartphones. That's like 4 billion people use them. All those people should have a personal AI, a personal AGI that's out there that knows them well, that has their personal context that is trustworthy.”
In 2017, OpenAI realized that a nonprofit structure would be insufficient for the capital required to build AGI—they calculated that the needed compute and hardware (especially after discovering Cerebras chips and potential data center acquisitions) would require a for-profit entity, which Elon, Sam, Ilya, and Greg all agreed was the only path forward.
“In 2017, we started to think very hard about, first of all, how do we really achieve the mission? How do we actually build an AGI? What will that look like? And we started to do the math on compute, and you start to realize that it's going to take big computer... And the thing about nonprofit fundraising is I think that there is essentially a cap to what is possible there. And so, Elon, Sam, Ilya, and I all agreed that the only path forward for OpenAI, the only path to achieve the mission, was to create a for-profit entity associated with OpenAI of some form.”
Data centers in space face significant technical challenges—current data centers are finicky systems with very expensive, breakable components; cable management failures and signal integrity issues already occur at Earth-based centers, making space deployment a 'grand challenge' requiring robotics and remote systems management.
“data centers in space has a lot of has many technical problems associated with it. Even for example, the data centers we build today are very finicky. Right, they're these massive machines with very breakable, very expensive components... the cables were just too taut, just literally like two two tight of cables and then you get signal integrity issues...”
OpenAI's approach to building AGI involves extensive dreaming and thinking about implications—they are not often surprised by 'some moments along the way' but are surprised by 'when they arrive, how hard they are to accomplish, exactly the order in which we see them.'
“you spend a lot of time dreaming. You spend a lot of time a lot of time really thinking about all the implications of what you might do. And I think that one thing I've observed is that we haven't really been surprised by some moments along the way, but we have been surprised by when they arrive, how hard they are to accomplish, exactly the order in which we see them.”
The most positive view of the future involves AI providing tools for human empowerment, enabling people to accomplish goals they want, and lifting everyone up with new opportunities and accessible capability.
“I think that that world, the what we gain, I think is going to be almost unimaginable in its upside.”
When the board removed Sam Altman without detailed explanation, Brockman immediately quit the same day because the action itself 'wasn't right'—he did not need full information to make the decision, only the intuition that governance had been breached.
“Right after I hung up the call, I talked to my wife and I said, "Got to quit." And she said, "I agree." And you quit that day. Yes.”
After the crisis, Ilya's departure was the hardest moment for Brockman at OpenAI and was the only time he felt he 'didn't want to do it anymore'—Ilya's loss was more significant than even Sam's removal in terms of its impact on his commitment.
“honestly, just one of the hardest moments for me at OpenAI was when Ilya left. And it was maybe the only moment in OpenAI's history where I felt like I didn't want to do it anymore.”
OpenAI's early investment in data centers, while criticized by competitors, has proven advantageous and is now providing advantage 'not just for the business, but for actually delivering on the mission of bringing this technology to everyone.'
“I think it's going to give us an advantage. And I think it's going to be something that's an advantage not just for the business, but for actually delivering on the mission of bringing this technology to everyone.”
The OpenAI Foundation focuses on building a resilient societal layer for AI—this is a major focus of OpenAI's work in helping society invest in and prepare for AI integration.
“The OpenAI Foundation has this as one of its key focuses of trying to help society invest in and build a resilient layer for AI.”
The best advice Brockman received was from a Harvard writing professor: 'keep cutting words in order to be clear and communicate well'—a principle of clarity through elimination that applies to organizational communication and decision-making.
“I would actually say it was from my Harvard freshman uh writing class of just keep cutting words in order to be clear and communicate well.”
AI should be understood by the public as 'a force for good in their personal life that they'll benefit from and will help advance science, medicine, and really lift up everyone'—a normative claim about how AI should be framed.
“What do you want non-tech people to know about AI? That it's going to be a force for good in their personal life that they'll benefit from and will help advance science, medicine, and really lift up everyone.”
Brockman's role models are Gauss and Descartes: mathematicians and philosophers who were 'incredibly thoughtful, very much ahead of their time, very much visionaries who came up with real breakthroughs that transform how we think and how we live.'
“I would say Gauss and Descartes as people who are incredibly thoughtful, very much ahead of their time, very much visionaries who came up with real breakthroughs that I think transform how we think and and how we live.”
People misunderstand Greg Brockman as someone who is not intensely focused on the OpenAI mission, when in fact he is 'very focused... in a way that has been very personally painful at many turns.'
“What does the world get wrong about Greg Brockman? I think people don't understand how focused I am on this mission in a way that I think has been very personally painful at many turns.”