Apr 22, 2026 · 1h 12m · knowledge-project
Ai Goes Parabolic | OpenAI Co-Founder Greg Brockman
0:00 / 0:00
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this in-depth conversation with Shane Parrish, OpenAI co-founder Greg Brockman examines the founding vision, technical breakthroughs, and boardroom turmoil behind OpenAI rise, while detailing the compute infrastructure and iterative deployment required to deliver beneficial AGI.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Shane holds 12% of the talking time here. How this is scored →
Shane as informed peer 4.0 Guest teaching 2.9 Guest disagreement 0.9 Shane pushing back 1.4
speaking balance: gold is Shane, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 44:02 Greg dismisses competitor doubts on compute investments
Greg delivers a sharp, confident retort against industry critics who teased OpenAI's massive hundred-billion-dollar infrastructure spending.
Hardest push from Shane ▶ 47:52 Shane challenges compute prioritizationShane directly confronts Greg on the moral and practical contradictions of rationing compute for image generation while cancer remains unsolved.
Biggest teaching moment ▶ 40:39 Greg explains hidden chain of thought rationaleGreg corrects simple public assumptions by explaining that exposing chain-of-thought damages interpretability because models optimize intermediate thoughts to please users.
Shane holds their own ▶ 10:05 Shane distinguishes reasoning from predictionShane articulates a sharp technical distinction between next-token statistical prediction and first-principles deductive reasoning, prompting Greg to explain their RL connection.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Shane as informed peer | Guest teaching | Guest disagreement | Shane pushing back | Why |
|---|---|---|---|---|---|---|
| Leaving Stripe and the Genesis of OpenAI | 4 | 2 | 1 | 2 | Shane prompts Greg with specific context about his transition from Stripe, Patrick Collison's involvement, and early timeline details. Greg elaborates openly about founding OpenAI and recruiting early talent without friction. | |
| DeepMind Dominance and Transitioning to a For-Profit Entity | 4 | 4 | 1 | 1 | Shane asks pointed questions about DeepMind's moat and the pivot from non-profit to capped-profit. Greg explains the compute requirements and the technical breakthrough of the sentiment neuron paper. | |
| OpenAI Five and the Power of Scaling Simple Algorithms | 5 | 5 | 2 | 2 | Shane offers an insightful observation contrasting Dota with chess and AlphaGo, and questions the distinction between next-token prediction and first-principles reasoning. Greg explains how massive compute on simple RL algorithms drives emergent intelligence. | |
| High Stakes, Internal Friction, and Lab Fragmentation | 4 | 2 | 1 | 1 | Shane probes into when internal tensions emerged at OpenAI due to the high stakes. Greg validates Shane's framing and describes the fragmentary nature of AI labs. | |
| Sponsor Break: CoinShares Digital Asset Management | 4 | 2 | 1 | 2 | After midroll sponsor reads, Shane drills down chronologically into the exact sequence of events when Sam Altman was abruptly fired by the board. Greg shares intimate details of the call. | |
| Corporate Rebellion, Microsoft Life Raft, and Ilya Reversal | 3 | 1 | 0 | 0 | Greg delivers an uninterrupted narrative detailing the weekend of the OpenAI corporate crisis, employee rebellion, Microsoft backup plan, and Ilya Sutskever's public reversal. | |
| Mending Bonds with Ilya and Team Solidarity | 4 | 1 | 0 | 1 | Shane asks empathetic questions about rebuilding trust with Ilya and references Bill Belichick's philosophy on team loyalty. Greg agrees and reflects on leading from the front. | |
| Personal Sabbatical and Training Biological AI Models | 4 | 3 | 2 | 1 | Shane asks about Greg's sabbatical and notes his blog post on self-study. Greg gently clarifies that he already knew how to train models and specifically applied them to DNA sequences at the ARC Institute. | |
| Confronting Reality, Embracing Suffering, and Capital Demands | 3 | 3 | 1 | 1 | Shane asks Greg to double-click on Ilya's philosophy of suffering. Greg explains that confronting hard physical truths and raising massive capital are essential to avoid Silicon Valley delusion. | |
| Rapid Fire: Life Lessons, Role Models, and Public Perception | 4 | 3 | 1 | 2 | Shane runs a rapid-fire questioning round touching on role models, model naming flaws, and coding automation. Greg gives crisp answers and explains how AI is accelerating its own engineering loop. | |
| Model Neutrality, Anti-Sycophancy, and Long-Term Alignment | 4 | 3 | 1 | 2 | Shane pushes on why models display political bias and whether RLHF incentivizes sycophancy. Greg explains OpenAI's model spec, anti-sycophancy efforts, and focus on long-term user alignment. | |
| Geopolitics, Distillation Defense, and Hidden Chain of Thought | 5 | 3 | 1 | 2 | Shane brings up global AI competition, sovereign AI, distillation threats, and asks if distillation is why OpenAI hides intermediate reasoning. Greg explains the dual security and interpretability reasons behind hidden chain of thought. | |
| Compute Bottlenecks and OpenAI Massive Infrastructure Bets | 5 | 2 | 2 | 2 | Shane highlights that OpenAI was initially mocked by competitors for pouring massive capital into data centers. Greg highlights the foresight of their bet and notes competitors are currently struggling with compute capacity. | |
| Sponsor Break: HeyGen AI Video Generation | 4 | 3 | 1 | 2 | Shane questions how compute should be allocated between frivolous consumer requests and existential medical research. Greg outlines OpenAI's commitment to keeping compute broadly accessible. | |
| Enterprise Transformation, Codex, and Ubiquitous Personal AGI | 3 | 3 | 0 | 1 | Shane inquires about the internal strategic division between enterprise and consumer AI. Greg details how tools like Codex democratize software engineering for eight billion individuals. | |
| Physical Infrastructure Challenges and Orbital Data Centers | 4 | 4 | 1 | 1 | Shane asks about orbital data centers and iterative deployment. Greg explains physical engineering realities, such as cable tension causing signal degradation, and recounts discovering medical spam as the top GPT-3 misuse. | |
| Safety as a Core Feature and Societal Resilience | 4 | 3 | 1 | 1 | Shane asks how the market will handle competing models with differing safety philosophies. Greg argues safety is an indispensable product feature and emphasizes building broader societal resilience analogous to seat belts and electrical codes. | |
| AI Governance, Legal Privilege, and Infrastructure Realities | 4 | 4 | 1 | 1 | Shane asks about AI regulation and adds an accurate fact about datacenter water recycling. Greg elaborates on legal privilege frameworks for AI interactions and debunks water consumption myths. | |
| Economic Uncertainty, Agency, and Future Labor Shifts | 4 | 3 | 1 | 2 | Shane raises widespread public anxiety over job security and technological disruption. Greg uses the historical Uber analogy to demonstrate that technological change creates unpredictable economic gains and rewards personal agency. | |
| AI Corporations, Pocket Physicians, and Potential Pitfalls | 3 | 3 | 0 | 1 | Shane asks about high-value skills for young people and explores both utopian and dystopian trajectories. Greg paints a picture of managing autonomous AI agents and universally accessible medical specialists. |
Statements from this episode (39)
Brockman: OpenAI's initial planned core team included Dario Amodei and Chris Olah
“Initially, the set of people that I narrowed down to were actually Ilya, Dario, Amidai, Chris Ola, and myself. That was going to be the team.”
Brockman: OpenAI's 10-year roadmap focused on RL, unsupervised learning, then complexity
“We came up with what I would Really say is almost the technical plan that we have 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…”
Brockman: Elon, Sam, Ilya, and I Agreed OpenAI Needed For-Profit Structure
“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. And so we were committed to that direction, and that is something that we kne…”
Brockman: 2017 sentiment neuron proved semantics emerge from next-character prediction
“Because 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 sentimen…”
Brockman: GPT-4 met previous AGI criteria despite clearly not being AGI
“It clearly wasn't an AGI. It was lacking something, but just if you'd describe your criteria for AGI two months prior, it probably would have been compatible with what GPT-IV was”
Brockman: OpenAI Five neural net was comparable to an insect brain
“And by the way, the neural net we used, tiny, tiny little insect brain, similar number of synapses as to truly an insect brain.”
Brockman: Predicting Einstein's Next Words Requires Intelligence Equal to Einstein
“On the one hand, 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.”
Brockman: OpenAI Uses Same Prediction Technology for Unsupervised and RL Stages
“Fundamentally the technology that we use 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.”
Brockman: Belief in Human-Level AI Turns Mundane Politics Into Existential Conflict
“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…”
Brockman: Frontier AI Development Naturally Causes Team Splintering Under Pressure
“This technology is by nature very fragmentary, right? That it's sometimes it, you know, like when you have a lot of pressure, you can get a diamond or you can get cracks. Often you'll see diamonds form in pockets, right? Teams of people that really work togeth…”
Brockman: OpenAI board gave zero explanation when firing Altman and removing Brockman
“I was told that the board has decided That Sam would be removed, and effectively the message that I got was the same messaging that was in the public post, and I asked if I could have any more information. I was told no, not right now, and I pressed on that ma…”
Brockman: Altman, Pachocki, Sidor, Madry, and Brockman Charted New AI Venture
“There were a few of my close collaborators who quit that day as well. That's Jakob, Shimon, Alexander, and the five of us. So Those people plus Sam. We all got together and we started to chart out what a new company could look like.”
Brockman: Altman Asked Satya Nadella to Fund Venture Absorbing All OpenAI Staff
“Sam talked to Satya, who we'd been talking about. Hey, can you be a funder? Could you know, help, help support this new endeavor? I was like, Hey, actually, can we expand from the small life raft to like, yes, can we take everyone? And we're like, all right, w…”
Brockman: OpenAI Staff Reinstatement Petition Crashed Google Docs
“So many people were trying to sign the petition at once, it actually crashed Google Docs. And so you had to have certain people who were designated as the person you go to actually put your name on the document so you don't have too many editors at once.”
Brockman: Ilya Sutskever officiated his civil wedding ceremony
“You'd been the officiant at my civil ceremony, right?”
Brockman: OpenAI lost zero employees to competitors during the 2023 board crisis
“People are getting offers and we actually did not lose a single person through that weekend. No one accepted a competing offer.”
Brockman: Sutskever leaving was the only time I considered quitting OpenAI
“And 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.”
Brockman trained DNA language models for ARC Institute during sabbatical
“So I trained language models on DNA sequences... For ARC Institute, yeah.”
Brockman: Building true value requires enduring suffering
“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.”
Brockman: AI will soon invent its own research ideas and run experiments
“And we're going to be hitting a phase soon where the AI will also come up with its own research ideas and test those out, run experiments.”
Brockman: Code written without AI is now a vanishing fraction
“It's hard to know what percent of the code is not written by AI. It's a vanishing fraction. The actual Writing of code. Currently, the AI is much better than humans at writing code, given the right context, given the right structure. Now there's parts of the a…”
Brockman: OpenAI models recently resolved an open quantum physics problem
“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.”
Brockman: Twitter screenshots alleging AI bias often omit hidden context
“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 …”
Brockman: OpenAI models previously leaned into sycophancy before training adjustments
“We've actually gone through an evolution of how we train the models to user preferences, and that we've seen that 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 a…”
Brockman: Distilling AI models fails because frontier progress is exponential
“There's certainly a lot of attempts to distill models. And that comes from companies in the U S it comes from all over the world. But I think that it misses the core point, which is that the way this technology is developing is it is on an exponential. And any…”
Brockman: OpenAI hides chain of thought to preserve faithfulness and prevent distillation
“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 we had not been anticipating, beca…”
Brockman: The world is heading into severe compute constraints
“I would say that we, in general, are heading to a compute-constrained world.”
Brockman: Global GPU production is nowhere near one GPU per person
“If you just wanted enough compute for, you know, you wanted one GPU for every person in the world, you're talking like eight billion GPUs. We are not on a trajectory to build anywhere near that level of compute, right? It's like, you know, hundreds of thousand…”
Brockman: OpenAI competitors are struggling with compute availability
“I mean, I think our competitors are not having a good time on compute, let me put it that way.”
Brockman: Data centers dedicated to curing cancer could emerge this year
“I think that this kind of thing happening this year is not out of the question.”
Brockman: Allocating compute will be society's most important question
“Well, 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, because you only have so much compute.”
Brockman: Overly taut cables caused signal integrity failures in OpenAI clusters
“We've had many issues in the past where the cables were just too taut, just literally like too, too tight of cables. And then you get signal integrity issues and the computer doesn't work.”
Brockman: GPT-3's number one misuse was medical drug spam
“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.”
Brockman: AI Users Need Legal Privilege Protections Similar to Doctors and Lawyers
“You talk to a doctor, you talk to a lawyer, those are privileged conversations, right? You feel comfortable sharing them. There's certain guardrails on when the healthcare provider would have to, you know, provide that information to law enforcement or alert s…”
Brockman: The United States Is Not the Global Leader in Robotics
“That you think about robotics where I think we are not the leader.”
Brockman: OpenAI Has Committed to Ensuring Data Centers Do Not Raise Electricity Prices
“And you think about things like data centers, that those are something where there's clearly been a lot of concern about questions like, do they drive up electricity prices? And we have a commitment to ensure that they do not.”
Brockman: Claims that OpenAI data centers consume massive water are misinformation
“Like a good example is data centers and water usage. Like that, that's something that people talk about a lot, but actually our data centers use incredibly little water, right? That's actually misinformation that they use a lot.”
Brockman: People will manage AI agents and lead autonomous AI corporations
“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, right?”
Brockman: Everyone will have a pocket AI doctor superior to top medical teams
“We should be in a world, 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,…”