Greg Brockman interview [video]

9 hours ago (fs.blog)

building products on top of their api makes these drama weekends terrifying. really makes you realize how fragile your whole stack is when a board decides to act up.

I just don’t understand why a non-profit was allowed to do this. Does this not set a precedent that non-profit doesn’t actually mean anything? You can just use a favorable structure until it’s time to enrich yourself.

  • I think it would be helpful for you to clarify which part of the chain you found objectionable:

    - in 2015, OpenAI was founded as a Delaware nonprofit

    - in 2017, OpenAI discovered the scaling laws and realized they needed far more compute (and thus money) than they had initially anticipated

    - that discovery precipitated a series of negotiations between the founders on how to restructure OpenAI to raise more money for compute, ultimately resulting in Musk’s departure when the other founders would not give him control

    - in 2018, OpenAI attempted to dramatically increase its fundraising despite Elon ending his contributions, but raised only $50M of its $100M goal

    - in 2019, OpenAI created a capped-profit subsidiary in order to attract funding from commercial entities

    - the nonprofit hired an independent assessor to value its IP, and then transferred that IP to the for-profit for fair value (around $60 million in 2019)

    - the OpenAI nonprofit received a right to 100x capped return on its IP investment, or $6B, once the for-profit began making a profit. The nonprofit also received the right to the residual profit after all future investors reached their caps

    - in 2019 OpenAI’s capped-profit received $1B in investment from Microsoft. OpenAI later received $2B from Microsoft in 2021 and $10B in 2023 as compute scaling continued

    - Microsoft received a cap of 20x on its $1B investment, and 6x on its $2B and $10B investments, for a total of $92B target redemption

    - in 2025, OpenAI’s for-profit entity recapitalized from a capped-profit entity with residuals flowing to the nonprofit to a traditional public benefit corporation with traditional equity

    - in exchange for the residual (and 100x profit cap on the original $60M transfer) the nonprofit received a 26% equity stake in the for-profit. That stake is currently valued at around $200B

    All of the above is from the record in Musk v. Altman, thanks to which we now have all the details. The upshot for the nonprofit is that it transferred IP worth around $60M in 2019 for rights to $6B in future profit, and then ended up with $200B in equity after the recapitalization. I see a lot of people in this thread assuming that the nonprofit no longer exists, which is not true.

    • The objectional part would be:

      - in 2019, OpenAI created a capped-profit subsidiary in order to attract funding from commercial entities

      Particularly if it creates a conflict of interest for anyone making decisions on behalf of the nonprofit

    • The OpenAI non-profit is now one of the biggest non-profits in dollar-denominated assets. If the goal was to make the non-profit really big and well-funded then that seems on track. But not clear to me what it would do to advance its mission.

  • Non profits have always been able to have for profit subsidiaries, owned by the non profit.

  • Most startups don't actually make profits and nonprofits can't give equity so it's not really a favorable structure.

    • It’s a favourable structure in many cases.

      Not everything is a business.

      OpenAI wasn’t, until it was.

  • This case was the first of its kind and it was never tested if OpenAI breached their charitable mission and the case was dismissed due to the statute of limitations.

    Other than researchers, nobody from big tech would ever see themselves wanting to work at a charity / non-profit. The moment the VCs came into the picture then all the grifters poured in and AGI meant IPO.

    > You can just use a favorable structure until it’s time to enrich yourself.

    Maybe that diary was made out of teflon.

  • Nonprofit doesn't mean anything, since people can just route the profits into salaries. It's just another legacy regulation that may have once once had a societally-constructive purpose that wealthy people just use as one of the array of financial tools to help implement their latest scams. IMO, here are no legitimate nonprofits.

    Western countries have been utterly strangled by nonprofits. Governments fund them with tax money in order to lobby themselves for legislation that financially benefits individuals in government and their donors. Obama even expanded the rules in the US to allow the government to unconstitutionally fund religious groups to accomplish functions that belong in government.

    They should all be either reformed so that their internal bylaws and compensation are strictly regulated or probably preferably, they should simply be destroyed. If you only pay taxes on your profits (and we get rid of legal vehicles to hide profits) and your employees are obligated to pay taxes on their incomes, there's no need for a nonprofit status. If nonprofits want to engage in business (religions included), let them pay taxes. If they engage in charity, they won't have anything to tax.

    • One reform I would make would be to limit tax breaks to actual charitable activity within an organization, instead of a blanket tax break to the whole organization. For example if a Church/Hospital runs a soup kitchen and homeless shelter, those resources should be tax free, but maybe the rest of their activities shouldn't be by default.

      Another reform I would make would be around independent governance and removing donor control of charities to reduce the number of sham Rich Guy foundations.

    • > Western countries have been utterly strangled by nonprofits

      To expand, there are two major problems with nonprofits in Western nations these days:

      1. Governments use them as a way to do things that they themselves are not allowed to do ("it's private charities that do this!", ignoring the fact that the charities get >90% of their revenue from government grants)

      2. Like you mentioned, the government grants to nonprofit back to politicians' campaign funds pipeline. Utterly egregious.

      > Obama even expanded the rules in the US to allow the government to unconstitutionally fund religious groups to accomplish functions that belong in government

      I wasn't aware of this being a big concern; more the other way around, like in my point 1.

  • It should not surprise you to learn that Greg Brockman is a Trumper and major donor.

    It should also not surprise you that the Epstein files have not been released.

    Everything is possible and not possible in a corrupted system.

I remember when computer magazines were aimed at programmers and had code listings in them.

Then there seemed to come a time when all they talked about was the IBM vs. Microsoft lawsuit. From then on they must have felt that they had discovered a formula, because all they ever yapped about after was insider baseball of computer companies.

I find this sort of corp. vs. corp. coverage boring, sort of like techie reality TV. Who will be voted out tonight, Debra, or Deborah...?

  • I remember when wired changed editors. After Chris Anderson left it became "gq but we talk about iPhones"

  • I dunno man, there are two "tech" industry worlds and the one you and I -- hacker types -- think of as "tech" is not what the rest of the world means. They mean the other one, which has almost nothing to do with the actual technology and instead everything to do with the absolutely apeshit amount of money, power, influence and intrigue that the technology enabled.

    This was a very large apeshit $$ amount back in IBM vs Microsoft but the scale of it now in the era of e.g. OpenAI etc is beyond imagination.

    There's a whole generation of people whose association with the engineering/technology side of things only happened because of their interest in the other side of things.

    I too miss old Byte magazine days.

    • This happened when business types realized technology could be enormously profitable (~PC era).

      Then inevitably, tech news turned into business news.

Why can't someone ask what happened in Ilya's mind. Firing Sam and then signing the solidarity letter of Sam to leave OpenAI if was fired. Other than that, all other information seems kind of just going over the surface.

As far as Brockman account of the past goes, there's also his personal diary which was made public as a part of that lawsuit by Musk. Includes for example the line: "Financially what will take me to $1B?". BTW, if you don't know, Musk lost it because he filed too late, lol.

  • If his entire personal diary got exposed and that's the worst that's in it, good for him.

    • What about stealing 12 million books of copyrighted human culture, at massive scale, and then enclosing the value created inside proprietary, investor-backed systems? Something wrong with that?

      What happens if you go tomorrow, downtown San Francisco, and leave a bookstore with one book without paying?

         "Behind every great fortune there is a crime"
               - Honoré de Balzac

      3 replies →

    • I'm curious what you're writing in your diary that's worse than blatantly admitting to fraud of this scale. He publicly misled people about OpenAI's "mission" as a nonprofit, while seeking to enrich himself to the tune of $1 billion(!!!) dollars.

      Also, his entire diary was not in fact made public. The attorneys only quoted the parts that were relevant to the case, which pertained to OpenAI's transition from non-profit.

      4 replies →

what's wild is they accidentally solved it — pretraining IS unsupervised learning at scale, RLHF IS reinforcement learning. they just didnt know the recipe yet

  • pretraining isn't unsupervised, it is self-supervised - meaning it is moderately more scale limited.

    • What would unsupervised mean, would unsupervised be something like alphago playing against itself trillions of times?

      Whereas self-supervised, allows learning without explicit annotation of data ; but it doesn't matter if the models already trained on the entire Internet, and it's not like a game where it can come up with effectively new training data for itself?

    • fair point — OpenAI's original plan literally said "solve unsupervised learning". the self-supervised distinction wasnt really standard til after BERT/GPT popularized it

not sure why but this episode feels v boring perhaps because he didn't share anything unexpected / unknown

  • I found it more interesting than I thought because I thought it'd all be stuff I knew but a lot was new to me.

Point of order: Anthropic is the most important AI company now.

  • People discount Google/Deepmind but a lot of the original research was done there including inventing transformers which form the basis of the other AI companies.

  • Anthropic is the most hyped AI company now. Their models aren't the best, but their marketing sure is.

  • It really does feel like OpenAI has lost their leadership. I haven’t used a model of theirs, let alone an app, in months.

Thankful for the mention of "AGI" in the first lines as I can bail out from reading the rest.

Whatever AGI is, it "AGI" is not glueing a load of text prediction machines together.

  • > Whatever AGI is, it "AGI" is not glueing a load of text prediction machines together.

    K don't think it would be that simple either, but for now we simply don't know.

    I would like to think that what I consider my intelligence will always be distinguishable from a cleverly built harness wrapped around text prediction, but I can't say for sure that's guaranteed.

    • Yes we do know. We’ve fully explored the possibilities of LLMs and they are nowhere.

      The latest efforts like agents are clearly showing the limitations and are nowhere near AGI.

      We’ve now reached the buzzwords and bullshit stage of the bubble where they cast around for problems shaped like the solution.

      1 reply →

  • How do you know that?

    • Shannon Got AI This Far. Kolmogorov Shows Where It Stops - Vishal Misra

      https://medium.com/@vishalmisra/shannon-got-ai-this-far-kolm...

      This article explains what's missing in terms of two kinds of complexity that oppose: Shannon complexity vs. Kolmogorov complexity.

      It introduces the opposition by an example of driving the value of pi as decimal number, which has no pattern and high complexity, and a formula for deriving pi that does have a pattern with low complexity, then observing that mind can work from the patternless high-complexity back to the patterned low-complexity without prior examples, while AI can't.

      LLMs encode and retrieve patterns in the training data, and doing so can connect data to the terminology of known principle, but mind can observe inconsistencies in data and to reason from first principles to resolve the inconsistency.

      The distinction between these two modes can seem blurry as AI can traverse the patterns of the known in ways that are extraordinarily revealing, but it's not structured to reason about the unknown.

      Inference is not sufficient for reason.

      For example, a conventional algorithm can search for patterns in text at a scale many orders of magnitude beyond a mind's capacity, and this can be very revealing, but to do so this algorithm need not read the text with comprehension.

      Regarding the question: can genAI be enhanced to reason? The answer is assumed to be "no", due to the categorical opposition of the two kinds of complexity and the lack of understanding of structures within genAI to handle the reasoning.

      Read the article, which includes other examples including a jump from Newtonian to Einstein physics in the history of astronomy, and a noodling on how to talk about the edge of the unknowable in AI.

>So many people were trying to sign the petition at once that it actually crashed Google Docs

I still wonder how much peer pressure was behind that. Like, what if you think Sam is a scumbag and you're glad he's gone, but people are waving this petition in your face. What would you do? It would be really bad for you if the emperor returned and you were one of the few who didn't sign it.

Also, going by this video, the first order of business for an AGI should be finding a cure for hair loss.

Obviously OpenAI betrayed their stated mandate by going to the largely closed-source API-access business model, which is the one that Anthropic, Google, and xAI also adopted, i.e. the high-margin hosted model businesses, while DeepSeek, Alibaba/Qwen, Baidu/ERNIE, and Tencent/Hunyuan are breaking that model by releasing various open weight models.

I think the Chinese labs have a fundamentally different viewpoint: they’re building infrastructure, and looking at it more like how a US corporation might like having some of its employees making core contributions to compiler ecosystems like LLVM/clang and so on. The payoff is down the road, partially reputational, but also having a great compiler is good for everyone in the computational business world. The rentier-finance capitalist instead wants to privatize the compiler and extract rents for access.

The thing about infrastructure is this: you don’t get a direct financial return on investment in infrastructure (think roads, which make other economic activity possible) unless you have some ridiculously corrupt system controlled by rent-collectors (which is how the US electricity grid and fiber optic backbone works). That’s all the major US LLM providers are doing: trying to collect rents on systems that were built using the global human knowledge base as inputs.

At the very least OpenAI should be releasing their older models on a steady timetable. Sure it might reduce some revenue streams but it would be good for their reputation.

Unfortunately they survived, not going to spend time with this.

From my point of view they are yet another big tech bros company.

Not feeling like 1 hour of my Sunday is worth listening to this, do anyone have the non-clickbait answers to the two "previews" mentioned in the description?

> Greg explains how the original Napa offsite produced the three-step technical plan OpenAI has followed for a decade and the real reason OpenAI had to abandon its pure nonprofit structure

What was the technical plan and what was the "real reason" they couldn't achieve their original goals?

  • > What was the technical plan

    "1. Solve reinforcement learning

    2. Solve unsupervised learning

    3. Gradually learn more complicated 'things'"

    That three point list is verbatim the extent of the technical plan mentioned.

    > what was the "real reason" they couldn't achieve their original goals?

    Paraphrasing, "we needed more money for compute and didn't think we could get enough as a non-profit". Brockman's diary might be a stronger indicator of the real real reason, though.

  • >> What was the technical plan and what was the "real reason" they couldn't achieve their original goals?

    Because they were still downloading from Anna's Archive and the lawyers were in panic?

  • 1. Solve reinforcement learning.

    2. solve unsupervised learning.

    3. gradually tackle more complicated things.

    > what was the "real reason" they couldn't achieve their original goals?

    I assume this is referring to why they gave up being a non-profit. The answer is that they needed more money.

    • Huh, I guess ML people weren't aware of "divide and conquer" that has been successfully employed in software engineering since basically forever?

      > I assume this is referring to why they gave up being a non-profit. The answer is that they needed more money.

      Ugh, that was more boring than even I expected, thanks a lot for saving me the time though, seems avoiding watching the full thing was worth it.

      2 replies →

    • > The answer is that they needed more money.

      isn't it still an odd choice for a nonprofit? it's hard to imagine a world without OpenAI and ChatGPT now, but at some point they decided being the best is most important. and presumably most profitable, since why just need a little more money?

      3 replies →

  • https://www.youtube.com/watch?v=6JoUcQ1qmAc

        00:00:00 Introduction
        00:00:49 Meeting Sam Altman and Starting OpenAI
        00:02:40 Building the Founding Team
        00:04:25 DeepMind's Lead Over OpenAI
        00:04:54 The Change from a Pure Non-Profit
        00:06:05 Breakthrough Moments at OpenAI
        00:08:22 What Dota 2 Meant for OpenAI
        00:10:04 Reasoning Versus Prediction
        00:11:59 Tensions Grow at OpenAI
        00:15:44 Sam Altman's Firing
        00:17:49 Greg Quits OpenAI
        00:19:56 Sam Explores Deal with Microsoft's Satya
        00:20:28 OpenAI Employees Sign Petition for Altman's Return
        00:23:43 Ilya Sutskever Leaves OpenAI
        00:24:59 Lessons Learned in Leadership after Sam Ousting
        00:28:22 The Thing Ilya Said that Greg Can't Forget
        00:32:22 Is AI Going Parabolic?
        00:33:24 How Much of OpenAI's Code is Written by AI?
        00:36:21 Are AI Chatbots Just Telling Us What We Want to Hear?
        00:38:06 The Global AI Race to Reach AGI
        00:38:40 What Happens if US Doesn't Reach AGI First?
        00:39:49 Are Competing Countries Stealing AI Advancements from U.S?
        00:40:38 Why ChatGPT No Longer Shows Reasoning
        00:41:47 The Finite Constraints of Compute
        00:43:38 On Investing Early in Data Centers
        00:46:31 The Future of Data Center Specialization
        00:47:52 How OpenAI Will Decide Whose Queries to Serve
        00:49:08 OpenAI on Consumer vs Enterprise Models
        00:53:05 Data Centers in Space?
        01:00:56 What Should AI Regulation Look Like?
        01:04:33 The Future of AI-Powered Entrepreneurship
        01:04:44 AI and Job Loss
        01:07:15 The Skills Young People Should Invest In
        01:11:30 What Does Success Look Like For You?

  • > Not feeling like 1 hour of my Sunday is worth listening to this, do anyone have the non-clickbait answers to the two "previews" mentioned in the description?

    I know HN is built around mostly not reading the articles linked but how about you click on the link and surprise, there is already exactly another link providing what you're asking for.

So firing a grifter means it would kill the company? Doesn't that mean the company is grifting? If no one else can possibly lead the supposedly the most important company, with billions/trillions (?) of so called value, do you have a good company and product?

Or do I forget that this guy sleeps with an Ayn Rand doll tucked under his arms?

  • ChatGPT or CoPilot were awesome products at the time. I do not use them anymore these days. But to me it felt never like i was abused. And Investment into companies is what it is, a risk. But the results remain forever, whoever wins.

  • Not a fan of Altman but the devil you know is a powerful argument. If you believe a CEO/Founder to be a grifter-position at its core (fake it till you make it etc etc), retaining the best grifter you can find is the optimal play.