murakamiiq84

Same, but some animals are more equal than others. I'd personally trust one of those Japanese companies that lasted hundreds of years more on questions like stability and carefulness, compared to a flash-in-the-pan SV startup that's been around since 2016.

And also how a board of academics + Adam + Ilya of a nonprofit ostensibly optimizing for "humanity" might not consider someone like that the best fit for representing the nonprofit, especially if he was gaslighting them and treating them like NPCs.

Right, it depends on the details of his firing, like was it more about Altman being spread too thin or if it was more about ethics.

My story: Maybe they had lofty goals, maybe not, but it sounded like the whole thing was instigated by Altman trying to fire Toner (one of the board members) over a silly pretext of her coauthoring a paper that nobody read that was very mildly negative about OpenAI, during her day job. https://www.nytimes.com/2023/11/21/technology/openai-altman-...

And then presumably the other board members read the writing on the wall (especially seeing how 3 other board members mysteriously resigned, including Hoffman https://www.semafor.com/article/11/19/2023/reid-hoffman-was-...), and realized that if Altman can kick out Toner under such flimsy pretexts, they'd be out too.

So they allied with Helen to countercoup Greg/Sam.

I think the anti-board perspective is that this is all shallow bickering over a 90B company. The pro-board perspective is that the whole point of the board was to serve as a check on the CEO, so if the CEO could easily appoint only loyalists, then the board is a useless rubber stamp that lends unfair legitimacy to OpenAI's regulatory capture efforts.

I think it was only a competitor app after GPTs came out. A conspiracy theorist might say that Altman wanted to get him off the board and engineered GPTs as a pretext first, in the same way that he used some random paper coauthored by Toner that nobody read to kick Toner out.

I think it's important to keep in mind that BOTH Altman and the board maneuvered to threaten to destroy OpenAI.

If Altman was silent and/or said something like "people take some time off for Thanksgiving, in a week calmer minds will prevail" while negotiating behind the scenes, OpenAI would look a lot less dire in the last few days. Instead he launched a public pressure campaign, likely pressured Mira, got Satya to make some fake commitments, got Greg Bockman's wife to emotionally pressure Ilya, etc.

Masterful chess, clearly. But playing people like pieces nonetheless.

Actually I think Bill would be a pretty good candidate. Smart, mature, good at first principles reasoning, deeply understands both the tech world and the nonprofit world, is a tech person who's not socially networked with the existing SF VCs, and (if the vague unsubstantiated rumors about Sam are correct) is one of the few people left with enough social cachet to knock Sam down a peg or two.

I couldn't find any source on her investing in any AI companies. If it's true (and not hidden), I'm really surprised that major news publications aren't covering it.

Wait what? She invested in a competitor? Do you have a source?

I guess Metz would have a lot of sympathy for someone drawing the heat of the entire internet, and even powerful VCs, by doing something that many people think is destructive "on principle".

Random fanfiction: it's also possible that it wasn't actually a 3-3 split but more like a 2-2 split with 2 people -- likely Adam and Ilya, though I guess Adam and Tasha is also possible -- trying to play nice and not obviously "take sides." And then eventually Sam thought he won Adam and Ilya's loyalty re: firing Helen but slipped up (maybe Adam was salty about Poe and Ilya was uncomfortable with him "being less than candid" about something Ilya care about. Or maybe they were more principled than Sam thought).

And then to Adam and Ilya, normally something like "you should've warned me about GPTs bro" or "hey remember that compute you promised me? Can I prettyplease have it back?" are stuff that they are willing to talk it out with their good friend Sam. But Sam overplayed his hand: they realized that if Sam was willing to force out Helen under such flimsy pretexts then maybe they're next, GoT style[1]. So they had a change of heart, warned Tasha and Helen, and Helen persuaded them to countercoup.

[1] Reid Hoffman was allegedly forced out before, so there's precedent. And of course Musk too. https://www.semafor.com/article/11/19/2023/reid-hoffman-was-...

I read it as a commitment to keep this costly signal, beyond just words. If an amoral oil company wants to keep employees/customers who care about the environment, they might both say that they care about the environment AND do costly signals that indicate that they care about the environment (like replacing plastic straws with inferior paper straws, even if it's annoying and costs money). This is different from the company actually caring about the environment. Maybe actually caring involves taking actions that matter more than paper straws. Which again is different from being good for the environment, overall.

I might be reading into the literal words too much though. I don't have a sense of how messages like that is read in political science academia and DC (the primary target audience).

Is this even true for for-profit companies? Like if a professor is on the board of a for-profit (which I think is pretty common for deeptech? Maybe companies in general too? https://onlinelibrary.wiley.com/doi/abs/10.1111/fima.12069#:....), is he/she banned from making a technical point about how a competitor's product or best practices is occasionally superior to the company's product?

She has a legally binding charter (well, assuming OpenAI's charter actually holds up in court, which we may end up seeing as the next logical step in the drama) to use board powers to further the non-profit mission. She does not have a legally binding obligation to commit crimes.

Yeah it sounded like 3-3 gridlock until someone flipped, either it was Sam/Greg/Ilya against Adam/Helen/Tasha or it was Sam/Greg/Adam against Ilya/Helen/Tasha.

Note that the quote about Anthropic is about Anthropic's desire to be perceived as a company that values safety, not a direct claim that Anthropic actually is safe, or even that it desires to value safety.

I think this is heavily editoralized. if you look at the 3 pages in question that the quotes are pulled from (28-30 in doc, 29-31 in pdf), they appear to be given as examples in pretty boring academic discussions explicating the theories of costly signaling in the context of AI.It also has lines like:

"The system card provides evidence of several kinds of costs that OpenAI was willing to bear in order to release GPT-4 safely.These include the time and financial cost..."

"Returning to our framework of costly signals, OpenAI’s decision to create and publish the GPT4 system card could be considered an example of tying hands as well as reducible costs. By publishing such a thorough, frank assessment of its model’s shortcomings, OpenAI has to some extent tied its own hands—creating an expectation that the company will produce and publish similar risk assessments for major new releases in the future. OpenAI also paid a price ..."

"While the system card itself has been well received among researchers interested in understanding GPT-4’s risk profile, it appears to have been less successful as a broader signal of OpenAI’s commitment to safety"

And the conclusion:

"Yet where OpenAI’s attempt at signaling may have been drowned out by other, even more conspicuous actions taken by the company, Anthropic’s signal may have simply failed to cut through the noise. By burying the explanation of Claude’s delayed release in the middle of a long, detailed document posted to the company’s website, Anthropic appears to have ensured that this signal of its intentions around AI safety has gone largely unnoticed. Taken together, these two case studies therefore provide further evidence that signaling around AI may be even more complex than signaling in previous eras."

https://twitter.com/willknight/status/1726793735143621058 This guy claims he has "reason" to believe it's a "complete fabrication."

No idea whether to trust this guy or his source more than the original article though.

Apparently Hoffman was kicked out by Sam, not just Musk: https://www.semafor.com/article/11/19/2023/reid-hoffman-was-...

Maybe the remaining board members could see the writing on the wall and wanted to save their own seats (or maybe he did move to coup them first and they jumped faster).

Either way, they got outplayed.

Unfortunately lawyers almost always tell you to be quiet, even when you should be talking. So in this case listening to legal advice might have screwed them over, ultimately.

Might be more for PR/regulatory capture/SF cause du jour reasons than the "prepare for later versions that might start killing people, or assist terrorists" reasons.

Like one version of the story you could tell is that the safety people invented RLHF as in a chain of steps eventual AGI safety, but corporate wanted to use it as a cheaper content filter for existing models.

Yeah even though I agree with the COI being a large concern and potentially a justifiable reason for firing him (especially if coupled with other instances of him withholding key information from the board), it's clear that the non-profit board might have emotional reasons aside from the common good.

Even just the "you keep lying to us, bro" emotional motivation might be enough (not necessarily rational) motivation to fire; despite the person being otherwise qualified for the job, loved by employees and investors, etc.

The claim is that he's starting different AI hardware companies on the side, not that he's doing it under OpenAI's umbrella. It's more like if Sundar tried to get funding for some side hustles while talking with potential customers or contractors of Google, without disclosing to Google's board ahead of time.

Yeah something can be potentially net good for a company but still be a COI. If I'm in charge of awarding military contracts and I give it to a armaments company my son is a VP of, and I didn't disclose to anybody that my son was a VP there, this is a clear COI even if I later argue (and people agree with me!) that my son's armaments company is best suited for the business.

At the very minimum I should not be the one to make that call.

Come on, I think it's really obvious that the largest AI nonprofit in the world might be interested in chips, to a much larger degree than a typical nonprofit would be interested in cafeteria food.

I think the analogy is more like someone on the board + Executive Director of your wildlife preservation nonprofit buying up land with potentially endangered animals on the side, with the (presumed) intention of selling the land back to the nonprofit.

Clearly a COI even if it's net good for the animals.

I thought non-competes are not legally enforceable in the state of California anyway.