No "malfeasance" behind Sam Altman's firing, OpenAI memo says
axios.com
axios.com
Even if I agree with Sutskever's general position, I think his actions here were colossally stupid. He's basically managed to piss off a ton of people, and I have no doubt lots of employees will just shift to whatever Altman and Brockman end up doing, or else there will otherwise be some huge splintering of OpenAI talent.
Like how much damage do they have to do in the name of efficiency, profit, “the future” before the “benefit to society” meme dies?
Although now Google is the undisputed leader, given that Microsoft does not have the internal resources to maintain this thing.
I’d expect a tough challenge ahead as msft/other stakeholders try to put guardrails in while dealing with the fact that whoever’s bright idea this coup was - is now in charge.
What a mess.
Pissing off all your stakeholders with a snap decision isn't great governance. Particularly egregious circumstances might make it necessary.
Finding something wrong might require action, but a board's responsibility is also to make the corrective action in a responsible manner that is in the organization's interests.
but the trust of the board completely braking down doesn't mean that e.g. MS doesn't trust or want him anymore as in the end the interest the board have to defend and the ones which stake holder like MS have might be in conflict
so even if it has negative consequences if it must be done it might be easier to get it done fast and then resolve the mess instead of giving other a chance to turn it into an even bigger mess
That very well may have happened here. It is important to consider that this is effectively an HR matter and that the results of employee investigations cannot be made public.
Here, there's been a statement that there's no malfeasance, too.
Maybe Ilya doesn't want to be at the forefront of AI building I guess. He probably thinks he can lead it better by not being at the table.
Or who knows.
So depending on the exact non-public reasons for the communication breaking down it might not be a disaster but the hybrid doing exactly what it was intended to do.
Now the real interesting question is how much MS must have been afraid to be technologically left behind to knowingly enter such a deal with such a huge investment?
To my eyes, and to many others, it appears the board is power-tripping. Also, OpenAI's unique structure means they have no checks on the board that an investor class would normally give.
It's difficult to envision a thriving future for OpenAI under these circumstances or a world in which they are a significant factor in AI policy. Their early market lead, still a considerable advantage, will not be sufficient to weather the storm brewed by the board's own actions. Losing their market leverage would mean ceding the influential role in shaping AI's future, potentially leaving the giants such as Microsoft, Google, Amazon, and others assuming control and steering the future course of AI, but without the creed to make AI "good."
It's a non-profit board, yea?
Are you expecting them to reach out to Microsoft and ask "we're thinking of firing our CEO. How does that grab you?"
I don't think this is the type of thing you can really do by consulting outside groups.
Note that the non-profit doesn't have shareholders, but the effect is much the same because the two are linked closely together and you can't affect the non-profit in such a drastic way without also affecting the for-profit arm. So they just pointed the double barrelled shotgun at their feet and pulled the trigger, twice. It will take a massive effort to restore trust in the brand, both on the commercial side and on the side of the people working there.
For a normal board, this stakeholder check would have been a simple check with internal leadership on positioning.
Really, most organizations try to maintain at least some facade of predictability and transparency.
The OpenAI board should have known not to piss off Satya.
"AI safety" is a euphemism for "aligns with our interests." "Roko's basilisk" is nonsense science fiction babble.
Or people who are mentally deficient and actually believe in rokos basilisk and its ilk.
It is a 501(c)(3) even if they had all of the board members Open AI, Inc can not be run for MSFT benefit or for profit. A sympathetic board member might have tipped MSFT off about the firing ahead of time though.
Frankly, as it should. If you are going to take $10B from another organization and you nuke the CEO during market hours and then later it turns out that it was just an ordinary disagreement, then you should answer for that action.
On a Friday, no less. This is actually the kind of thing that can sink MSFT's apparent "first mover" advantage in the eyes of big money. OpenAI is now fractured, and whatever work they were going to be doing on tech is now going to be devoted to figuring out WTF to do with this mess the board just created all by themselves for no apparent reason.
The board has the power to dissolve itself, I am sure there(edit typo) are legal mechanisms for illegal acts, I think the IRS can revoke 501(c)(3) status but not dissolve, not sure if there are other mechanisms.
OpenAI might just show the world how the naive ideal we are not driven by profit but spending billions every year is never a good idea.
Those board has no stake in the company, on the other side means their have no accountability either. It is like they remote controlling the titanic, and who cares it crashes, because they are not on it.
Regardless, they made very stupid decisions now sabotaging interests and the company itself, brutally beheaded itself.
By firing the guy who founded Loopt?
I’m genuinely mystified, where does this theory come from. What makes this guy essential?
He elected himself to be the face of OpenAI
That is why he is now essential. He is the hinge of trust, and people believed his vision of OpenAI, as a company.
Ilya might be the science guy, but he 0 track record to run a company. And now he executes this badly planned coup, how would people trust him not to play such shticks in the future?
Arguing about semantics is useless
The board in a real way decides what the interests of the company are since they are the majority owner. The interests they set however may or may not bring about the results they want though.
Assuming such situation happens in is noticed there is no reason to consult because it wouldn't change anything and just make things more complicated as the only action which can be taken is to fire the manipulator in power asap.
Especially some recent actions where not so much in line with the mission the board is supposed to protect push for but very much in line with giving especially Microsoft a additional competitive advantage (i.e. additional to what Microsoft payed for).
And I have seen many comments like but MS and other stake holder should have been part of the decisions and similar. But here is the think, with the specific structure of OpenAI MS and other knew from the get to go that it's a high risk investment where even through they in vest a ton of money they won't have _anything_ to say at all when it comes to decision done by the outer non-profit which is implicitly the majority owner of the for-profit OpenAI.
It's so strange to hear this regarding a CEO, a single person, effectively going rogue against a company's charter, but then layoffs of tens, hundreds, and thousands of workers, on the basis of zero fault of their own, by these CEOs and boards are considered good governance.
The rank-and-file serfs don't matter, they're expendable and interchangeable. When Dear Leadership is shaken up, it's newsworthy. When ten thousand labourers are let go, it's just a statistic.
This is EXACTLY what late stage "political" companies are like. There is no rational way to decide who gets the window office vs. the internal office. Everything is "political"
Pointing out that modern capitalism is comprised of actors that have large structural similarities to failed communist government seems like more valuable comment than your shallow and dismissive response.
I'll try to expand on my previous comment without using the "C" word itself, prior programming seems to have triggered an emotional reaction there.
Here are some qualities that many (if not most) corporations seem to share:
- centralized power
- collective goals (Team A has an 18 month plan to deliver V18)
- top-down approach
- bureaucratic hierarchy
- collectivist ideology (down to "we're a family, we're all in it together, do it for the company")
- controlled expression (approved list of questions for the all-hands meeting, anyone?)
- resistance to change
- top-down reforms (vc / top brass reorgs)
- limited transparency
Does that sound like traits characteristic of capitalism, or more like spooky ivan papa bear, sickle and hammer stuff?
Where the similarities end is around ownership. In an (idealized) communist system, people own the means of production; in a corporate setting, everything is owned by the organization. Of course this is a key, essential, and critical difference, and it goes all the way back to maybe the 1600's when chartered corporations first appeared; but the roots of that reach even earlier, to the aristocracy disempowered by the rise of the merchant class in the middle ages.
The idea of parasitic Majority Shareholders siphoning off any benefits created by the collective efforts of individually irrelevant and easily replaceable Individual Contributors has been with us ever since. That the corporate structures would adopt the social organization aspects of a system that was created in direct response to their relentless exploitation, well, that's just irony of the highest caliber.
Yes, it is good governance to cut unproductive jobs.
There is no problem here then, is there
It definitely used to be considered as such. There has been a cultural shift over the last 40 years.
That usually means making money, but it's not a certainty that's what the primary purpose will be.
If cigarette companies could cause 5% less cancer by selling 5% fewer packs of cigarettes (resulting in 5% lower profits) do you believe they would do it?
You might say that companies have strayed from the Platonic ideal of a corporation and that cigarette companies have strayed from their “purpose” but the reality is that purely prosocial companies have never been the norm.
You might also suggest that the very fact that people are purchasing cigarettes demonstrates that they are providing “valuable” goods/services but this is to water down the definition of value to the point of meaninglessness.
The truth is that a corporation is a sort of alien entity that almost entirely operates to maximize profits. I say “almost” entirely because it is possible in extreme cases for their leadership to face personal responsibility, but it’s pretty rare.
http://www.antipope.org/charlie/blog-static/2010/12/invaders...
I think you mean "by providing anything anyone will pay for" in this universe you describe. There are a ton of companies that provide zero or even negative valuable services to society.
And what you describe is unbridled capitalism. That's not what most people consider viable. A company is not a logical entity that satisfies "if company, then capital at all costs", although there are the few (the super wealthy) that would like it to be that way. A company is part of a complex system and thus it needs to satisfy a wide variety of conditions.
And note that my comment is reacting to the idea that firing the CEO is some sort of crime against humanity.
If I am Vanguard, I am on the phone right now with Satya telling him to get control of this shitshow of a BOD before it nukes my positions.
Professional boards of directors are supposed to be attentive to the risks to the organization as a whole, regardless of whatever weird ass governance structure exists.
Blowing out your CEO on a Friday afternoon and then accusing him of lying is not the way to do this, and they are likely to find out why come Monday morning at about 9:30AM Eastern.
If Satya is as pissed as it sounds like he is, and if Ilya et al double down on their madness, Microsoft could back out and shut that all off tomorrow. OpenAI would be dead in the water, with no running product and no path to recovery.
This needs to shake out. Sam Altman was clearly not the best fit anymore. Thats the moral of the story.
Why are executives special? They aren’t. It’s a damn myth
If it was a normal boardroom difference of opinion, how would it have gone down? You think there would have been public along the process?
From an outsiders perspective I think both situations would have looked the same; we have no idea how long the process went on. If the board wanted to get rid of Altman but couldn’t convince him to resign, the announcement was always going to look like this.
And if Sutskever's beef was that OpenAI was going too fast at the expense of responsible AI development, he just completely screwed himself, as basically everyone who wants to move fast now will leave OpenAI and join the companies that do move fast. This is basically exactly what happened to Google Brain going too slowly so everyone leaving to join OpenAI and others. At least OpenAI was previously in the position of being able to guide responsible AI development, now companies will probably all go by the "drive fast or die" ethos.
the thing is there is (legally) a huge gap between lying and "not being candid"
in the later case you can have a lot of subtle manipulative formulations, phrasing misleading emphasizing and de-emphasizing of parts misleading representation of what certain decisions imply, intentional not correcting when you notice that someone did misunderstand something and it will lead to them supporting decisions of you they would normally not have supported etc.
all of this are legal actions (and in-actions; at least if subtle done) which still can get you fired for "breakdown of communication and/or trust" in pretty much any job, especially a position like CEO
outright lying on the other hand likely can entail a lot of legal trouble in many ways
so them saying that he "wasn't candid and there was a break down in communication" then it likely means that and a break down in trust but not outright lying (except if they want to downplay the situation, but why should they)
1. The OpenAI board certainly gave the impression that something egregious had occurred (just look at all the speculation on HN and elsewhere). I have literally never seen a separation announcement that was that pointed where there wasn't significant malfeasance. In the US, basically anyone can sue anyone else for anything. It would be quite easy to argue in court that the board gave the impression to a reasonable observer that Altman lied over something important - Kara Swisher tweeted that basically the board is screwed unless they have an extremely unambiguous smoking gun.
2. On a practical level, the legalese really doesn't matter. Then general SV opinion already seems to be that the board did Altman and Brockman dirty, it's already been reported that Altman is starting a new company, and I have no doubt heaps and heaps of OpenAI employees will jump ship to this new venture, as many high level employees have already announced their departure.
but there is still a huge different between something which is done subtle enough to not realistically be able to be pinned down by any court but entails a fundamental break of trust (which is egregious) and what legally could count ad fraudulent actions and similar (and is eregious and can additional lead to legal cases being brought against you)
through AFIK the general opinion I have heard outside of SV by people working with AI is that Altman is the one who had been acting "dishonest, manipulative" in a way which endangers open ai (not OpenAI but AI which is open) in recent times
and lets be honest SV isn't really known for caring about morals and bad consequences their tech can have, but the non-profit which made the decision has roughly committed to upholding such standards
> that basically the board is screwed unless
I highly doubt this, probably his comments (some of which where very misleading and manipulative i.e. not candid at all AFIK) in front of congress in context with the goals of the (non profit parent of) OpenAI which fired him has committed to would be enough reason to fire him. If we used similar manipulative speech also in board meetings this would be a pretty clear open and shut case in favor of the board which also would thoroughly destroy his reputation (iff that is the case and it's brought in front of court) so from an outsider perspective this would be a very interesting case.
Can you please add a link to that post? It’s been so frustrating trying to follow this story on Twitter. Kara’s feed has a single pinned post with no indication from the UI about where the rest of her posts are.
The typical approach is to let the CEO know they are being replaced, line up a successor (even if temporary) and arrange an orderly transition (message key partners, customers, etc). This avoids the narrative getting away from you ("omg, the product is just mechanical turk, about to go bankrupt, have some serious accounting issues, CEO did crimes, etc"). These bad vibes have a way of persisting and can cause a loss of trust with key stakeholders (previously overlooked issues now get revisited with a more jaundiced eye)
The way the action was taken (suddenly and with accusations) gives the appearance of personal animos rather than measured, sober judgement. Of course, there might still be good reason for taking this approach but so far, it is not looking good.
But I've seen "on the spot firings" somewhat with Open Source-based, high profile startups. There's always a point in a FOSS startup's life where Investors/Board/Founders/SrStaff have core communications break down in one day (aka yelling matches) and people leave immediately--typically from debate on GPLv3, legal violations w/GPLv3, profit strategy, and 'community' support. Makes me now wonder what was the orig exit strategy for OpenAI.
"I'm rubber, you're glue. What bounces off me, sticks to you"
The fact they had to make a release saying there was no malfeasance means there was malfeasance on their part.
The board are looking very bad here imo. Either it was serious misconduct from Sam and that should be at least leaked, or they staged a coup.
anyone telling me this guy's the only one who can achieve the nonprofits mission seems to be high on occult business logic
QRD? Haven't caught this detail yet.
That still may be good life advice in general (even if it wasn't for Sam in this case) but what I really don't get is the fact that OpenAI's board governance was structured in a way such that this was even possible.
I also don't understand what is to be gained from the perspective of the remaining senior leaders at the company. This is a tremendously momentum-killing event. I cannot think of a single facet of their day-to-day operations, product roadmap, competitive position, etc. that would be improved by this decision.
Yesterday, when this was announced, I was bracing myself for some truly awful news about something that Sam had done in his personal life that was about to be divulged, since that is the only possible rational reason for the board to make the decision it did.
What am I missing? It's all so strange.
This is true in the abstract but it is not common to see the narrowest possible majority on a six-person board oust the co-founder/CEO (also a board member) while excluding the board chair (who was also the company's President and co-founder) from the process.
> So it would be odd if it was structured in a way that this wasn’t possible.
You are wrong here. Boards are usually structured so that what just happened at OpenAI is not possible. The exclusion of the board chair from the process is especially egregious and the involvement of board chairs in decisions of this magnitude is often mandated in documents of incorporation.
Matt Levine is going to have a field day with this, LOL.
Yet another reason why it's a bad idea for a CEO to also be on the board. And probably also evidence that it's a bad idea for conflating the jobs of CEO and chief spokesperson.
Dropping this link to another comment I made because people often forget what "fiduciary" actually means.
You see it all the time in legal dramatizations where the lawyer is telling the client that they can get more money, but the client tells the lawyer they just want the legal process to stop. The client gets their way because the lawyer has a fiduciary duty to them.
Edit to add: I'm curious what word "Revlon" was supposed to be.
I'd guess "revenue", but who knows. Autocorrect is a disaster for clear communication these days. It's strange to me that it's as bad as it is. Is there a patent on using context to decide which word is appropriate? If not, I can't figure out why it isn't done more commonly. It seems certain the poster didn't type "Revlon" with a capital R, and it also seems certain that any statistical model wouldn't think "Revlon" would be the most likely word here.
At least that’s a layman’s guess
Uh.......
You might want to spend a second researching what "Revlon" is before speculating like this. I'm 100% positive that they actually typed Revlon with a capital R, because I know what Revlon is, and it's completely topical and contextual. It's very well known in certain circles, much like Chevron, Howey, and Brandenburg.
I surmise that "Revlon" stands for "an entity worth billions". IOW having a component with large monetary value to manage.
Board members are the fiduciaries who steer the organization towards a
sustainable future by adopting sound, ethical, and legal governance and
financial management policies, as well as by making sure the nonprofit has
adequate resources to advance its mission.
[0] https://www.councilofnonprofits.org/running-nonprofit/govern...
In the case of nonprofits, the fiduciary duties of board members pertain to keeping the organization financially healthy as opposed to representing the interest of shareholders. But you already knew this, right?
If the dispute is over whether they should be a pure research nonprofit or not, then operations, product roadmaps and competitive positions are things they would not even want to begin with. It was only four of them that booted the other two out, and of those at least two seem to be professional non-profit board sitters. The transition of OpenAI into an actual product company doing something useful for other people is incredibly recent, it's barely a year old. It may be that they simply felt that offering products isn't what OpenAI is meant to be doing.
Alternatively it could be some blowup related to doomerism, although the memo saying it's not related to "safety practices" may rule that out.
The safety comment seemed more about reassuring people there wasn’t, like, a data breach or “escaped model” or something crazy. In other words, clarifying that the disagreement was about the future, not the present.
Think this is good advice for day-to-day interactions for everyone but it will always run into limitations. Especially for an edge case like OpenAI
I'm unbiased and not super familiar with this situation either but the quote above makes me think of "Never attribute to malice that which is adequately explained by stupidity."
imo "never" ignores cases where malice can be proven, so i don't like it as a saying
Sufficiently advanced stupidity is indistinguishable from malice.
This is a genuine question, can anyone explain to me what the actual value of San Altman is?
I don’t know anything about the guy except the core public outline of his career but I find his timeline confusing.
He founded Loopt which by all accounts was unremarkable.
Then he ran YC during a period where at best you can say it didn’t get wrecked, though it seems to have lost focus and influence. He left YC in a situation that seems similarly abrupt minus the PR drama which was kept completely under wraps.
Then OpenAI which is a genuinely impressive organization but by now he’s been in personal conflict with basically everyone involved and doesn’t seem to have been a driver of the actual technology.
Other than that there’s world coin which is a total grift at best and dystopian at worst. And stuff like Bohemian Grove or whatever.
With all that said I don’t have a personal opinion just observing the broad strokes and trying to figure out how exactly he ended up some kind of kingpin.
Weird as it may seem: that's not all that important. What is important is that the rest of the world sees OpenAI as a mature and stable organization and that image has now been seriously damaged. Even if they wanted to fire Sam there must have been a dozen ways in which that could have been done more effectively and without this much damage to the OpenAI brand. It's almost comical, they were on track to outrun Google and Facebook combined and now they're on track to be an also-ran. You can expect a large number of resignations because no matter what you think of Sam Altman you want to work for a company that is mature and stable, not one where execs are replaced on a moments notice on a clear and sunny day without a direct and clear cause.
But what if they don’t want that. Those are rapacious and malevolent organizations.
Regardless, even if that is the goal what’s the evidence Altman is the right guy for that? It’s not like he’s done that before. The relationship between his stature and his track record is the part that’s confusing me.
And if the board wanted the company to be a pure research organization they should have acted much, much earlier, and the company probably should not have hired Sam Altman in the first place but someone like Geoffrey Hinton or another respected name from academia, provided they would be available in the first place.
I mean sure. But if true that’s hardly an argument for why they shouldn’t do it now as the stakes grow increasingly higher.
By whom?
Either way, normally you'd have a pre-written resignation letter drafted where the only thing missing is their signature, and you'd confront them in a meeting: resign voluntarily or we'll put your continued presence up for a vote at the next board meeting. Of course the board could try to eternally vote itself back in but that usually doesn't work because a board has to be able to serve in its oversight role and one part of that is that the board has to have broad support, both legally and from within the organization. For instance: the board might no longer be able to find a CEO that is acceptable to the rest of the C-level. That would be a major problem.
Corporate governance is hard, non-profits have a bunch more twists but in the end nobody's position is carved in stone. Note that even non-profits have bylaws and these usually detail clearly how board members are to be proposed and what the procedure is to get them to become established as well as how they can be removed. If it can be proven that a board member has acted against the interests of the legal entity they represent then they usually can be removed even easier because that's a clear conflict with the statues of the non-profit, then there are potential conflicts of interest (for instance sitting on the board of another entity that has goals that are not compatible with those of the non-profit).
Who outranks them? I don’t get the premise. You’ve implied twice that they could lose their seats over it.
How? By what mechanism? Challenged in court or presented with a resignation letter by whom?
Seems to me they’re the top of the org chart. They’re only accountable in the sense that employees can quit or they can run out of money.
But they happen to have direct control of one of the most valuable pieces of technology ever created so what exactly are you suggesting will happen.
Depending on the bylaws it could be stakeholder: donors, beneficiaries or the employees of the non-profit in some organized form. All of these could petition the court if they feel that the non-profit wasn't governed properly. Note that it isn't known if the board acted unanimously (likely it it didn't) and what the grounds were. That will make a big difference to any outcome.
Non profits that lose their donors (especially if the money was pledged but not yet committed) usually don't live long so the board has some incentive to play ball.
But you said: “I fully expect the board to be replaced”
Ok, sure. By who? When?
I don’t see any reason to think that unless there’s a really genuinely large scale staff revolt.
That's already underway, see other news about resignations at OpenAI, also donors have a very strong play to make.
Donors could simply withhold the next tranche, but possibly they can funnel enough money upwards from the for-profit to compensate for that.
Even so I would expect the other shareholders in the for-profit to be furious, especially those that liked Altman. So they're going to have to do some explaining because right now this does not deserve the beauty prize, to put it mildly.
Isn't that exactly why you should outrun them?
Can you quantify and specify the brand damage? I only see some possible damage to current and possible future employee morale (in that some of them are quitting and others may be less inclined to take a job there). Do you see this as seriously affecting relationships with companies such as Microsoft? With end-users?
Absolutely. If there isn't a very good reason why they did this in this way then I fully expect the board to be replaced.
And yes, there are plenty of precedents of board members being recalled, most of them aren't stupid enough to fight it, especially not non-profit board members, who are supposed to do this all for glory and sunshine. Typically they are presented with a pre-written one pager with date and place already filled in and all they do is sign it or they'll find their position to be up for a vote. And of course, the ranks could close around this decision but that might make some extremely powerful enemies. Think 'Microsoft', Sam Altman, Greg Brockman, Reid Hoffman, YC, Peter Thiel, Elon Musk, Amazon Web Services (AWS), and Infosys.
The combined onslaught of that would annihilate the board. So unless they made very sure they had plenty of backing on this decision they set themselves up for a very difficult situation.
They will remain aboard for as long as it takes to find a suitable replacement.
Why the dominant narrative isn’t “they’re probably disagreeing over the AGI status of a GPT5 ensemble because it affects their relationship w/ Microsoft” I have NO idea, and I’m trying to keep my head down and not let it all drive me crazy with anxiety…
Much love to the fellow hackers out there. If I’m anywhere close to right-ish, then I’m looking forward to exploring the post-SV/VC world with you.
It is like saying Jobs doesn’t make iPhone, so he is not useful either. I don’t like the Sam is the new Jobs narrative but here it fits.
OpenAI is no small company now, it needs a head to pull itself in a single string.
In real world, power or trust goes to the person people who actually trusts. A manager isn’t a manager just by calling oneself manager. Without trust, people will find or form other decision centers around them. The result is execution became watered down or outright not being executed.
That is how I see the value those high level executives really offer. They are the center of trust and accountability. If people trust them, people beneath him will make sure things happen fast and hold themselves accountable. On the other hand, you just end up with an ineffective company
He’s run exactly two impressive organizations, wasn’t the rainmaker at either, and left both somewhat abruptly. I don’t quite get it.
But aw shucks it’s a terrible fit actually, because this company has a non-profit, idealistic mission. In the face of unprecedented high demand, Sam Altman is like fuck the nonprofit hippie shit, let’s launch this thing into outer space. Let’s build fast and ship too many products too early, use investor money to give discounts and build moats, let’s go for regulatory capture! It’s the opportunity of a lifetime guys!
But the board is like nah, fuck that tiring ‘ship broken shit’ tech bro shit, let’s stick to our vision. This is too serious and important to treat it like a fucking Airbnb or Coinbase.
I think the reason so many people on HN can’t come to this obvious conclusion is they share the same sociopathic Silicon Valley worldview he does.
The "difference of opinion" is a fundamental difference in ideology and motivations, not to mention day to day operating differences.
Even if this is the case, and Jeremy Howard’s Twitter thread on this would lend credence to this version, OpenAI butchered the execution. If firing Altman really is about not letting profit ruin a very important advancement in technology, the board would have the moral upper hand. And yet, their communication has been so poor that right now, they’re not exactly winning the PR battle.
If you’re gonna stage a coup for such a high profile company, you better know what you’re doing. So far, it’s been looking like amateur hour.
That they weren't concerned about day-to-day operations, employee morale, product roadmaps or any of that boring stuff. They were operating on the precarious calculus of their "friendships" and commitments.
They had to act fast. It is possible, if Altman or Brockman had more chance to talk to board people, they could have changed at least one member's mind.
What you’re missing is that the board (at least what remains) seems to disagree strongly with the direction in which the organization was headed — then why would they care to maintain momentum along the “wrong” roadmap? The changes seem (deliberately) driven by the intent to steer in a different direction and with slower momentum. Others might disagree with that vision, but the board feels that it aligns better with the founding charter.
Yeah that is clearly the case, but when you have a situation where four of the six board members execute a coup against the two most powerful and important board members (the CEO and the chair), the company looks like a shitshow and nobody is going to want to work there or partner with them, at least until they are able to present a credible narrative about what they did (which may never happen).
Totally get the fact that the four remaining board members didn't like where things were headed, but their actions yesterday created a tremendous amount of collateral damage that will massively impede OpenAI in its journey toward whatever new azimouth they choose.
Umm, if it's "one human, one vote" then the board was working exactly as designed.
Being "important" may not save your job. Been there, done that.
Board charters are often written so that this kind of thing doesn't happen - especially in the case of startups with small boards and charismatic founders.
Before this I wasn't interested in joining OpenAI despite having seen colleagues jump ship to them, but I now very much am. It works both ways.
That 501(c)3s like OpenAI are entities where the shared interests they are designed to serve are ideological, not generating profits to return to investors; OpenAI (the governing nonprofit) is in some respects more like a church than a commercial enterprise.
Who's the COO talking on behalf of when he says "we have had multiple conversations with the board"?
Who's full support does Mira (Ilya's choice) have? Why does it need saying?
The rapid escalation from "we seem to disagree on this" to "walk them off the premises" was standard practice at OpenAI, along with firing people on Friday, at noon.
Many people were "resigned" while they were actively discussing (or at least so they thought) any disagreements with their immediate management, or the higher echelons of OpenAI.
Coordinating this internally took more than a few days, and there must have been a few middle-managers involved in several meetings to coordinate this.
Details on Ilya's selection function on who to read in to this would inform a lot of his most intrinsic motivations for this.
It's a pity to see brilliant minds unravel under the pressure of their own invention . . . and so publicly.
If they just said philosophical differences or something that would have been sufficient.
Now that is over, and Google has taken the lead now that Microsoft crashed into the wall.
Was OpenAI just supposed to live on the good graces of MSFT?
Do you know how much GPT4 costs? Plus training for GPT5? Plus supporting all those alignment efforts which just hinder the utility?
The onion layers let the for profit subsidiary make some money and get taxed on it, but they are ultimately beholden to the nonprofit parent's wishes.
I guess OpenAI just expects to get a lot of donations / quasi-investments. Hell, Mozilla survived for two decades on Google's teat, and all they do is make a browser and act as a thinktank.
If OpenAI can keep up the momentum, there'll probably be funders lining up for new arrangements after the Microsoft one expires. Shrug.
But none of this explains wtf happened yesterday.
The question is how is this going to impact their API and the increasing demand for lower-cost, higher-performing models to be used by the masses? If they want to be like Google and keep everything behind closed doors for years that's their decision, but they need clear communication about it.
Lots of people were launching things that were basically gpt(3/4) + some rag + managing state and a prompt.
Openai made it so that someone who can just describe the problem can create a solid version of that. Without managing hosting or keys or anything.
My guess (and its a guess, not knowing anything but excerpts from the memo which we aren't even told who it was addressed to) is that its on behalf of the executive team of OpenAI Global LLC, which, as a reminder, is separated from the OpenAI nonprofit by one (by ownership) or two (by control) intermediary organizations (the holding company that is the direct parent of OpenAI Global LLC, and the OpenAI GP LLC which is the organization through which the nonprofit exercises control of the holding company.)
> Who's full support does Mira (Ilya's choice) have?
The rest of the OpenAI Global LLC executive team.
> Why does it need saying?
Fairly common thing to say after an externally-imposed change of leadership, especially one that is or has the appearance of being a radical shift in direction.
> The rapid escalation from "we seem to disagree on this" to "walk them off the premises" was standard practice at OpenAI, along with firing people on Friday, at noon.
> Many people were "resigned" while they were actively discussing (or at least so they thought) any disagreements with their immediate management, or the higher echelons of OpenAI.
Where does this come from? Its not from the article (either the quotes from the memo or other material in it.
> Coordinating this internally took more than a few days, and there must have been a few middle-managers involved in several meetings to coordinate this.
Same issue as above, plus this contradicts the information that has previously come out that the only person outside the participating board members who was told anything before the dismissed parties were informed was Mira, and even that wasn't days in advance.
I predict that this is the beginning of the end for OpenAI. The company won't collapse overnight or anything, but its best people will start leaving in January, then its average people will leave over the course of 2024, and there will be little left by 2025. They may have enough money to live on as a zombie company for several years after that, but everyone will know that it is no longer what it once was or could have been.
If the board ended up acting rashly I expect the next thing to happen is that the board itself is exiled.
They all have enough money to be able to fight any legal challenges. This is not your regular, poor nonprofit paying slave wages.
You always claim "precedent" - so please show me, when has the board of a nonprofit even close to as high profile as open AI ever been removed by external pressure? As far as I can tell, we're on entirely new territory here.
I for one am glad Ilya is prevailing here - sama is the typical SV sociopath pushing growth at any cost. We really don't need any more of those guys at a place this important.
It is going pretty much according to the few ways in which it could go: lawsuit with unknown outcome (most probably by MSFT or other minority shareholder), reversal, board (or selected board members) resigns. The future timeline in which nothing changes and the board holds fast is not one that I think is viable, assuming Altman didn't murder anybody in broad daylight.
And while I in principle agree with your character assessment (though I wouldn't put it in those terms, I'm not qualified to diagnose people to that degree) that still doesn't mean you can do what you can do in any way that you want to do it without consequences. The board forgot for a hot moment that they are first and foremost representatives of the stakeholders and govern accordingly. To oust one founder and cause another founder to walk you have to have grave reasons and solid support from all of the stakeholders.
So far I have not seen these so the decision can't stand as it is. It's entirely possible that those grave reasons exist, and even then this should have been handled with some tact, not the blunt side of the axe.
> Ego is the enemy of growth
https://twitter.com/ilyasut/status/1707752576077176907
The real mystery that we're all trying to solve in realtime is: who was the biggest ego here? Like a good game of Clue, everyone is still a suspect...
Sam's own related tweet: "curiosity powers progress, but ego often does the heavy lifting during the slogs"
My take is that while Sam's statement is not a popular thing to say or admit, it's more honest than Ilya's tweet, which sounds more like a grandiose ego trying to pretend it has no ego. I know which I prefer.
But on a cursory look it seems to me like the boards' intentions are the right ones, while Sam seemed intent on making the most money possible with no regard for any of the Org's founding principles. It's why the "Open" in the name had become a joke.
I had no hope that OpenAI would ever deliver on its original mission, because when has a commercial takeover ever been reversed?
This may or may not kill OpenAI's lead, but maybe at least it will bring back openness ...
For people saying they shouldn't have done the MSFT deals, how else would they have gotten the money to build anything at all considering how much GPUs cost? Their competitor Anthropic is doing the same raising money from all of big tech, made possibly only due to the ridiculous success of ChatGPT. For others saying Ilya is the key, Google had a big lead on AI researchers and the only reason Google is not the undisputed king is due to bad leadership and product development.
So yeah, to me, really backs up the narrative that the board and sama were in disagreement about some key things.
Now there is a good chance that the “true believers” in AGI have taken over who will want to focus on just trying to achieve that. There is a good chance that true AGI is decades or more away (if ever). Without a product to push, this pure research organization will produce a lot of really cool papers(maybe) but not much else. As the talent see that there are better economic prospects elsewhere, they will leave.
If it's not malfeasance about finance, business, safety or security practices and is merely a "breakdown in communications" then that rules out virtually everything that could justify such an abrupt firing. Sounds like there's no legal problems to worry about, leaving an incompetent board as the only obvious possibility left.
When I was younger I remember thinking non-profit orgs were awesome and hoping to work in that space one day. This + the Wikimedia Foundation funding scandal + the proliferation of "misinformation" NGOs has really turned me off the whole concept. Nonprofits seem to have more than their fair share of problematic governance. There's something about commercial trade that grounds people.
It's called shareholder lawsuits. I really do wonder about boards of non-public companies, though. I'd expect them to be, on balance, not much different from NGO boards or even HOA boards.
So let me frame it another way, how does research expect to be supported if it's not providing material value? Research doesn't exist in a vacuum, if you don't want to be in industry you don't have to, we have academia for that.
I'm cool with OpenAI becoming more like Google brain, but then its Microsoft calling the shots. Except wait the company structure is weird, so they're not really and after seeing how Ilya handled this I don't expect him to be able to stomach that relationship long term.
I'm not bullish on OpenAI right now.
Before that, as a pure research organization, capital requirements were much smaller and (probably) could have chugged along without a player like MSFT, or at least a much lighter partnership.
If OpenAI ends up like a Google Brain, that doesn't mean Microsoft calls the shots, because Microsoft doesn't own any of the IP or researchers. Microsoft just has a a license to something that is almost certainly contingent on funding, and as long as the research is good, others will be happy to take their place.
If true, Sam would have kept his job just by communicating more and better.
If true, the board wouldn’t have a lose CEO in the wild and talents now possibly trying to build a competitor…
https://www.levels.fyi/blog/openai-compensation.html
Only a few days ago there was a post here about them poaching researchers with “10M” equity compensation packages..
How?
> It’s pretty textbook right?
No? Offering a compensation package can't be tortious interference, tortious interference requires an existing contract, and action by a third party that (along with a whole bunch of other conditions) causes a breach of that contract.
1. an economic relationship existed between the plaintiff and a third party which contained a reasonably probable future economic benefit or advantage to plaintiff;
2. the defendant knew of the existence of the relationship and was aware or should have been aware that if it did not act with due care its actions would interfere with this relationship and cause plaintiff to lose in whole or in part the probable future economic benefit or advantage of the relationship;
3. the defendant was negligent;
4. and such negligence caused damage to plaintiff in that the relationship was actually interfered with or disrupted and plaintiff lost in whole or in part the economic benefits or advantage reasonably expected from the relationship
It’s pretty open and shut, no?
And you can't make that case with the later decision by the nonprofit board (even if one assumes that an entity that has complete control of a party can even count as a "third party" for tortious interference rather than just being a source of breach of contract if it actually induces a breach), because the OpenAI Global LLC operating agreement expressly sets this out as normal, so there is no reasonable expected benefit of the relationship that is being interfered with.
If OpenAI Global LLC offered its employees the PIU/PPU as compensation but did not disclose (or, a fortiori, actively concealed), you could make a case for fairly simple fraud. But tortious interference is just not a tort that works here.
Yes, a profit sharing claim in an LLC whose operating agreement explicitly says it will be operated for the charitable purpose of a controlling nonprofit and that profit will not be its guiding principle does not create a reasonable expectation that it will be managed for profit.
Either the employees have that information disclosed, and there is no expectation, or it is concealed by the subsidiary employing them, in which case there might be a reasonable expectation, but its due to fraud by the subsidiary, not interference by the parent.
A lawsuit based on what? The OpenAI Global LLC operating agreement has a prominent notice...
https://images.openai.com/blob/142770fb-3df2-45d9-9ee3-7aa06...
Live and learn I guess.
OpenAI is massively ahead of anyone else in the AI space, anything that shows problems in the organisation seems like it could set us back massively in the development of AI.
Several key people at OpenAI have also resigned, including Greg Brockman who was removed from the board but told he was too important so would remain in his other roles, he also tweeted not to worry bigger things are coming.
If Sam breaks away and does something else in this space, I can see others joining him and although competition is deathly, I feel like it is just going to fracture our AI developments, everyone is trying to catch up with OpenAI, any new company would be starting from scratch and be years away from anything useful while also leaving OpenAI a shell of it's former self.
Hopefully not the start of bigger issues at openAI.
In the very short term it’s disruptive. In the medium and long term it’s probably a wonderful thing to soften the ground a bit and let competition take hold.
Possibly followed by Murati, but also maybe her comments won't be that relevant if Sutskever is really as into shutting down product development as it seems.
Interestingly, the situation is fully reversible for now:
1. Majority of employees sign a strong letter condemning the board and calling for their resignation. 2. MSFT threatening to activate whatever protective clause they have 3. Other investors/donors threatening lawsuits. 4. Sustkever et all resign, new board appointed and Sam and Greg come back.
Things could be back to business as usual by Monday morning.
Given the disclaimer attached to the OpenAI Global LLC (the entity Microsoft is directly invested in) operating agreement that is also reposted on OpenAI's own description of its structure about how investment in the firm should be treated as a donation, and that the firm has no obligation to seek profit, and its primary goal is to serve the charitable mission of the nonprofit and that that will trump all other concerns including profit, I doubt very much that that is the case.
Now, if there was actual evidence that the board (which is, remember, the board of the nonprofit) was not making a bona fide effort to serve its understanding of the charitable purpose of the nonprofit, but instead serving some unrelated (especially profit-oriented) reason, that would perhaps be grounds of the lawsuit, but most of the concerns being expressed are exactly the opposite -- that the board is putting its view of the nonprofit's purpose ahead of commercial interests that people wish the firm would pursue. Which is exactly what the entire structure (and the laws governing charities) are set up to assure...
How in the world is losing Sutskever business as usual?
As in – without him – this LLM craze, as we know it and as it has manifested – would not exist today?
My understanding is that he is at the center of all of it.
NY Mag said that Sam Altman is the Oppenheimer, I think it's Ilya.
The research led to a technology with a simple use case: input / output.
The "product" is a chatbot. input / output.
There is no innovation in interface or product design.
All the magic – all the "product innovation" is in it creating coherent text output.
(OpenAI's UI is not really that great).
All attempts at "product" (as in serving a specific use case tailored to a specific business or commercial workflow) that OpenAI have led to date – i.e. plugins – have been a huge flop.
I am uncertain who is behind assistants interface design to make RAG so easy. It's usability is underpinned by the larger context window of GPT-4T.
All of the magic is in the underlying technology. And I think Ilya's at the core of it. Yes, Attention Is All We Need is the Google paper that invented the transformer.
Ilya also worked at Google and was a key part of the ecosystem, and I imagine had indirect (and most likely very direct) contributions.
He is one of the most cited computer scientists of all time.
Without Ilya, I don't think we'd have the hype right now.
Someone would have come to the same conclusion, I'm sure.
Trillion $ companies are not forces you should be meddling with.
Which feels like complete science fiction, but it comes closest to explaining why the non-profit board would move so quickly and disruptively.