9 ms·

I am still so confused by this whole saga. No one has explained how or why the entire board decided to do a coup, and then simply changed their minds a few days

by reso 3y ago

I am still so confused by this whole saga. No one has explained how or why the entire board decided to do a coup, and then simply changed their minds a few days later. I have to assume that the individuals involved are under a mix of social and legal pressure to not talk about the circumstances.

richbell 3y ago
The speculation seems to be that Sam was being duplicitous and trying to oust a board member he didn't like. Members of the board compared notes and realized he had misrepresented his conversations with other members in an attempt to build concensus. Throw in the rapid expansion of OpenAI (commercialization being in conflict with the board's vague mandate to do what's best for humanity) and the rumored (now confirmed) deal he was arranging with a company that he had a financial stake in, and they felt like they needed to remove him. However, by trying to do so swiftly and not allowing him a chance to retaliate they pissed off Microsoft and the employees. At that point, they were basically forced to reinstate him or the entire company would collapse — and if they do that, they can't also go on record clarifying why he's bad. * this is my vague recollection based on reading past discussions. I'm on my phone right now and unfortunately don't have any sources I can link, take this with a grain of salt. Edit: a few links https://news.ycombinator.com/item?id=38559770 https://news.ycombinator.com/item?id=38559770 https://news.ycombinator.com/item?id=38548404 https://news.ycombinator.com/item?id=38548404
- JumpCrisscross 3y ago
  Their communication strategy was also juvenile at best.
- ethbr1 3y ago
  Thanks for the citations. That was my guess... but only because it was the only scenario I could think of where the board being curiously and obviously intentionally vague about 'why' in the announcement, but still saying more than nothing, made sense.
pixelmonkey 3y ago
This recent TIME article lays out the saga pretty straightforwardly and makes it a bit less confusing. https://time.com/6342827/ceo-of-the-year-2023-sam-altman/ https://time.com/6342827/ceo-of-the-year-2023-sam-altman/ At least, that was how I felt after reading it. Basically, within the span of a year, OpenAI transformed from a research lab inside a non-profit that was pursuing a seemingly-Quixiotic dream of artificial general intelligence (AGI)... into one of the fastest-growing for-profit software companies of all time via its creation of the chat-based generative AI category (aka ChatGPT) and its consumer/enterprise SaaS and API offerings. The board -- or, at least, its 4 remaining non-CEO members -- thought that this was too much, too fast, and too soon, and that there was a narrow window of time where they could slow things down and refocus on the original non-profit mission. They also felt that Altman was a bit of a force of nature, had his own ideas about where OpenAI was going, and treated "board management" as one of his CEO skills to route around obstacles. Once a board loses trust of their CEO, unfortunately, there is usually only one blunt and powerful tool left: firing the CEO. And this happens pretty often. As the investor Jerry Neumann once put it, "Your board of directors is probably going to fire you."[1] Boards have very few ways to actually take action when they are worried about a company or institution; firing management is one of the few "course correction" actions they can take quickly. In OpenAI's case, if they had a for-profit board, that board would probably have been ecstatic with Altman and the company's progress. But this was not a for-profit board. It was a non-profit (mission-oriented) board meant to oversee the safe rollout of AGI. Those board members weren't sure the best way to do that was to become one of the world's largest for-profit software companies. I'd speculate that it was probably an emotional decision and the full implications were not entirely thought through until it was too late. I'd also speculate that this explains why Ilya Sutskever felt some immediate regret, because his goal wasn't to destroy OpenAI (or inspire an employee revolt) but to put its non-profit mission back into focus. I like to practice the principle of charity[2], and, in this case, I think the non-profit board was not acting maliciously but simply did not realize the knock-on effects of trying to replace a CEO when everything at the company seems to be "going right." I suspect Altman thought the best way to roll out AI was via iterative product development and fast revenue growth to finance the GPU demands, utilizing corporate partnerships (Microsoft), viral word-of-mouth marketing, and SaaS/API fees (ChatGPT). Running out of data center compute started to become a primary concern, so it wouldn't surprise me if safety took a backseat to this. Remember, all this growth happened in the span of a year. Perhaps Altman thought he was satisfying the safety concerns simply by talking to regulators, making iterative releases, and going on a speaking tour about it, but the board thought the only way to go safely was to go slower. I'm sure we'll learn more after some books are written about the episode. [1]: https://reactionwheel.net/2021/11/your-boards-of-directors-is-probably-going-to-fire-you.html https://reactionwheel.net/2021/11/your-boards-of-directors-i... [2]: https://en.wikipedia.org/wiki/Principle_of_charity https://en.wikipedia.org/wiki/Principle_of_charity
tivert 3y ago
> I am still so confused by this whole saga. No one has explained how or why the entire board decided to do a coup, and then simply changed their minds a few days later. I don't know how anyone can call the board's action a "coup." Calling it that seems to be a propagandistic abuse of the term. The board was in charge, and it's not a coup if it fires a subordinate (the CEO). If anything, the coup was getting the board ousted.
- reso 3y ago
  Sure, it was not my point to suggest it wasn’t legal, just that it was a bold action which has gone completely unexplained.
- darkerside 3y ago
  I think the coup is on the part of other employees who advocated for the board to fire Altman. No judgement in whether that was justified or not.
  - tivert 3y ago
    > I think the coup is on the part of other employees who advocated for the board to fire Altman. No judgement in whether that was justified or not. The GP was pretty clear that he thought "the entire board decided to do a coup," which does not fit that interpretation. But even the scenario you describe isn't something that can be properly described as a "coup." In that case, the employees are just appealing to a legitimate higher authority, which is a totally OK thing to do (e.g. it's not wrong to report a bribe-taking boss to the company ethics hotline). IMHO, a coup is where subordinates illegitimately usurp and depose the highest authority from below.
    - darkerside 3y ago
      I probably missed that in the GP comment. I don't think invoking a higher power necessarily makes it not a coup. To me, a coup is when you angle to relieve someone from office above you by any means. But reasonable people can disagree on that, I suppose.
bogomipz 3y ago
There was a bit of insight today from a board member on her perspective. See: https://archive.is/Sy3Xm https://archive.is/Sy3Xm
- Clubber 3y ago
  That certainly doesn't leave a good taste in my mouth.
- huytersd 3y ago
  Honestly who cares what Helen has to say. She didn’t belong on a board in the first place.
  - bogomipz 3y ago
    Maybe try to follow the thread where the OP states: >"No one has explained how or why the entire board decided to do a coup ..." To which I responded that there was in fact a very recent and relevant update to this. Your opinion on whether she belonged on the board is completely irrelevant to this fact and also childish.
nicce 3y ago
> No one has explained how or why the entire board decided to do a coup, and then simply changed their minds a few days later. Board’s job is to hire or fire CEO. Technically CEO made the coup since he managed to throw his bosses out of their positions.
marvin 3y ago
A majority of the OpenAI board members were heavily sympathetic to the AI doomsday cult mostly known through MIRI. A novel discovery spooked them, made them believe that The End Is Near, and that continuing on the path led by Altman would lead to imminent disaster. OpenAI is designed so that the board do pretty much whatever they deem necessary if this happens, and the board interpreted the necessary action as unilaterally halting AI development -- a staple objective of the AI doomsday cult. So the board did this. They jumped the gun - this extreme move might have become justified at a later point in time, but definitely not yet. In a pedestrian understanding of game theory typical of AI doomsday cult members, the board failed to anticipate that other actors may not agree with their interpretation of reality, and perform their own actions in response. This response largely boiled down to scattering the institutional knowledge of OpenAI to the winds, in their eyes greatly exacerbating the safety concerns that led to firing Altman in the first place. Altman, meanwhile, as a strong believer in OpenAI's mission, attempted to salvage the situation by brokering a deal to transfer OpenAI's institutional knowledge to Microsoft. After having their shitty options explained to them - exacerbate a potential safety risk by seeding 750 new AI companies, leave control of the situation to Microsoft's board or step down and retain control under the OpenAI non-profit, they folded like a wet paper bag. The whole saga was a waste of their board seats and influence on humanity's first deployment of AGI. They all got the outcome that their competence deserved.
- fireflash38 3y ago
  This doesn't sound biased at all
  - marvin 3y ago
    It's obviously a hypothesis, stitched together from extremely incomplete data. I've said from day one that I'm looking forward to read about this in the history books in 20 years. Maybe Altman's eventual bio will contain the truth.

by reso 3y ago

richbell 3y ago

The speculation seems to be that Sam was being duplicitous and trying to oust a board member he didn't like. Members of the board compared notes and realized he had misrepresented his conversations with other members in an attempt to build concensus. Throw in the rapid expansion of OpenAI (commercialization being in conflict with the board's vague mandate to do what's best for humanity) and the rumored (now confirmed) deal he was arranging with a company that he had a financial stake in, and they felt like they needed to remove him. However, by trying to do so swiftly and not allowing him a chance to retaliate they pissed off Microsoft and the employees. At that point, they were basically forced to reinstate him or the entire company would collapse — and if they do that, they can't also go on record clarifying why he's bad. * this is my vague recollection based on reading past discussions. I'm on my phone right now and unfortunately don't have any sources I can link, take this with a grain of salt. Edit: a few links https://news.ycombinator.com/item?id=38559770 https://news.ycombinator.com/item?id=38559770 https://news.ycombinator.com/item?id=38548404 https://news.ycombinator.com/item?id=38548404

JumpCrisscross 3y ago
Their communication strategy was also juvenile at best.
ethbr1 3y ago
Thanks for the citations. That was my guess... but only because it was the only scenario I could think of where the board being curiously and obviously intentionally vague about 'why' in the announcement, but still saying more than nothing, made sense.

pixelmonkey 3y ago

This recent TIME article lays out the saga pretty straightforwardly and makes it a bit less confusing. https://time.com/6342827/ceo-of-the-year-2023-sam-altman/ https://time.com/6342827/ceo-of-the-year-2023-sam-altman/ At least, that was how I felt after reading it. Basically, within the span of a year, OpenAI transformed from a research lab inside a non-profit that was pursuing a seemingly-Quixiotic dream of artificial general intelligence (AGI)... into one of the fastest-growing for-profit software companies of all time via its creation of the chat-based generative AI category (aka ChatGPT) and its consumer/enterprise SaaS and API offerings. The board -- or, at least, its 4 remaining non-CEO members -- thought that this was too much, too fast, and too soon, and that there was a narrow window of time where they could slow things down and refocus on the original non-profit mission. They also felt that Altman was a bit of a force of nature, had his own ideas about where OpenAI was going, and treated "board management" as one of his CEO skills to route around obstacles. Once a board loses trust of their CEO, unfortunately, there is usually only one blunt and powerful tool left: firing the CEO. And this happens pretty often. As the investor Jerry Neumann once put it, "Your board of directors is probably going to fire you."[1] Boards have very few ways to actually take action when they are worried about a company or institution; firing management is one of the few "course correction" actions they can take quickly. In OpenAI's case, if they had a for-profit board, that board would probably have been ecstatic with Altman and the company's progress. But this was not a for-profit board. It was a non-profit (mission-oriented) board meant to oversee the safe rollout of AGI. Those board members weren't sure the best way to do that was to become one of the world's largest for-profit software companies. I'd speculate that it was probably an emotional decision and the full implications were not entirely thought through until it was too late. I'd also speculate that this explains why Ilya Sutskever felt some immediate regret, because his goal wasn't to destroy OpenAI (or inspire an employee revolt) but to put its non-profit mission back into focus. I like to practice the principle of charity[2], and, in this case, I think the non-profit board was not acting maliciously but simply did not realize the knock-on effects of trying to replace a CEO when everything at the company seems to be "going right." I suspect Altman thought the best way to roll out AI was via iterative product development and fast revenue growth to finance the GPU demands, utilizing corporate partnerships (Microsoft), viral word-of-mouth marketing, and SaaS/API fees (ChatGPT). Running out of data center compute started to become a primary concern, so it wouldn't surprise me if safety took a backseat to this. Remember, all this growth happened in the span of a year. Perhaps Altman thought he was satisfying the safety concerns simply by talking to regulators, making iterative releases, and going on a speaking tour about it, but the board thought the only way to go safely was to go slower. I'm sure we'll learn more after some books are written about the episode. [1]: https://reactionwheel.net/2021/11/your-boards-of-directors-is-probably-going-to-fire-you.html https://reactionwheel.net/2021/11/your-boards-of-directors-i... [2]: https://en.wikipedia.org/wiki/Principle_of_charity https://en.wikipedia.org/wiki/Principle_of_charity

tivert 3y ago

> I am still so confused by this whole saga. No one has explained how or why the entire board decided to do a coup, and then simply changed their minds a few days later. I don't know how anyone can call the board's action a "coup." Calling it that seems to be a propagandistic abuse of the term. The board was in charge, and it's not a coup if it fires a subordinate (the CEO). If anything, the coup was getting the board ousted.

reso 3y ago
Sure, it was not my point to suggest it wasn’t legal, just that it was a bold action which has gone completely unexplained.
darkerside 3y ago
I think the coup is on the part of other employees who advocated for the board to fire Altman. No judgement in whether that was justified or not.
- tivert 3y ago
  > I think the coup is on the part of other employees who advocated for the board to fire Altman. No judgement in whether that was justified or not. The GP was pretty clear that he thought "the entire board decided to do a coup," which does not fit that interpretation. But even the scenario you describe isn't something that can be properly described as a "coup." In that case, the employees are just appealing to a legitimate higher authority, which is a totally OK thing to do (e.g. it's not wrong to report a bribe-taking boss to the company ethics hotline). IMHO, a coup is where subordinates illegitimately usurp and depose the highest authority from below.
  - darkerside 3y ago
    I probably missed that in the GP comment. I don't think invoking a higher power necessarily makes it not a coup. To me, a coup is when you angle to relieve someone from office above you by any means. But reasonable people can disagree on that, I suppose.

bogomipz 3y ago

There was a bit of insight today from a board member on her perspective. See: https://archive.is/Sy3Xm https://archive.is/Sy3Xm

Clubber 3y ago
That certainly doesn't leave a good taste in my mouth.
huytersd 3y ago
Honestly who cares what Helen has to say. She didn’t belong on a board in the first place.
- bogomipz 3y ago
  Maybe try to follow the thread where the OP states: >"No one has explained how or why the entire board decided to do a coup ..." To which I responded that there was in fact a very recent and relevant update to this. Your opinion on whether she belonged on the board is completely irrelevant to this fact and also childish.

nicce 3y ago

> No one has explained how or why the entire board decided to do a coup, and then simply changed their minds a few days later. Board’s job is to hire or fire CEO. Technically CEO made the coup since he managed to throw his bosses out of their positions.

marvin 3y ago

A majority of the OpenAI board members were heavily sympathetic to the AI doomsday cult mostly known through MIRI. A novel discovery spooked them, made them believe that The End Is Near, and that continuing on the path led by Altman would lead to imminent disaster. OpenAI is designed so that the board do pretty much whatever they deem necessary if this happens, and the board interpreted the necessary action as unilaterally halting AI development -- a staple objective of the AI doomsday cult. So the board did this. They jumped the gun - this extreme move might have become justified at a later point in time, but definitely not yet. In a pedestrian understanding of game theory typical of AI doomsday cult members, the board failed to anticipate that other actors may not agree with their interpretation of reality, and perform their own actions in response. This response largely boiled down to scattering the institutional knowledge of OpenAI to the winds, in their eyes greatly exacerbating the safety concerns that led to firing Altman in the first place. Altman, meanwhile, as a strong believer in OpenAI's mission, attempted to salvage the situation by brokering a deal to transfer OpenAI's institutional knowledge to Microsoft. After having their shitty options explained to them - exacerbate a potential safety risk by seeding 750 new AI companies, leave control of the situation to Microsoft's board or step down and retain control under the OpenAI non-profit, they folded like a wet paper bag. The whole saga was a waste of their board seats and influence on humanity's first deployment of AGI. They all got the outcome that their competence deserved.

fireflash38 3y ago
This doesn't sound biased at all
- marvin 3y ago
  It's obviously a hypothesis, stitched together from extremely incomplete data. I've said from day one that I'm looking forward to read about this in the history books in 20 years. Maybe Altman's eventual bio will contain the truth.