Firstly, if you are based in the US, I hope you have a wonderful Thanksgiving break over the coming days. So, I’ll keep this week’s post short enough to read while you’re queuing for your gate.
Many of us who are interested in AI knew Sam Altman was the CEO of OpenAI on Friday afternoon. Usually, when a CEO is in place on Friday afternoon till close of business, you kind of expect they will be on Monday morning, too. This actually could have been one of the outcomes, strangely enough.
As I write this, the saga that has robbed tech folks of sleep since Friday night continues. It still seems somewhat possible that Altman may return to OpenAI as CEO, if the board resigns.
For those of you who weren’t paying attention to this, here is a little bit of context. OpenAI is probably the world’s leading company in the AI space - certainly the world’s leading independent company (DeepMind is part of Alphabet).
OpenAI has a strange corporate structure, where there is a holding non-profit company at the top of the group - this is where the board sit and they have the power to hire and fire executives as they see fit, including Altman. The profit-making company is a subsidiary and this is where the investment into OpenAI has gone. Let’s not gloss over the investment… it’s huge - Microsoft(MSFT) alone have poured in about $13 billion to date. This triggered an AI investment arms race, with Alphabet even making a huge investment into Anthropic, which feels like a hedge against DeepMind performance.
The board decided to fire Altman on Friday night, and it has since been revealed that two reasons for this have been shared with OpenAI staff:
One explanation was that Altman was said to have given two people at OpenAI the same project.
The other was that Altman allegedly gave two board members different opinions about a member of personnel. An OpenAI spokesperson did not respond to requests for comment.
These explanations didn't make sense to employees and were not received well, one of the people familiar said. Internally, the going theory is that this was a straightforward "coup" by the board, as it's been called inside the company and out. Any reason being given by the board now holds little to no sway with staff, the person said.1
These are the most ridiculous reasons to fire anyone, let alone a CEO, and especially where firing said CEO would jeopardise the company.
It is common at big tech companies for multiple similar projects to be run concurrently. If something is important enough and you have the resources, you can hedge your bets, especially if there are competing philosophies on approach. So, this is not a reason to fire a CEO. It’s arguable that, with the fragile competitive edge OpenAI has, this is in fact a responsible thing to do.
The second reason doesn’t even sound like a problem. It doesn’t even say “conflicting” opinions. You can hold different opinions about a person concurrently. They can change within the course of a day, depending on their behaviour.
I agree with the OpenAI staff, this is a coup - but why? The real reasons are still elusive.
Breach
Early on, it was thought that what Altman was not “entirely honest” about, was that there had been a data breach at OpenAI and data had leaked into training that shouldn’t have. This is beginning to seem less and less likely, as no further details have emerged or any evidence of this. Microsoft offering Altman and all OpenAI staff roles doesn’t fit well with this theory, either.
Convergence/Alignment
There has been a lot said about the risks of AI and lobbying government to regulate AI. The AI that has triggered this - LLMs - is not AGI. Even if we made it supremely capable of any task, it could not take over the world or become sentient. Not without a great deal of engineering by bad actors, and even then this motive and work would be what the problem was, rather than the LLM. Altman and others including board member Sutskever, have been lobbying government for regulation for AI - I believe with a bootlegger’s mentality. For this reason, I have no sympathy for Altman - maybe this is karma. At the same time, it doesn’t mean it’s right.
Politics and Coup
This seems like a frontrunner. Perhaps the board just didn’t like Altman. Perhaps he has been able to ride roughshod over their concerns and ideas for a long time with the tailwind he’s had. Perhaps they felt this was the last time they could do this before he became unassailable. We’ll probably never know if this is the case, but if we don’t ever find out then it it could well be likely that this was the reason.
Feigned Politics and Coup
“You follow drugs, you get drug addicts and drug dealers. But you start to follow the money, and you don't know where it's gonna take you.” - Det. Lester Freamon
If you follow the money and the ownership - at the beginning of the weekend, Microsoft was a minority shareholder or a subsidiary of what is evidently a poorly-structured group. They were apparently not consulted about firing Altman despite their huge investment, which, if true, would be very unusual. If I were the OpenAI board, I wouldn’t want to be poking Microsoft’s legal army department.
Right now, it looks like one of two things will happen. Either the board at OpenAI quits and a qualified board is appointed (as in, one that has either Satya Nadella or someone at MSFT on it) bringing Altman back OR MSFT ends up effectively owning all of OpenAI, because it hires most of their people at once.
There will be setbacks initially if the latter happened, as MSFT won’t own OpenAI assets like training datasets, but these would definitely be temporary. MSFT would have the right to use OpenAI’s models as they do today, until such a time as they have parity or superior models of their own to replace them with.
For all we know, the terms of MSFT’s investment might state that any disruption in service grants them access rights to OpenAI assets, like datasets and weights. If most of OpenAI’s staff leave, we would most likely see some degradation in service or lack of maintenance.
In short, OpenAI would get restructured after a board change - both in MSFT’s favour OR MSFT would acqui-hire OpenAI with modest setbacks, also hugely in their favour. So, if you follow the money in this situation, it could be seen as logical for MSFT to have triggered this chain of events. They get to control or own the jewel in the crown of AI after the events of one weekend - absolute bargain.
Something else
There are other actors out there who are desperately trying to catch up with OpenAI and the possible disruption to OpenAI progress from the weekend’s events could be the leveller that these companies and states need. The motivations behind the coup are not well understood at a personal level, yet.
The board responsible for firing Altman is comprised of Adam D'Angelo, CEO of Quora; Tasha McCauley, a tech entrepreneur; Helen Toner, of the Georgetown Center for Security and Emerging Technology; and Setskever. They don’t seem to have done anything illegal, but whoever ends up governing OpenAI after their time is over should lift the veil and allow for a civil investigation of these board members personally. They are either incompetent, corrupt, hiding a material truth or all of the above. While Microsoft will likely benefit, the other investors in OpenAI will lose out, and unfairly so.
Disruption
I do really feel for the OpenAI staff, who have had the rug pulled out from under them. They have more right to truth and good governance than anyone externally. Many joined OpenAI with a clear understanding of the value of their option equivalents and where their company was going. This has been completely shattered.
For the rest of the industry, many of whom build upon OpenAI’s products, myself included, this is a concerning time.
Will the maintenance and stability of OpenAI’s existing products be affected by the disruption?
Will confidence in the AI sector be reduced in the minds of customers and investors?
If MSFT do end up owning OpenAI in effect, it becomes an even more powerful company than it was. If MSFT had tried to buy OpenAI at the point they made their last investment, the US FTC, UK CMA and the European commission would have likely blocked or slowed and investigated the acquisition. The acqui-hire route allows them to avoid all of this. MSFT have recently had a tough time with these regulators.
There is still a lot of uncertainty about what will happen next, with new facts being released regularly. We could still see a very different outcome. We could see the OpenAI board resign, a new board appointed which restructures OpenAI without ceding control to MSFT, and with Altman returning - as close to back to BAU as possible.
Update - I found this enlightening after finishing writing this post:
Microsoft CEO Satya Nadella appeared on CNBC and Bloomberg TV. When asked directly by CNBC’s Jon Fortt if Sam Altman and OpenAI’s staffers would join Microsoft, Nadella said “that is for the OpenAI board and management and employees to choose.” He followed by saying that Microsoft “chose to explicitly partner with OpenAI [and] obviously that depends on the people at OpenAI staying there or coming to Microsoft, so I’m open to both options.”
On the topic of whether Microsoft needs a seat on OpenAI’s board, he said that “it’s clear something has to change around the governance — we will have a good dialogue with their board on that, and walk through that as that evolves.”
On Bloomberg TV, Nadella told Emily Chang that “surprises are bad” and Microsoft will “definitely want some governance changes. This idea that changes happen without being in the loop is not good.”2
https://web.archive.org/web/20231120233119/https://www.businessinsider.com/openais-employees-given-explanations-why-sam-altman-out-2023-11
https://www.theverge.com/2023/11/20/23969586/sam-altman-plotting-return-open-ai-microsoft