What’s happened now?
The artificial intelligence company OpenAI fired its co-founder Sam Altman, who then joined Microsoft. Some other people have changed their jobs too but none is as famous as Altman.
Do I need to care?
Not really, unless you’re a Microsoft shareholder, in which case you still don’t need to care that much.
Isn’t OpenAI meant to be a charity?
It was set up as a non-profit in 2015 but grew much too fast over the past couple of years, so it now resembles a bottom-heavy startup that’s overseen by a Temple of Grand Wizards.
The non-profit had struggled to raise money, because obviously, so it spun out a limited partnership in 2019 with two holding companies placed in-between the topco and the opco. The Temple has outright ownership of one holding company that controls the operating division of OpenAI, and there’s a separate holding company that’s an equity warehouse for charity employees and small investors. This second holding company has majority ownership of the operating division, which has Microsoft as a minority shareholder.
That’s complicated. Can I see it as a picture?
What’s the point of all these subsidiaries?
They’re to give Microsoft a stake in the commercial bit while wearing no responsibility for the Asimov-y stuff that concerns the management layer. This seemed to suit both parties. “From the beginning, [Microsoft] accepted our capped equity offer and our request to leave [artificial general intelligence] technologies and governance for the Nonprofit and the rest of humanity,” they said in 2019.
And the money goes from where to where in the structure?
Money goes into the lower levels, where it gets incinerated. OpenAI last year lost $540mn on $28mn in revenue. According to reports, Altman had wanted to raise another $100bn to reach the aim of creating artificial general intelligence (AGI).
OpenAI calls the operating division a capped-profit company because returns for the first round of investors are capped at 100 times the initial investment. “All residual value created above and beyond the cap will be returned to the Nonprofit for the benefit of humanity,” it says.
Microsoft has reportedly put in at least $13bn into OpenAI (though according to a Semafor report, it’s mostly been in comped cloud-compute usage rather than cash). Over-simplifying, therefore, Microsoft’s profit cap might kick in when the value of its minority stake exceeds $1.3tn — or to put it another way, when OpenAI approximately matches the value of Microsoft.
Do they realise how unnerving it is to keep saying things are done for the benefit of humanity, as if the plan is to colonise Mars and they want to select breeding pairs?
Why was Altman fired?
He was not “consistently candid in his communications”, OpenAI said, without elaborating. The general theme of reporting so far is that the board was split over Altman’s commercialisation efforts, including his tapping up SoftBank for an AI chip venture.
But also, the Temple was running low on Wizards. Three of the more commercially minded board directors resigned earlier this year — Presidential candidate Will Hurd, LinkedIn cofounder Reid Hoffman and Shivon Zilis, a VC who’s been friendly at least once with OpenAI founder investor Elon Musk — which reduced the board to six and gave doomwarts like OpenAI chief scientist Ilya Sutskever greater influence.
Things came to a head at OpenAI’s maiden developer day on November 6, where Altman announced moves to give developers more freedom to develop custom versions of ChatGPT. Paying customers will be able to make use-specific bots that can into external sources of information by feeding the machine up to 300 pages of prompts. But the safety-minded types weren’t keen on the idea of letting loose millions of baby ChatGPTs on corporate intranets and the open web.
If he upsets the geeks so much, why has Microsoft hired Altman?
To remove the risk that he sets up a competitor to OpenAI. The thing Microsoft wants least is to have sunk $13bn on backing the Betamax of AGI.
There’s been no suggestion yet that Altman will be restricted to one job. As well as the AI chip startup he has a nuclear fission company, an AI device developer and the Worldcoin eyeball-scanning project, so the time he can give to running Microsoft’s new AI special-ops division might be limited. But he won’t be running someone else’s AI special-ops division and that’s apparently what matters most: Microsoft shares are up a bit in Monday premarket.
Here’s what Macquarie analyst Fred Havermeyer thinks:
We think (Microsoft CEO) Satya Nadella may have just pulled off a coup of his own. If many OpenAI employees choose to migrate to Microsoft to join Mr. Altman and Mr. Brockman, then not only would Microsoft hold a license to OpenAI’s IP up to AGI, but Microsoft would also be effectively acquiring OpenAI’s core differentiation — its ambitious and experienced technical talent.
We think this team is well positioned to hit the ground running with experienced leaders who would likely be equipped with familiar Microsoft Azure infrastructure (OpenAI’s exclusive cloud provider) and Microsoft’s robust balance sheet. We believe Microsoft has both preserved its AI product roadmap and assembled formidable competition for generative AI startups and Google DeepMind. We can imagine a future where Mr. Altman’s team could specialize in not just AI model development but also vertically optimizing AI models from silicon to software and building new, category-defining apps.
What happens to OpenAI?
One possibility is that the weight of money will flip its upside-down structure — though for those of us on the outside looking in, it’s debatable how much that changes anything. As JPMorgan says this morning:
We knew prior to last week that OpenAI was ultimately controlled by a non-profit set up to “ensure that safe artificial general intelligence is developed and benefits all of humanity,” but no one had really considered the potential ramifications of that.
And it’s an interesting time for a coup. Before Altman was ejected, OpenAI had been in the middle of selling approximately $1bn of employee stock at a possible valuation of $86bn. Cleaning up the control structure is suddenly a priority for investors, while employees probably quite fancy bigger houses, so go-slow campaigners like Sutskever might not be too popular internally right now.
Emmett Shear, new CEO of OpenAI, wrote a lot on X about why he was breaking a paternity career break to take the job:
OpenAI is one of the most important companies currently in existence. When the board shared the situation and asked me to take the role, I did not make the decision lightly. Ultimately I felt that I had a duty to help if I could. I have spent today drinking from the firehose as much as possible, speaking with the board, a small number of major partners, and listening to employees.
“Drinking from the firehose”?
Yep. That’s how they talk.
Our partnership with Microsoft remains strong, and my priority in the coming weeks will be to make sure we continue to serve all our customers well. OpenAI employees are extremely impressive, as you might have guessed, and mission-driven in the extreme. And it’s clear that the process and communications around Sam’s removal has been handled very badly, which has seriously damaged our trust.
Shear says has a three-point plan for his first month. He’ll commission a report on how he got the job, organise a lot of meetings, and reform the management team (“including pushing strongly for significant governance changes if necessary”).
Before I took the job, I checked on the reasoning behind the change. The board did *not* remove Sam over any specific disagreement on safety, their reasoning was completely different from that. I’m not crazy enough to take this job without board support for commercializing our awesome models.
Does any of that mean anything?
Not really. But everyone seems to have been a bit freaked out by recent events so they’re all trying to sound reassuring, so as not to scare away the money.
Remind me, why don’t I care about any of this?
The power battle over the future of AI is an important and exciting story. It’s less exciting when a company hires the former boss of a dysfunctional subcontractor for defensive reasons.
The latter event has definitely happened. How it will affect the path of the former story is unknown but, for scare-away-the-money reasons, a lot of the people involved are trying to memory-hole the entire incident. No one worried about OpenAI’s weird management structure and grandiose mission last month; and the same may be true by next month.
In the meantime, maybe you’re among the 3 per cent of people who use Bing for searching, or the less than 11 per cent who use Edge for browsing, or the approximately 100mn users who ask ChatGPT to generate doggerel or make foraging guides or whatever. For you, absolutely nothing has changed either.