- By Amy Castor and David Gerard
- We need your support for more posts like this. Send us money! Here’s Amy’s Patreon, and here’s David’s. Sign up today!
No one has ever attempted to move a 700-person polycule from SF to Redmond but I’m told the resources exist.
— Jacob Silverman
After Sam Altman was booted from OpenAI on Friday, the one question was: what will Microsoft do?
Microsoft has put $13 billion into OpenAI. It put in $1 billion in 2019 and another $2 billion in the years since. In January 2023, the company pledged an additional $10 billion in capital and infrastructure credits — i.e., compute time on the Azure cloud — though not all of that has been drawn.
Microsoft CEO Satya Nadella announced early on Monday morning that Microsoft would be starting its own AI unit — with Sam at the helm. Microsoft would also hire Greg Brockman, the former OpenAI president who was also booted from the board and quit his job in solidarity with Sam.
Microsoft’s stock price shot up on the news — even as the deal isn’t signed as we write this. [Twitter, archive]
Restoring the rightful order in Silicon Valley
OpenAI is not a regular tech company. It was formed as a nonprofit to develop artificial general intelligence (AGI) to benefit “humanity as a whole” and keep developers from creating a Torment Nexus. [Twitter, archive]
OpenAI’s chartered duty is to humanity, not to making big money into bigger money. But it turns out big money is also important. Training machine learning models isn’t cheap.
If OpenAI does put Altman back on the board, he wants the remaining board members — Adam D’Angelo, Helen Toner, Ilya Sutskever (maybe), and Tasha McCauley — replaced with Sam-friendly people who aren’t full-on AI doomsday cultists. He also wants his name cleared.
If a Sam-friendly board is put in place, then no enterprise or government will take anything OpenAI says about AI safety seriously henceforth — and OpenAI will just become a regular tech company focused on number going up, not a research company with AI doomers guiding its ethics.
Sam is venture capital’s guy. He’s the face of “AI” for the VC world. The VCs simply could not suffer the humiliation of Sam being ousted. They want him back.
Sam’s VC buddies have been working the press hard since Friday, trying to pressure the remaining OpenAI board members. If you see an article sourced to “multiple people familiar with discussions,” think to yourself which of the warring factions the “multiple people” likely belong to.
Right now, the VCs who put money into the for-profit arm of OpenAI are talking up the idea of suing the nonprofit board for their losses, to put added pressure on the board to resign. [Reuters]
If Altman and Brockman do go to Microsoft, OpenAI becomes an empty shell with no funding — a nonprofit board of nothing.
Sutskever, who led the coup against Sam, has already crumbled. He says he’s sorry — “I deeply regret my participation in the board’s actions. I never intended to harm OpenAI. I love everything we’ve built together and I will do everything I can to reunite the company.” [Twitter, archive]
Altman now needs two of the three remaining Sam-opposed board members to flip. [Verge]
Palace intrigue
Nobody has revealed precisely why Sam got booted — but several past and present OpenAI employees who spoke with the Atlantic said the tension started with the release of ChatGPT. The company was growing too quickly.
“After ChatGPT, there was a clear path to revenue and profit,” one source said. “You could no longer make a case for being an idealistic research lab. There were customers looking to be served here and now.” [Atlantic, archive]
OpenAI was torn between two growing factions at the company — the idealistic, like Sutskever, who feared AI taking over the world, and the commercial, like Altman and Brockman, who were pushing for more product releases, sometimes before the products were ready.
Sutskever began to behave like some sort of cultist:
At OpenAI’s 2022 holiday party, held at the California Academy of Sciences, Sutskever led employees in a chant: “Feel the AGI! Feel the AGI!” The phrase itself was popular enough that OpenAI employees created a special “Feel the AGI” reaction emoji in Slack.
Altman, meanwhile, was trying to drum up money from Softbank and Middle Eastern investors to build a chip company so OpenAI could own its computation. He wanted an OpenAI that worked like any other fast-growing Silicon Valley startup. [Bloomberg, archive]
New seeker falls off broomstick
OpenAI has hired a second interim CEO, Twitch cofounder Emmett Shear, to replace Mira Murati, who held the position for two days. Shear wants to hire an independent investigator to find out what the heck happened here. [Twitter, archive]
Shear appears to fall into the idealistic category. He takes Eliezer Yudkowsky seriously. He also had a cameo in Yudkowsky’s Harry Potter and the Methods of Rationality. [Twitter, archive; HPMOR]
But unlike Yudkowsky, who thinks that the rogue superintelligence will absolutely, positively, destroy humankind one day, Shear puts the probability of AI doom at a mere 5% to 50%. [Twitter]
In the manner of AI doom prophets throughout recent history, Shear has never done anything so tawdry as showing how he worked out these numbers. You might be forgiven for thinking that these guys pull this sort of number out of their backsides so that they can announce scary numbers in a confident voice.
Altman’s job at OpenAI was not in any way technical. He dropped out of Stanford computer science after two years to chase money in startups. But he inspired the team. So OpenAI’s 770 employees want their old CEO, not this new guy.
When Shear called for an all-hands meeting on Sunday at the company’s San Francisco headquarters, the employees refused. One responded in Slack with a rude emoji. [Verge]
By Monday afternoon, 700 of the OpenAI staff had signed a letter saying that they would quit and go to Microsoft if Sam didn’t return. They wrote that they were “unable to work for or with people that lack competence, judgment and care for our mission and employees.”[Wired; archive]
Sutskever also signed the letter — because hey, we all want to find the guy who did this. Murati signed too.
Ask Clippy
Microsoft also has rights to OpenAI’s source code and training data. It could start a unit with Altman as an inspirational tech leader and let him cherry-pick who he wants there. Microsoft could effectively buy OpenAI for nothing. The company is already extending feelers out to OpenAI staff — though on a very noncommittal “if needed” basis. [BBC]
Whether Microsoft actually wants to swallow OpenAI is another question. Working for a startup is a very different experience from working for a large corporate office supply company. Nobody who thought they were changing the world is going to stick around to work on text generation for Outlook email, including Sam. And Microsoft is smart enough to realize this.
Microsoft’s ideal outcome is that Altman goes back to OpenAI and the flows of cash and firewalling from culpability continue as they did before all this unpleasantness.
What Microsoft wants is to rent out computation on Azure. Cloud computing is a commodity, and one that’s only getting cheaper. But the supply of graphics cards for number crunching is rather more constricted. [Paris Marx]
If there’s demand for “AI” products — whether or not they even work — then there’s money renting out the number crunching the machine learning will need. That’s what Microsoft is in this for.
This provides a more robust and business-friendly substrate — without those annoying “ethics” people — for AI’s real use case: abusing labor and customers.
Update 11/22/2023: Sam has been reinstated. Venture capital won and OpenAI is now just another startup whose goal is to grow like a cancer. The paperclip maximizer is satisfied. [Twitter; archive]
Image: Sam Altman
Amy Castor 