The AI Arms Race Begins
The OpenAI board took a shot at the king. Instead, they shot themselves in the foot
This is my off-the-cuff attempt to make sense of what happened over the weekend regarding OpenAI and Sam Altman and what's to come. As you'll soon realize, I'm as confused as anyone.
I’m posting updates to the bottom of this as necessary/ as long as I can be bothered.
In 'The Techno-Optimist Manifesto,' Marc Andreessen labeled trust and safety "the enemy" of technology progression. He decried that there was no place for people who wanted to hold tech companies and their overlords accountable, as all they did was slow down the march of progress. You can read my thoughts on why that's wrong here.
Well, this weekend, we watched the battle of those two philosophies play out in real time.
If you believe the hype, OpenAI is one of the most important startups to appear in a long while. And so, when news broke that the board had forced co-founder Sam Altman out, seemingly at random, it was a kinda-big deal. I'll spare you the headache of piecing it together. What you need to know is that:
On Friday, Altman was kicked out in a surprise announcement, citing that he had been dishonest and "not consistently candid in his communications."
Soon after, President and co-founder Greg Brockman announced he was leaving.
It becomes increasingly clear that OpenAI's Chief Scientist, Ilya Sutskever, was at the center of the coup.
Then, VCs got involved, pressuring the board to reverse the decision (and likely Microsoft too, who own 49% of OpenAI. It's thought that Microsoft CEO Satya Nadella was blindsided and informed only minutes before the news went public. He wasn't happy.)
A board meeting was held, with a deadline of 5 pm, to sort the mess out. Rumors were that the board was considering leaving, choosing new members, and reinstating Altman.
The deadline was missed.
Numerous OpenAI employees resigned in support of Altman.
Another board meeting was held the following day, with another deadline of 5 pm.
Despite various media outlets reporting Sam Altman was back in as CEO with a new board, the current board chose an interim CEO, Emmett Shear, the former chief executive of Twitch. Altman’s potential return was over.
Then, Microsoft announced it was bringing Altman and Brockman in to lead a new advanced AI division.
What's crazy is that a few days on, we're still left doing guesswork and theorizing what exactly happened. The board's communication gets a big fat 1/10. There's still no final word on why Altman was kicked out. It meant that all weekend, there were rumors abound of scandals. Had he been sacked for misconduct? Financial naughtiness? Malpractice? Data breaches? Was he lying about something serious to the board?
With the dust settling, it appears to be a little less scandalous, but certainly not without future ramifications. It's possibly the result of a tug-of-war between ethics/safety and the relentless pursuit of progress.
OpenAI's business structure is interesting. It has a non-profit board that acts as a balance and check, ensuring the company's work aligns with human interests. It controls OpenAI, Inc., the non-profit, which takes "precedence over any obligation to generate a profit."
Sam Altman had become the face of OpenAI Global, the capped-profit company. He was pushing it forward as a for-profit company, driving users, growth, VC investment and valuation (it was about to be over $90 billion.) The way the structure works also gave Altman no control, evidenced by the board's ability to fire him on a whim. Ultimately, it looks like the board couldn't handle how OpenAI was operating. It had lots of employees, lots of deals with corporations, and was on track to make over $1 billion in revenue, clashing with the aims of the top level.
We can figure this out based on the snippets that appeared over the weekend. Ilya Sutskever told staff that kicking out Altman was the "only path" to do AI right at OpenAI. The new CEO has already come out and spoken similarly, saying,
"I specifically say I'm in favor of slowing down, which is sort of like pausing except it's slowing down. If we're at a speed of 10 right now, a pause is reducing to 0. I think we should aim for a 1-2 instead."
Until we hear otherwise, the theory that board members lost trust in Altman seems most plausible. With rumors surfacing that OpenAI is getting close to creating AGI, Altman's inability to pause for a second to consider the ramifications of the work could have been the breaking point. In order to get back some control, a few board members orchestrated a coup.
But let's not beat around the bush. The OpenAI board may have got its wishes — the chance to pause — but it's been a disaster for them. If the board wanted to build safe AI and was worried about how serious Altman was about AI risk, they have miscalculated their actions. By sacking him, they unintentionally created a new AI division inside a profit-maximizing $2.7 trillion tech company led by the very person they had doubts about.
Ilya Sutskever appears to regret the decision already. A few hours ago, he tweeted,
"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."
As of today, OpenAI is a different company. From the world's most important AI startup to, well, nobody is entirely sure yet. What does it mean for its flagship product, ChatGPT? While I agree that the pace should be slowed, it will likely be a death sentence for the product because we can guarantee its competitors — Altman now included — will not be slowing down. Altman appears to have rubbed the magic lamp and got his wishes. He has broken free of the restraints put in place by a board that wanted to ensure humanity's best interests are at the heart of OpenAI, and now he can run free. And run free, he will.
The OpenAI drama shows that when powerful companies are left with a choice to slow down or speed up, only one option wins out, no matter the consequences. It's a sad outcome and a sorry reflection of what tech companies truly value. OpenAI's vision of being a true non-profit for the people, a company producing AI for the benefit of humanity, is a noble goal. But alas, that goal wouldn't enrich VCs or bestow power into the hands of a few individuals. That just won't do.
The progress OpenAI has made with guardrails has been astounding. Now, they’re likely gone.
I think this is the beginning of the real AI arms race.
One where it's less about why we should build the technology, less about what purpose the technology should serve, and more about who can build it first and get the biggest valuation. And with how this saga has unfolded — calamitous, miscalculated and messy — I'm not certain these people should be trusted with superhuman AI.
(It’s important to iterate this is a fluid story. Even as I type this, some 500 of the 700 employees have threatened to leave OpenAI and join Microsoft’s new AI division run by Sam Altman if the board doesn't resign and reinstate Altman and Brockman. This even includes Ilya Sutskever! If this were to happen, then forget everything I’ve written above, and expect OpenAI to become what it likely already was becoming… a for-profit organisation. And if that doesn’t happen — would Satya Nadella sanction it now that he got basically got OpenAI for $0? — what happens to OpenAI? It could well be the end of the road.
I may update this story, or write a follow-up, depending on how it all plays out over the coming days. What a batshit crazy story.)
Update 1: The number of employees signing the letter has now reached 700 (out of 770 total). Begs the question… what is the board going to be a board of?
Update 2: For fuck sake. This is getting tedious. According to Alex Heath, the deal with Microsoft isn’t actually signed, sealed delivered, and pressure is mounting on the board to resign “gracefully.”
Update 3: As of Tuesday afternoon, Sam and the OpenAI board are once again in talks over his possible return. Possible scenarios include Sam serving as a director.
Update 4: Well, there we go folks. It appears Sam Altman will return as CEO, with a new board made up of Bret Taylor, Larry Summers, and Adam D’Angelo (who initially voted to fire Altman). This temporary board will eventually pick new members to “reset the governance of OpenAI.” In other words, Altman and the VCs got what they wanted — the non-profit side has no power anymore.