A week after Sam Altman resumed his role as OpenAI’s CEO, the most salient explanation for the board’s decision to fire him remains the only one it has publicly given so far, leaving some OpenAI employees feeling uneasy, people familiar with the matter said.
The board said at that time that Altman was “not consistently candid in his communications,” which one of the people said on Wednesday made it impossible for directors to uphold OpenAI’s mission to ensure advanced AI benefits all of humanity.
But the lack of details has been unnerving for some OpenAI workers. A number of them began pursuing job opportunities at competing AI companies even after it was clear Altman would return. For instance, a person familiar with the matter said that Cohere has received several inquiries from OpenAI employees in recent days.
“We are happy that Sam is back but we are still in the dark about a lot of things,” a senior OpenAI staffer told Semafor. “We’ve been told repeatedly that the heart of the board’s dispute with Sam was not over AI safety, but we haven’t been told what that is. So things still feel shaky.”
As part of the deal to return, Altman agreed for OpenAI’s new nonprofit board of directors to commission an independent investigation into the events surrounding his outster. He and fellow OpenAI co-founder Greg Brockman also did not retain their board seats, which they initially pushed to keep, according to a person familiar with the negotiations.
The person said that the board did not fire Altman because of risks posed by OpenAI’s technology, nor was it worried about the pace of the company’s AI advancements. The members also never received a letter from staff researchers about a breakthrough called Q* as had been reported earlier by some news outlets.
For now, OpenAI’s research team is back working on their laptops, which had been locked during the five-day saga last week over fears that staff would possibly copy the company’s codebase or other intellectual property, other people familiar with the matter said.
A spokesperson for OpenAI declined to comment.
— Gina Chon contributed to this report
Even though Altman is back as CEO after an embarrassing debacle that almost sank the entire company, the business is in a different place than it was two weeks ago. Something caused the board of OpenAI to fire Altman, and whatever that something was, it hasn’t been resolved.
Anyone looking to invest in OpenAI or join its ranks now has to consider what just happened and what it means for the future of the company.
OpenAI still owns GPT-4, the most advanced AI model on the market. That alone makes it a valuable startup that will surely attract talented employees and investors.
But it’s an increasingly competitive field, and OpenAI’s rivals can use the drama to lure talent. Seeing a vulnerable market leader will also motivate others to seize the moment to try to surpass it.
Venture capitalists who passed on OpenAI are feeling vindicated. That’s probably a bit of schadenfreude, but there’s also some truth there. To put money into a nonprofit, or a for-profit controlled by a nonprofit, required a bit of a leap of faith.
What might worry some of OpenAI’s investors now, though, is that the company is in a race so competitive that practically every day matters. OpenAI just wasted a bunch of those crucial days and, worse, the distraction is still not over.
Its key customers are other businesses that utilize OpenAI’s technology to create their own applications. A lot of the work OpenAI has been doing is finding ways to reduce costs and make its products so user-friendly that customers flock to them and don’t even bother with the competition.
But those customers aren’t locked in. In fact, it’s rare to find a company that utilizes OpenAI models exclusively. If another firm comes up with a better model, OpenAI will lose business.
The company is also facing a threat from less advanced technology. Meta’s free, open-source AI model called LLama2 is rapidly catching on, according to people privy to cross-industry data. Llama isn’t just cheaper, it’s also faster and can be run on most cloud platforms.
If OpenAI had a defensive moat before the coup, it was a shallow one built in large part on the company’s mystique. Some of that is gone now and the moat is that much shallower.
Room for Disagreement
Last week’s chaos prompted public praise for Altman from a number of tech luminaries and OpenAI employees, further elevating the CEO’s personal brand. That suggests the startup will be just fine, argued Jeff Beer at Fast Company. “Altman brings with him an incredibly positive brand halo that may just result in the overall OpenAI brand being even stronger than ever,” he wrote.
- Breakthroughs in artificial intelligence are highly subjective when they first happen, and it takes a long time for other researchers to “build upon and bear out how replicable, effective, and broadly applicable the idea is,” Karen Hao wrote in The Atlantic.
- OpenAI is facing a plethora of other challenges beyond its governance structure, including addressing security vulnerabilities in its new custom chatbots, according to WIRED.