Adam runs a company called Quora, which has a product called Poe, which uses large language models, including those of OpenAI and some of its competitors.
The way I perceived it was after GPT-4 was demoed to the board in summer 2022, Adam began taking his responsibilities as a board member more seriously, because the technology seemed to be advancing, and he became a more engaged board member.
In the lead-up to — between that time in summer 2022 and April 2023, we had had several conversations as a board about what kinds of conflict of interest were acceptable or unacceptable on the board, because many potential board members, and current board members, had various involvements with various AI companies.
So we had fairly detailed discussions about what was an unacceptable conflict of interest and had decided that being closely involved with a company that was training its own large language models, you know, highly advanced frontier language models that would compete with OpenAI’s, was the bar for excessive conflict of interest.
So it was surprising to me when Sam emailed the board in April 2023 saying that Adam’s conflict of interest had grown too large and seemed like he needed to step off the board, and did we agree. Because Adam’s company produced a product that used others’ LLMs, they didn’t — they weren’t training their own. So clearly it didn’t meet the conflict of interest criteria we had all discussed.
When I said as much via email, Greg Brockman chimed in with a different reason to remove Adam, namely that his position as both a customer and a board member was creating communication difficulties internally. I forget who exactly said what on the email chain, but other board members raised questions about that or wanted to know more about that.
Ultimately, I spoke to Sam on the phone, and we sort of — at my urging, we agreed that, surely, the step before just removing Adam from the board, if the problem was how he was communicating inside the company, surely, the next step would be to discuss that with him and see if we can improve the situation. Sam said he would do that, he would have a conversation with Adam, to try and improve how he was communicating inside the company. And then the situation seemed to go away.
I later found out that Sam had never had that conversation with Adam, or that he had talked with him but had never actually tried to solve that problem, but, instead, had just said the only thing that he, Sam, didn’t like about Adam’s product Poe was that it used Anthropic models, because Anthropic was a competitor.
So, all in all, the situation seemed to me like there wasn’t actually a clear, concrete reason to ask Adam to move off the board, but that Sam and Greg were sort of searching for an excuse because he had been providing more active governance of the company.