The national security dimension of OpenAI’s leadership struggle
As the very public custody battle over OpenAI’s artificial intelligences winds down, I would like to point out a few facts, and then comment briefly on their possible significance.
It has already been noticed that “at least two of the board members, Tasha McCauley and Helen Toner, have ties to the Effective Altruism movement”, as the New York Times puts it. Both these board members also have national security ties.
With Helen Toner it’s more straightforward; she has a master’s degree in Security Studies at Georgetown University, “the most CIA-specific degree” at a university known as a gateway to deep state institutions. That doesn’t mean she’s in the CIA; but she’s in that milieu.
As for Tasha McCauley, it’s more oblique: her actor husband played Edward Snowden, the famous NSA defector, in the 2016 biographical film. Hollywood’s relationship to the American intelligence community is, I think, a little more vexed than Georgetown’s. McCauley and her husband might well be pro-Snowden civil libertarians who want deep state power curtailed; that would make them the “anti national security” faction on the OpenAI board, so to speak.
Either way, this means that the EAs on the board, who both presumably voted to oust Sam Altman as CEO, are adjacent to the US intelligence community, and/or its critics, many of whom are intelligence veterans anyway.
((ADDED A WEEK LATER: It has been pointed out to me that even if Joseph Gordon-Levitt did meet a few ex-spooks on the set of Snowden, that doesn’t in itself equate to his wife having “national security ties”. Obviously that’s true; I just assumed that the connections run a lot deeper, and that may sound weird if you’re used to thinking of the intelligence community and the culture industry as entirely separate. In any case, I accept the criticism that this is pure speculation on my part, and that I should have made my larger points in some other way.))
Those are my facts. Now for my interpretation.
Artificial intelligence is a national security issue. It is a technology with strategic implications. OpenAI is an American company. There is no way that such a company is allowed to operate without some kind of national-security oversight existing. One way to do that, is to have people on the board.
The way things are going, OpenAI is going to end up even more tied to Microsoft than it already was. Of course, Microsoft would already have a relationship to the state organs of American national security. In fact, those relationships would be decades old, and very familiar on both sides.
So from a national security standpoint, OpenAI is transitioning from, say, ad hoc supervision by its board, to integration with the existing relationship between Microsoft and the deep state—however that works.
As we try to understand what happened over the past few days, I would suggest considering whether national security representatives played a role, especially given the composition of the OpenAI board. To me, it’s conceivable that e.g. GPT-5′s progress set off alarm bells, and someone said we need to bring OpenAI into a more closely supervised relationship, right now.