So it's been about a week since Elon Musk and Sam Altman and that whole crew put out the announcement about OpenAI, and I've been chewing on it instead of writing about it right away, which is probably the opposite of how blogging is supposed to work. Everybody else had their hot take up within about six hours. Mine's late. I'm okay with that.
For anyone who missed it in between wrapping presents and dodging Star Wars spoilers (more on that in a second): a group of very rich, very online tech people announced a new nonprofit called OpenAI. Musk is in it, obviously. Sam Altman, who runs Y Combinator, is in it. Peter Thiel's name is on there too, which is its own kind of interesting given some of the things he's said about AI risk versus what he funds. Greg Brockman is the CTO, and they pulled Ilya Sutskever away from Google to be director of research, which if you follow the machine learning world even a little is a pretty big deal, kind of like poaching a starting pitcher mid-season. The headline number everyone's repeating is a billion dollars pledged, though from what I can tell that's a commitment over years, not a check somebody cut on day one. Amazon's kicking in cloud credits too.
The pitch is that AI research has gotten too concentrated inside a handful of giant companies, and that's dangerous long-term, so this outfit is going to do the research in the open and publish it, patents and all, instead of locking it behind a corporate wall. Ensure it "benefits humanity," is the phrase they keep using.
I want to like this more than I actually do. Not because I think AI research being secretive is fine, I don't, but because "trust us, we're doing this one for humanity" is a sentence that a group of billionaires has said about several other things over the years and it doesn't always age well. Also there's something a little funny about the guy who's constantly warning that AI could be an existential threat to civilization turning around and personally funding a lab whose whole purpose is to build more of it, just under a nonprofit banner instead of a corporate one. I get the logic, better it happens in the open with good intentions steering it than entirely inside Google's black box, I just don't think "open" automatically means "safe," and I don't think a mission statement is a substitute for actual constraints on what gets built. We'll see. Ask me again in five years.
What I do think is true is that this is a genuinely different structure than what we're used to seeing. A nonprofit lab with that kind of money and that kind of talent, explicitly saying it won't be optimizing for shareholder returns, is not nothing. Most of what passes for "AI news" in 2015 is a chatbot demo or a self-driving car clip, and this is neither of those. It's infrastructure-level, further upstream. Whether it actually publishes the good stuff once the good stuff exists is the part I'm reserving judgment on.
Unrelated tangent, because it's my blog and I get to do this: I have tickets for the Thursday night showing of The Force Awakens tonight, 7pm, and I have successfully avoided every trailer breakdown, every leaked plot summary, every well-meaning coworker who wanted to tell me "just one thing" about it. My brother-in-law has not managed the same discipline and has already told me he cried at the theater during a screening two days ago, so now I'm sitting here trying to guess what part that was instead of just waiting three hours to find out. This is worse than not knowing anything. Half-information is always worse than none.
Anyway. Somewhere between a research nonprofit promising to save us from AI and a movie theater full of people who've been waiting since 1983 (or 1999, if you count the prequels, which I try not to), it's been a strange week to be paying attention to technology. I don't have a clean way to tie those two things together and I'm not going to force one.