Notes from EA Global Boston 2017
This weekend, I attended EA Global 2017 in Boston. Unlike some previous EA conferences, this one explicitly focused more on long-term issues, science, technology, and object-level improvements, rather than short-term causes and the marketing of EA itself. Some quick notes on things that especially stood out, since I hope future conferences continue with this direction.
I was really impressed by how much professors Kevin Esvelt and Allan Dafoe cared about existential risk and working to prevent it. While I disagree with some of their planned mitigation strategies, it’s obvious that they’ve put a lot of work into analyzing the situation and doing what they can to improve it. I worry sometimes that people who talk about x-risk are just imitating Yudkowsky and Bostrom, and wouldn’t take independent initiative or try to figure things out for themselves. I’m now a bit less concerned about that.
Elizabeth Edwards, a New Hampshire state representative, talked mostly about getting elected to office as an EA. While she covered a lot of the depressing stuff about the cynical world of political maneuvering and rounding up votes, I think most of us already kind of knew (or at least suspected) that. For me, the most important takeaway was probably that people can be good or bad candidates based on seemingly random factors, often ones they have no control over. For example, someone who happened to have been born and raised in Berkeley would have a big advantage running there. Even silly things, like how easily your name fits on a sign, can help a person win. Someone who happened to be a good candidate could easily find elections to be their comparative advantage, even when others with an identical resume wouldn’t; formal qualifications are only moderately important nationally, and even less locally.
Jason Matheny and the IARPA team seem extremely skilled at choosing good technology development/research programs for defense purposes (including x-risk mitigation). However, I worry they’re overly optimistic about the federal contracting process, which all of their money has to flow through. Even a perfect program won’t work well if the best teams never apply, or if they’re all turned off by security clearance requirements. At least as an outsider, it seems hard to discover how much real benefit past IARPA programs had, even when nominally successful. Technology being funded, built, and then left to rot on the shelf seems like a real concern.
Jim O’Neill opened his talk with the joke, “It’s great to be back at the Singularity Summit”. One of the main EAG organizers was Amy Willey, who also helped organize several Summits, and the two shared many topic areas. However, the conference seemed to lack many of the business leaders and “big names” who attended the Summits. This might be caused by not having Kurzweil as a name recognition draw, or the EA brand being most popular with students. I’m guessing it might also come from the conference website, which required writing an application before buying tickets. Although (I’m assuming) it’s intended to make the conference higher-quality, I’d intuit that very busy or famous people might not want to bother. (IIRC, at least one billionaire bought a Summit ticket and then didn’t come. This type of person seems likely to be deterred by trivial inconveniences.)
The informal events during the conference were pretty good, although not spectacular. They were helped a lot by holding the conference in dense, highly-walkable Cambridge. The easiest thing to improve here would just be doing more marketing, as things are limited when there’s only a few hundred attendees.
I didn’t go to the career coaching session by 80,000 Hours. Although this seems potentially very high value, I was a bit discouraged by many informal past reports from people who wanted new careers, were very active in EA and sought out lots of help, and got little actionable advice. I suspect the root cause of this is, to paraphrase Eliezer a bit, people “not accepting the implications of nobody knowing anything better to do about x-risk and having to channel all his concern into calling for more people to be concerned, a useless meta-strategy with no object level that I suppose still feels to him like doing something”. This is a hard problem, but I hope this group can help address it a bit as it develops.