143Joined Dec 2017


Sorted by New


Narrow AIs have moved from buggy/mediocre to hyper-competent very quickly (months). If early AGIs are widely copied/escaped, the global resolve and coordination required to contain them would be unprecedented in breadth and speed.

I expect warning shots, and expect them to be helpful (vs no shots), but take very little comfort in that.

"Leadership" and "eco-systems" sound very nice as far as they're described here but I find this post unhelpful as a guide to what "EA" should do. 

Assuming this post is addressing EA funders – rather than the collection of diverse, largely uncoordinated, people, organizations, and perspectives that 'EA' is – is the claim that funders should open 20 of these offices? Who do they pay to do that and apply the "high standards" for early membership? What are the standards? Should people have models of the world that distinguish good/bad opportunities and big/small ones? At what point does answering these questions become too analytical? 

"Find all the smart altruistic people, point them to each other, give them some money, and let them do what they want" sounds nice, but aren't there hundreds or thousands of organizations interested in funding various projects, not least of which the whole VC industry? My sense is that not analyzing why you might be the funder of last resort, at least a little bit, is a recipe to crash and burn very quickly. $1m/yr/office could feed a handful of people and keep the lights on, but it's not scaling any projects. "EA" doesn't have enough money to last long without a lot of analysis and it's only been around for ~10 years.

People with diverse, niche interests and moxie have had really outsized influence on the world. It's easy to say "go find them," but the ones who will actually make a difference are very very few and far between and it takes some analysis to find them. There are a million people in Port Au Prince and probably hundreds of discernible perspectives on how to make things better there. Some multiplication for other localities. The Future Fund has 30 categories of ideas they want to pursue. Maybe that's "too small," but they're largely unaddressed and really big in scale. If they wanted to count all wins as equal, I don't doubt they could rack up a lot of very concrete wins and cool stories, but that seems to be what... all the rest of philanthropy is doing. And I'm glad they are! 

There's an undergrad econ thing where burning a dollar lowers the price level for all other dollar holders and increases their welfare, but everyone thinks you can do better than that by being more discerning. So just saying "more causes/ideas!" isn't really helpful without some limiting principle.

I took a development class in law school and thought the total focus on aid/politics/culture was a feature of it being in a law school. I guess not.

It seems like some discussion of s-risks is called for as they seem to be assumed away, though many longtermists are concerned about them. 

I was wondering if someone was looking into far UVC devices. I did briefly and it seems they're rare and maybe only available on a B2B basis. Also, I'd guess someone is currently working on a post about how EAGx Boston caused some higher-than-expected number of cases, so there's an update in favor of extra caution there. 

This interview with Obama Allan Defoe pointed to once is pretty instructive around these questions. On reflection, reasonable government actors see the case, it's just really hard to prioritize given short-run incentives ("maybe in 20 years the next generation will see things coming and deal with it").

My basic model is that government actors are all tied up by the stopping problem. If you ever want to do something good you need to make friends and win the next election. The voters and potential allies are even more short-termist than you. Availability bias explains why people would care about private nuclear weapons. Superintelligence codes as dinner party/dorm room chat. It will sell books, but it's not action relevant.  

Am I violating Reddiquette by advising people to browse the thread, use ctrl+F, and sort by new to find comments they might enjoy?