Figure/Ground Illusions · Read Scott Alexander

There’s a social justice concept called “distress of the privileged”. It means that if some privileged group is used to having things 100% their own way, and then some reform means that they only get things 99% their own way, this feels from the inside like oppression, like the system is biased against them, like now the other groups have it 100% their own way and they have it 0% and they can’t understand why everyone else is being so unfair.

I’ve said before that I think a lot of these sorts of ideas are poor fits for the one-sided issues they’re generally applied to, but more often accurate in describing the smaller, more heavily contested ideological issues where most of the explicit disputes lie nowadays. And so there’s an equivalent to distress of the privileged where supporters of a popular ideology think anything that’s equally fair to popular and unpopular ideologies, or even biased toward the popular ideology less than everyone else, is a 100%-against-them super-partisan tool of the unpopular people.

So I want to go back to Dylan Matthews’ article about EA. He is concerned that there’s too much focus on existential risk in the movement, writing:

Effective altruism is becoming more and more about funding computer science research to forestall an artificial intelligence–provoked apocalypse.

And:

EA Global was dominated by talk of existential risks, or X-risks.

And:

What was most concerning was the vehemence with which AI worriers asserted the cause’s priority over other cause areas.

And:

The movement has a very real demographic problem, which contributes to very real intellectual blinders of the kind that give rise to the AI obsession.

It sounds like he worries AI concerns are taking over the movement, that they’ve become the dominant strain, that all anybody’s interested in is AI.

Here is the latest effective altruist survey. This survey massively overestimates concern with AI risks, because only the AI risk sites did a good job publicizing the survey. Nevertheless, it still finds that of 813 effective altruists, only 77 donated to the main AI risk charity listed, the Machine Intelligence Research Institute. In comparison, 211 – almost three times as many – donated to the Against Malaria Foundation (note that not all participants donated to any cause, and some may have donated to several)

An explicit question about areas of concern tells a similar story – out of ten multiple-choice areas of concern, AI risks, x-risks, and the far future are 5th, 7th, and last respectively. The top is, once again, global poverty.

I wasn’t at the EA Summit and can’t talk about it from a position of personal knowledge. But the program suggests that out of thirty or so different events, just one was explicitly about AI, and two others were more generically x-risk related. The numbers at the other two EA summits were even less impressive. In Melbourne, there was only one item related to AI or x-risk – putting it on equal footing with the “Christianity And Effective Altruism” talk.

I do hear that the Bay Area AI event got special billing, but I think this was less because only AI is important, and more because some awesome people like Elon Musk were speaking, whereas a lot of the other panels featured people so non-famous that they even very briefly flirted with trying to involve me.

And – when people say that you should donate all of your money to AI risk and none to any other cause, they may well be thinking in terms of a world where about $50 billion is donated to global poverty yearly, and by my estimates the total budget for AI risk is less than $5 million a year. There are world-spanning NGOs like UNICEF and the World Bank working on global poverty and employing tens of thousands of people; in contrast, I bet > 10% of living AI risk researchers have been to one of Alicorn’s weekly dinner parties, and her table is only big enough for six people at a time. In this context, on the margin, “you should make your donation to AI” means “I think AI should get more than 1/10,000th of the pot”.

I suspect that “AI is dominating the effective altruist movement”, when you look at it, means “AI is given an equal place at the effective altruist table, compared to being totally marginalized everywhere else.” By figure-ground illusion, that makes it seem “dominant”.

Or consider me personally. I probably sound like some kind of huge AI partisan by this point, but I give less than a third of my donations to AI related causes, and if you ask me whether you should donate to them, I will tell you that I honestly don’t know. The only reason I keep speaking out in favor of AI risks is that when everyone else is so sure about it, my “I don’t know” suddenly becomes a far-fringe position that requires defending more than less controversial things. By figure-ground illusion, that makes me seem super-pro-AI.

In much the same way, I have gotten many complaints that the comments section of this blog leans way way way to the right, whereas the survey (WHICH I WILL ONE DAY POST, HONEST) suggests that it is almost perfectly evenly balanced. I can’t prove that the median survey-taker is also the median commenter, but I think probably people used to discussions entirely dominated by the left are seeing an illusory conservative bias in a place where both sides are finally talking equally.

Less measurably, I think I get this with my own views: – I despair of ever shaking the label of “neoreactionary sympathizer” just for treating them with about the same level of respect and intellectual interest I treat everyone else. And I despair of ever shaking the label of “violently obsessively anti-social-justice guy” – despite a bunch of posts expressing cautious support for social justice causes – just because I’m not willing to give them a total free pass when they do something awful, or totally demonize their enemies, in the same way as the median person I see on Facebook.

Or at least this is how it feels from the inside. Maybe this is how everybody feels from the inside, and Ayatollah Khameini is sitting in Tehran saying “I am so confused by everything that I try to mostly maintain an intellectual neutrality in which I give Islam exactly equal time to every other religion, but everyone else is unfairly hostile to it so I concentrate on that one, and then people call me a fanatic.” It doesn’t seem likely. But I guess it’s possible.