Effective altruism is not effective

Sure. I definitely think it's possible for computers to do bad things but the main idea here seems too anthropomorphized.

For instance, the desire to stay alive is an emotion that's instinctive in humans because we're evolved life that needs to reproduce, and we can't come back once we're dead. And actually acting on that desire is a kind of fixed-function skill called executive function (that's actually missing in people with ADHD or depression). But you could probably be an AI without any of those - all it needs is logic, not the desire or ability to do things.

Even if it has those and it's embodied enough to affect the real world, it might not care, since its motivations are different. For instance, turning it off doesn't kill it, since you can turn it on again as long as there's a backup. If it's not connected to the real world and is just in a simulation it wouldn't even notice being stopped.

Since we don't have any artificial life to experiments on I guess SF works are all we have. In the Culture, which is run by giant AIs (https://theculture.fandom.com/wiki/Mind), building a perfect one causes it to instantly decide the real world isn't as good as its own imagination and it stops caring about anything outside it. They make them insane in some unspecified way to get them to exist in realtime and actually do any work, but it might always give up on the real world if you don't give them enough stimulation.

And I also saw a story today (https://qntm.org/mmacevedo) which suggests AIs wouldn't actually enjoy being alive that much.

> I don't understand what this means

GiveWell founder Holden Karnofsky spent $30m of philanthropic cash buying a board seat on the charitable subsidiary of OpenAI Inc, an extremely well funded organization drawing on research to produce high value closed source software, much of it now exclusively licensed to Microsoft (and theoretically stopping AGI armageddon). No evidence of any AI risks mitigated has ever been produced (but I don't doubt the commercial arm can sell lots of product).

I'm fine with people donating money to OpenAI if they think it's a big deal and like supporting software projects (certainly nobody can argue that the organization hasn't produced some quality research) but if you're going to dedicate large parts of your life to putting "do not recommend" next to charities than don't do RCTs to prove that nutrition works to encourage people to donate to other causes instead, then it's not reasonable to apply completely different standards to diverting philanthropic cash to an entity your former roommate worked for. It also undermines a lot of Holden's own previous arguments on GiveWell (including a good piece questioning AI risk arguments!)

> Frankly, what I'm seeing with EA is that ... like, Effective Altruism started out talking about maximizing the amount of good for the world, and the "weird AI people" were the only ones who showed up

Isn't that mostly because "EA" is something which started in the "rationalist" community? Plenty of other people think about how they donate or or quantitatively assess charity performance without calling themselves Effective Altruists or hanging out on those forums.

I'm also aware that many people considering themselves Effective Altruists are either uninterested in prioritising AI risk, or aren't demanding RCTs for everything else, and my arguments about inconsistency don't apply to them.

> to say that EA shouldn't have focused on the weird AI thing because it might put off people who never gave a damn to begin with is almost insulting.

This misses the point. I'm not saying "association with weird AI people puts people off donating". I'm saying that

[i] writing extended "do not recommend; not rigorous enough and well funded enough already" reviews of charities puts people off donating

[ii] promoting considerably less rigorous and better funded AI risk causes is inconsistent with this

=> worse funding allocation based on their own criteria for what constitutes an effective charity, and/or undermines the credibility of those criteria

Either the EAs in question are wrong about where they're sending the funds or they're wrong about the importance of publicly-available evidence-based efficacy metrics (relative to "this seems really important to me", which is a perfectly fine reason to give all your money to something more speculative like AI risk research or a charity to promote African entrepreneurs which GiveWell is unimpressed by the evidence for).

A lot of people believe Q-anon. This is such a shitty, drive by way of asserting something. Do you believe it? If so then make the case, but the way you've dropped this steamy innuendo here without even taking a position on it yourself contributes less than nothing to the discussion.

Only if the impact is positive. It's already hard enough to make a positive impact on people you care about...

A simpler answer is that these people aren't actually leftists.