The Singularity Institute's Arrogance Problem

lukeprog

I intended Leveling Up in Rationality to communicate this:

Despite worries that extreme rationality isn't that great, I think there's reason to hope that it can be great if some other causal factors are flipped the right way (e.g. mastery over akrasia). Here are some detailed examples I can share because they're from my own life...

But some people seem to have read it and heard this instead:

I'm super-awesome. Don't you wish you were more like me? Yay rationality!

This failure (on my part) fits into a larger pattern of the Singularity Institute seeming too arrogant and (perhaps) being too arrogant. As one friend recently told me:

At least among Caltech undergrads and academic mathematicians, it's taboo to toot your own horn. In these worlds, one's achievements speak for themselves, so whether one is a Fields Medalist or a failure, one gains status purely passively, and must appear not to care about being smart or accomplished. I think because you and Eliezer don't have formal technical training, you don't instinctively grasp this taboo. Thus Eliezer's claim of world-class mathematical ability, in combination with his lack of technical publications, make it hard for a mathematician to take him seriously, because his social stance doesn't pattern-match to anything good. Eliezer's arrogance as evidence of technical cluelessness, was one of the reasons I didn't donate until I met [someone at SI in person]. So for instance, your boast that at SI discussions "everyone at the table knows and applies an insane amount of all the major sciences" would make any Caltech undergrad roll their eyes; your standard of an "insane amount" seems to be relative to the general population, not relative to actual scientists. And posting a list of powers you've acquired doesn't make anyone any more impressed than they already were, and isn't a high-status move.

So, I have a few questions:

What are the most egregious examples of SI's arrogance?
On which subjects and in which ways is SI too arrogant? Are there subjects and ways in which SI isn't arrogant enough?
What should SI do about this?

I intended Leveling Up in Rationality to communicate this:

Despite worries that extreme rationality isn't that great, I think there's reason to hope that it can be great if some other causal factors are flipped the right way (e.g. mastery over akrasia). Here are some detailed examples I can share because they're from my own life...

But some people seem to have read it and heard this instead:

I'm super-awesome. Don't you wish you were more like me? Yay rationality!

This failure (on my part) fits into a larger pattern of the Singularity Institute seeming too arrogant and (perhaps) being too arrogant. As one friend recently told me:

At least among Caltech undergrads and academic mathematicians, it's taboo to toot your own horn. In these worlds, one's achievements speak for themselves, so whether one is a Fields Medalist or a failure, one gains status purely passively, and must appear not to care about being smart or accomplished. I think because you and Eliezer don't have formal technical training, you don't instinctively grasp this taboo. Thus Eliezer's claim of world-class mathematical ability, in combination with his lack of technical publications, make it hard for a mathematician to take him seriously, because his social stance doesn't pattern-match to anything good. Eliezer's arrogance as evidence of technical cluelessness, was one of the reasons I didn't donate until I met [someone at SI in person]. So for instance, your boast that at SI discussions "everyone at the table knows and applies an insane amount of all the major sciences" would make any Caltech undergrad roll their eyes; your standard of an "insane amount" seems to be relative to the general population, not relative to actual scientists. And posting a list of powers you've acquired doesn't make anyone any more impressed than they already were, and isn't a high-status move.

So, I have a few questions:

What are the most egregious examples of SI's arrogance?
On which subjects and in which ways is SI too arrogant? Are there subjects and ways in which SI isn't arrogant enough?
What should SI do about this?

Thanks for clarifying what part of my statement you were objecting to.

Mostly what I was thinking of on that side was the idea that actually building a powerful AI, or even taking tangible steps that make the problem of building a powerful AI easier, would result in the destruction of the world (or, at best, the creation of various "failed utopias"), and therefore the moral thing to do (which most AI researchers, to say nothing of lesser mortals, aren't wise enough to realize is absolutely critical) is to hold off on that stuff and instead work on moral philosophy and decision theory.

I recall a long wave of exchanges of the form "Show us some code!" "You know, I could show you code... it's not that hard a problem, really, for one with the proper level of vampiric aura, once the one understands the powerful simplicity of the Bayes-structure of the entire universe and finds something to protect important enough to motivate the one to shut up and do the impossible. But it would be immoral for me to write AI code right now, because we haven't made enough progress in philosophy and decision theory to do it safely."

But looking at your clarification, I will admit I got sloppy in my formulation, given that that's only one example (albeit a pervasive one). What I should have said was "throughout the sequences EY frequently presents himself as possessing the intellectual horsepower and insight to transform the world in "impossible" ways, one obvious tangible expression of which (that is, actual AI design) he holds back from creating only because he possesses the unusual wisdom to realize that doing so is immoral."

"You know, I could show you code... it's not that hard a problem, really,

I'd actually be very surprised if Eliezer had ever said that - since it is plainly wrong and as far as I know Eliezer isn't quite that insane. I can imagine him saying that it is (probably) an order of magnitude easier than making the coded AI friendly but that is still just placing it simpler on a scale of 'impossible'. Eliezer says many things that qualify for the label arrogant but I doubt this is one of them.

If Eliezer thought AI wasn't a hard problem he wouldn't be comfortable dismissing (particular isntances of) AI researchers who don't care about friendliness as "Mostly Harmless"!

84

The Singularity Institute's Arrogance Problem

84

84

84

The Singularity Institute's Arrogance Problem

84

84