Miles_Brundage

532Joined Jul 2022

Comments
16

I would have to think more on this to have a super confident reply. See also my point in response to Geoffrey Miller elsewhere here--there are lots of considerations at play. 

One view I hold, though, is something like "the optimal amount of self-censorship, by which I mean not always saying things that you think are true/useful, in part because you're considering the [personal/community-level] social implications thereof, is non-zero." We can of course disagree on the precise amount/contexts for this, and sometimes it can go too far. And by definition in all such cases you will think you are right and others wrong, so there is a cost. But I don't think it is automatically/definitionally bad for people to do that to some extent, and indeed much of progress on issues like civil rights, gay rights etc. in the US has resulted in large part from actions getting ahead of beliefs among people who didn't "get it" yet, with  cultural/ideological change gradually following with generational replacement, pop culture changes, etc. Obviously people rarely think that they are in the wrong, but it's hard to be sure, and I don't think we [the world, EA] should be aiming for a culture where there are never repercussions for expressing beliefs that, in the speaker's view, are true. Again, that's consistent with people disagreeing about particular cases, just sharing my general view here.

This shouldn't only work in one ideological "direction" of course, which may be a crux in how people react to the above. Some may see the philosophy above as (exclusively) an endorsement of wokism/cancel culture etc. in its entirety/current form [insofar as that were a coherent thing, which I'm not sure it is]. While I am probably less averse to some of those things than the some LW/EAF readers, especially on the rationalist side side, I also think that people should remember that restraint can be positive in many contexts. For example, I am, in my effort to engage and in my social media activities lately, trying to be careful to be respectful to people who identify strongly with the communities I am critiquing, and have held back some spicy jokes (e.g. playing on the "I like this statement and think it is true" line which just begs for memes), precisely because I want to avoid alienating people who might be receptive to the object level points I'm making, and because I don't want to unduly egg on critiques by other folks on social media who I think sometimes go too far in attacking EAs, etc.

FWIW despite having pretty diametrically opposed views on a lot of these things, I agree that there is something to the issue/divide you reference. It seems correlated with the "normie-EA vs. rationalist-EA" divide I mentioned elsewhere on this page, and I think there are potential tradeoffs from naively responding to the (IMO) real issues at stake on the other side of the ledger. How to non-naively navigate all this seems non-obvious.

I think it's a bit more nuanced than that + added some more detail on my views below.

Happy to comment on this, though I'll add a few caveats first:

- My views on priorities among the below are very unstable
- None of this is intended to imply/attribute malice or to demonize all rationalists ("many of my best friends/colleagues are rationalists"), or to imply that there aren't some upsides to the communities' overlap
- I am not sure what "institutional EA" should be doing about all this
- Since some of these are complex topics and ideally I'd want to cite lots of sources etc. in a detailed positive statement on them, I am using the "things to think about" framing. But hopefully this gives some flavor of my actual perspective while also pointing in fruitful directions for open-ended reflection. 
- I may be able to follow up on specific clarifying Qs though also am not sure how closely I'll follow replies, so try to get in touch with me offline if you're interested in further discussion.
- The upvoted comment is pretty long and I don't really want to get into line-by-line discussion of specific agreements/disagreements, so will focus on sharing my own model.

Those caveats aside, I think some things that EA-rationalists might want to think about in light of recent events are below. 

- Different senses of the word racism (~the "believing/stating that race is a 'real thing'/there are non-trivial differences between them (especially cognitive ones) that anyone should care about" definition, and the "consciously or unconsciously treating people better/worse given their race"), why some people think the former is bad/should be treated with extreme levels of skepticism and not just the latter, and whether there might be a finer line between them in practice than some think.
- Why the rationalist community seems to treat race/IQ as an area where one should defer to "the scientific consensus" but is quick to question the scientific community and attribute biases to it on a range of other topics like ivermectin/COVID generally,  AI safety, etc. 
- Whether the purported consensus folks often refer to is actually existent + what kind of interpretations/takeaways one might draw from specific results/papers other than literal racism in the first sense above (I recommend The Genetic Lottery's section on this).  
- What the information value of "more accurate [in the red pill/blackpill sense] views on race" would even be "if true," given that one never interacts with a distribution but with specific people.
- How Black people and other folks underrepresented in EA/rationalist communities, who often face multiple types of racism in the senses above, might react to seeing people in these communities speaking casually about all of this, and what implications that has for things like recruitment and retention in AI safety.

(will vaguely follow-up on this in my response to ESRogs's parallel comment) 

Note that there is now at least one post on LW front page that is at least indirectly about the Bostrom stuff. I am not sure if it was there before and I missed it or what.

And others' comments have updated me a bit towards the forum vs. forum difference being less surprising. 

I still think there is something like the above going on, though, as shown by the kinds of views being expressed + who's expressing them just on EA Forum, and on social media. 

But I probably should have left LW out of my  "argument" since I'm less familiar with typical patterns/norms there. 

Thanks for clarifying on the censorship point!

I do think it's pretty surprising and in-need-of-an-explanation that it isn't being discussed (much?) on LW - LW and EA Forum are often pretty correlated in terms of covering big "[EA/rationalist/longtermist] community news" like developments in AI, controversies related to famous people in one or more of those groups, etc. And it's hard to think of more than 1-2 people who are bigger deals in those communities than Bostrom (at most, arguably it's zero). So him being "cancelled" (something that's being covered in mainstream media) seems like a pretty obvious thing to discuss.

To be clear, I am not suggesting any malicious intent (e.g. "burying" something for reputational purposes), and I probably shouldn't have used the word censorship. If that's not what's going on, then yes, it's probably just that most LWers think it's no big deal. But that does line up with my view that there is a huge rationalist-EA vs. normie-EA divide, which I think people could agree with even if they lean more towards the other side of the divide than me.

Seeing the discussion play out here lately, and in parallel seeing the topic either not be brought up or be totally censored on LessWrong, has made the following more clear to me: 

A huge fraction of the EA community's reputational issues, DEI shortcomings, and internal strife stem from its proximity to/overlap with the rationalist community. 

Generalizing a lot,  it seems that "normie EAs" (IMO correctly) see glaring problems with Bostrom's statement and want this incident to serve as a teachable moment so the community can improve in some of the respects above, and "rationalist-EAs" want to debate race and IQ (or think that the issue is so minor/"wokeness-run-amok-y" that it should be ignored or censored). This predictably leads to conflict.

(I am sure many will take issue with this, but I suspect it will ring true/help clarify things for some, and if this isn't the time/place to discuss it, I don't know when/where that would be)

[Edit: I elaborated on various aspects of my views in the comments, though one could potentially agree with this comment/not all the below etc.]

The Aaron Swartz thing is not really the same kind of thing as the others, on various levels. See: https://en.wikipedia.org/wiki/United_States_v._Swartz 

A few aspects of my model: 

- Compute cost reduction is important for driving AI capabilities forward (among other things), and historically is mostly driven by things other than deployment of the most powerful systems (general semiconductor progress/learning curves, spillover from videogame related investments, deployment of systems other than the most powerful ones, e.g. machine translation, speech recognition, etc.). This may be changing as the share of NVIDIA's datacenter revenue increases and more companies deploy powerful LMs but for a long time this was the case. 

- Other drivers of AI progress such as investment in new algorithms via hiring people at top labs, algorithmic progress enabled by more compute, and increasing the amount of money spent on compute are only somewhat tied to deployment of the most powerful models. Again this may change over time but see e.g. all of DeepMind's value provided to Google via non-cutting-edge (or fairly domain specific) things like speech synthesis, as well as claims that it is a long-term investment rather than something which requires immediate revenue. Again these things may change over time but they have been true for some time and still have at least some truth. 

- Restraint is not all or nothing, e.g. given deployment of some system, it can be deployed more or less safely, there can be more or less alignment across companies on best practices, etc. And on the current margin, doing better w.r.t. safety is mostly bottlenecked by good ideas and people to execute on those ideas, rather than adjusting the "safety vs. speed" knob (though that's relevant to an extent, too). Given that situation, I think there is a lot of marginal additional restraint to be done without preventing deployment or otherwise significantly compromising lab interests (again, this could change eventually but right now I see plenty of opportunity to do "restraint-y" things that don't entail stopping all deployment). 

Load More