Effective Altruism Forum
EA Forum

Hide table of contents

Comment Permalink

Guy RavehJun 23 20222

I like the format.

The interesting ones IMO (meaning "the ones that may convey some kind of important truth") are 1, 6, 8, 17, 19. And maybe 10 but I can't see the thread so I don't know what it's saying.

See in context

20 Critiques of AI Safety That I Found on Twitter

by Daniel Kirmani

Jun 23 20221 min read 13

14

AI safetyBuilding effective altruismCriticism of effective altruismCriticism of effective altruist causesTwitter

Frontpage

20 Critiques of AI Safety That I Found on Twitter

Conclusions

13 comments

In no particular order, here's a collection of Twitter screenshots of people attacking AI Safety. A lot of them are poorly reasoned, and some of them are simply ad-hominem. Still, these types of tweets are influential, and are widely circulated among AI capabilities researchers.

1

2

3

4

5

(That one wasn't actually a critique, but it did convey useful information about the state of AI Safety's optics.)

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20 Conclusions

I originally intended to end this post with a call to action, but we mustn't propose solutions immediately. In lieu of a specific proposal, I ask you, can the optics of AI safety be improved?

14 Reactions

Comments13

Sorted by

New & upvoted

Click to highlight new comments since: Today at 7:18 PM

Asa Cooper SticklandJun 23 202229

"Still, these types of tweets are influential, and are widely circulated among AI capabilities researchers." I'm kind of skeptical of this.

Outside of Giada Pistilli and Talia Ringer I don't think these tweets would appear on the typical ML researcher timeline, they seem closer to niche rationality/EA shitposting.

Whether the typical ML person would think alignment/AI x-risk is really dumb is a different question, and I don't really know the answer to that one!

Asa Cooper SticklandJun 23 20223

Although to be clear it's still nice to have bunch of different critical perspectives! This post exposed me to some people I didn't know of.

Angelina LiJun 24 202212

As someone who is not really on Twitter, I found this an interesting thread to read through, thanks! :)

I'd enjoy reading periodic digests like this of "here's surveys of what people are saying about us, and some cultural context on who these people are". I do feel a bit lost as to who all of these people are, which I think would help me parse what is going on here a bit better.

Guy RavehJun 24 20221

Strongly agree.

TW123Jun 23 20228

I think it's essential to ask some questions first:

Why do people hold these views? (Is it just their personality, or did somebody in this community do something wrong?)
Is there any truth to these views? (As can be seen here, anti-AI safety views are quite varied. For example, many are attacks on the communities that care about them rather than the object-level issues.)
Does it even matter what these particular people think? (If not, then leave them be.)

Only then should one even consider engaging in outreach or efforts to improve optics.

ojorgensenJun 23 20223

Could someone explain the “e/acc” in some of these? I haven’t seen it before.

Guy RavehJun 23 202211

Neither have I, but judging by one of the tweets it stands for "effective acclerationist"? Which I guess means trying to get as much tech as possible and trusting [society? markets? individual users?] to deal effectively with any problem that comes up?

Elias SchmiedJun 26 20223

It's something that was recently invented on Twitter, here is the manifesto they wrote: https://swarthy.substack.com/p/effective-accelerationism-eacc?s=w
It's only believed by a couple people afaict, and unironically maybe by no one (although this doesn't make it unimportant!)

Guy RavehJun 26 20223

We expect e/acc to compile as “scary” for many EAs, although that’s not the goal. We think EA has a lack of focus and is missing an element of willingness to accept the terms of the deal in front of humanity — i.e. to be good stewards of a consciousness-friendly technocapital singularity or die trying.

Unlike EA, e/acc

Doesn’t advocate for modernist technocratic solutions to problems

Isn’t passively risk-averse in the same way as EAs that “wish everything would just slow down”

Isn’t human-centric — as long as it’s flourishing, consciousness is good

Isn’t in denial about how fast the future is coming

Rejects the desire for a panopticon implied by longtermist EA beliefs

Like EA, e/acc:

Is prescriptive

Values more positive valence consciousness as good

Values zero recognizable consciousness in the universe as the absolute worst outcome.

I agree with some of these allegedly not EA ideas and disagree wih some of the allegedly EA ones ("more positive valence consciousness = good"). But I'm not sure the actual manifesto has anything to do with any of these.

Guy RavehJun 24 20222

Abridged version of #10, as I understand it, after looking it up on Twitter: Pistilli was aware of the threat of superintelligente, but eventually chose to work on other important AI ethics problems unrelated to X-risk. She was repeatedly told that's negligent and irresponsible and felt very alienated by people in her field. Now she refuses to delve into sentience/AGI problems altogether, which seems like a loss.

The lesson is that the X-risk crowd needs to learn to play better with the other kids, who work on problems that will pop up in a world where we aren't all dead.

Guy RavehJun 23 20222

I like the format.

The interesting ones IMO (meaning "the ones that may convey some kind of important truth") are 1, 6, 8, 17, 19. And maybe 10 but I can't see the thread so I don't know what it's saying.

VerdenJun 25 20221

Okay, this is very off-topic, but I just really want more EAs know about a browser extension that has massively improved my Twitter experience.

https://github.com/insin/tweak-new-twitter/

Tweak New Twitter is a browser extension which removes algorithmic content from Twitter, hides news and trends, lets you control which shared tweets appear on your timeline, and adds other UI improvements

Madhav MalhotraJun 25 20221

I also find the screenshotted post in #7 problematic: "Once AGI is so close to being developed that there's no longer sufficient time for movement building or public education to help with AI safety, I guess I can go on holiday and just enoy the final few months of my life."

I'd be doubtful that official AI safety organisations or their representatives would communicate similarly. But a good takeaway in general is to not promote content on global priorities that insinuates a sense of powerlessness.