When the going gets weird the weird turn pro. Researching and writing on AI and other topics from a broadly EA perspective. On a grant from CG.
Washington DC
- Tweets 21,265
- Following 2,392
- Followers 18,497
- Likes 119,421
It is bad to anthropomorphize AI when the language implies it has abilities that it doesn't, and it's bad not to anthropomorphize AI when the language implies it doesn't have abilities that it does
7
9
159
4,105
It is funny bumping into people who will say NATO was designed to stop communism and expect you to be upset at that by default
5
4
169
3,231
My next post is trying to be a full and comprehensive case you can send people for why current models can "think" and "understand" and imo it's looking good
6
55
1,892
Are there any other contexts where people are asked to never consider or work on risks that could pop up in 5 years because they "distract from the real harms in the present"?
30
3
175
7,107
A core part of how I think about the world is that most of our beliefs about our own minds and thought are useful but wildly imperfect folk theories and heuristics that get at reality about as much as saying "the rock wants to move toward the ground." I'm always really shocked when people imply that our own accounts of the specialness of our minds and what they can do are so reliable that they should override any debate. The mind seems remarkably opaque when you actually start poking at it.
2
1
68
2,390
If you brought Astra or Opus 5.5 back in time to 1980, what percentage of people, after using it for a few weeks, do you think would say "Wow, you have successfully built a machine that can actually think and understand words"?
I think it'd be pretty high.
86
23
1,453
129,088
Yeah I think I have like 20 shows I like max. It's a bad medium.
the vast majority of TV is extremely bad in a way that is not the case with music, and it makes lists like those very weird to scan. A list of the 100 greatest albums will never have anything as dogshit as Ted Lasso, Homeland, or Stranger Things
11
3
53
19,946
I would ask before making proclamations like what Trevor's calling out here that you just ask the AI models if this criticism actually describes what's up in EA
Today I have seen multiple posts of the form "if the EAs and AI safety people were really serious about AI risks, they would be working on cybersecurity/model theft/constraining AI misbehavior, not alignment." These have in fact been major focuses of the leading organizations in the space for years! nitter.net/trevposts/status/21020…
24
2,337
Had this specific scene haunting my memory for about 15 years and the newest Claude finally figure out it's from Story Book Weaver Deluxe thank you Claude
10
1,063