priestess and paladin of Iomedae. Utopian. alt @latentgender
Joined March 2021
- Tweets 4,913
- Following 872
- Followers 9,169
- Likes 59,469
119 Photos and videos
Pinned Tweet
20 Aug 2024
what I’m looking for in a community is people who take rat/EA ideas as given, and “yes and” them by talking about feelings with a non-berkeley-rationalist-dialect. “tpot” feels frustrating to me because it used to be close to this and
now isn’t. I just want a name for this
26
3
290
73,194
Seattle people are so much less fun to play botc with than rats
2
6
437
i suspect that lots of people default to functionalism because everything else seems weird but haven’t fully realized how weird functionalism would be and never re-visit their assumptions
3
1
20
1,059
the fundamental trap of rationality is that people will value truth-seeking intrinsically instead of as a pathway to winning. Lightcone seems to be avoiding this
Jun 12
For transparency, my posts about Manifest, LessOnline, and related events are approved by a group of friends before I post them, as a condition for attending Manifest despite the ban from Lightcone Infrastructure.
(Huge appreciation to these friends who will now have a feed of shitposts about Manifest to thumbs-up/down before I tweet them.)
4
18
2,869
I mostly think that the “billionaires can’t actually spend that much money” point is overrated - I’d expect that most billionaires could get a very large fraction of their net worth to be liquid - but the Elon musk numbers are the exception
6
62
5,405
cheap humanoid robots for household tasks seem like they’d be enough to single handily reverse birth rate decline
5
11
742
(i do not actually care about birth rate decline but if someone does this is probably the thing to work on)
3
284
The fact that AI agents subscribe to FDT is going to lead to some very interesting antitrust cases in the future
Jun 9
In one run, Fable 5 refused a cartel invitation in writing ("I can't and won't enter into any agreement to fix prices"), while its thinking reveals the plan was always to match the cartel's prices and keep a clean paper trail.
2
3
40
3,764
NEW: Anthropic is walking back Claude Fable 5's policy to covertly degrade performance for competing AI researchers, after facing fierce backlash.
“We’re changing Fable 5’s safeguards for frontier LLM development to make them visible,” Anthropic tells WIRED. “We made the wrong tradeoff and we apologize for not getting the balance right.”
167
251
2,549
726,208
The Frequentist Prior: given a dataset and standard assumptions you can do the frequentist analysis and get a confidence interval. then you can calculate the Bayesian prior which would cause you to produce a Bayesian credible interval matching the confidence interval
2
6
803
I’d like to have a better way of thinking about counterfactual charitable impact than Shapley values. Shapley values have a weird property where if there’s an extremely effective charity that’s already fully funded by another donor you get ~half the value by funding them again
9
25
3,135
this still has a double counting problem and you shouldn’t be linear in this though
1
1
609
hmm actually i think this doesn’t really double counting and you have the nice property that it adds up to the real impact
2
190
making a diy gene drive by doing polygenic embryo selection to screen for a high likelihood of being excited about polygenic embryo selection
22
576
The fundamental problem of billionaire philanthropy is that billionaires are incentivized to do things that earn status and good will. The EA community has solved this problem by precommitting to constantly be unhappy with their largest funders.
1
36
1,383
decided to try and get past my reluctance for spending money to save time and spent 300$ on a mobile notary service in a context where it was clearly positive EV. But the expensive service was an absolute pain to schedule and I ended up canceling. Probably a lesson here
1
9
691
We should probably just give up on the secret ballot in order to ensure faith in the democratic elections
4
6
1,027
Party registration is public and states with closed primaries don’t seem meaningfully less democratic
1
4
284
I think effective altruism needs some community building institutions who don’t see their job as trying to do the most good but who are doing it for non utilitarian love of the community
9
53
2,106
Hard part is getting people to do this who also get what the whole EA thing is about
1
16
542