Pinned Tweet
new blog post! i explore a symbolic reasoning method (EMPA) and test it out on ARC-AGI-3, give my opinions on it and on the future of such methods in world modelling, check out the link below for results and discussions on the same
5
7
54
3,810
I finally watched the AlphaGo documentary. more than anything else I came away with a deep respect for Fan Hui
I finally watched the AlphaGo documentary. more than anything else I came away with a deep respect for Lee Sedol
2
8
351
to go through what fan did and still go onto advising the team, now that is what is actually the differentiator
2
14
weeye retweeted
wrote a technical blog on how to train nanogpt2 in moe style without a load balancer. do give it a read. happy sunday you all! now i should log out, rest and sip some matcha.
3
12
29
592
i think with all the php and emudra stuff it is a reminder call to shut down the web once and for all, i have some backup plans but the question is do you!?
7
204
weeye retweeted
am sick so spent the night with opus 5.5 working on a chart (inspired by the one from lem's golem xiv) how did we got here / model genealogy / the heroic age of language modeling
20
48
502
21,941
yes i work as a "slop refactorer" cleaning up all the slop in progress so your chud mind can comprehend it
7
259
pretty wild when you think we changed our perceptions about mech interp and flying cars both in just about 24 hours
1
16
558
weeye retweeted
Interpretability blogs are the recipes blogs of ML. Not sure, why I have to read through the story of your grandmother's first house to find out you used gradients.
22
41
1,074
44,998
oh look how far we've come
Anthropic released their 2021 research roadmap earlier this year (because of copyright litigation); I think some people might find it interesting: i.sm4.ca/concise-research-ro…
2
155
billions must picture-maxx (added 3 more figures in my draft after reading this btw)
5
150
i think i might be getting the infamous llm psychosis finally, what does one do to counteract these symptoms...
3
5
372
i agree a lot with this take spiritually, this paper was my first introduction to such an intersection (pretty easy to follow through too), of course there are far better works in this regard but this one makes you go through the motions: > start reading through it > realise you dont know certain terms (eg. price system, relative scarcity, equilibrium, mutually compatible plans etc) > your fav llm expands on it, you start a temp econ arc surfing through more papers > you give up thinking it is too dense and requires more time (just don't btw and get a phd)
Replying to @willcb
econ theorists also likely have some of the best intuitions about managing agent swarms we're now wielding armies of boundedly-rational partially-aligned agents with communication and transaction costs at mind-boggling levels of abstraction they've been doing this for 50 years
2
2
12
464
reminds me of the 1000 pending papers saved from such a run and of the blog i promised to write
27
New post: why the life of the mind is not purely private, the importance of intellectual community, and how AI undermines these communities.
12
29
277
16,827
grey goo hall of fame
You asked, we delivered. The @JoinFAI Field Guide to the Tribes of Silicon Valley is live now. 27 tribes, a few easter eggs, and a 3-minute quiz (just for fun). tribes.thefai.org
1
13
322
ruined morning
15
weeye retweeted
Replying to @0xdiicell
i have a theory that as the year progresses everyone just gets super busy until the holidays hit and they realise half of it could be avoided and used for better activities
1
3
47
we need more santa believers after the sudden demise of many of them last night
3
116
in the paper writing mines, we produce no slop in any form whatsoever
2
13
238
more results from the trenches, also i think idealens is a pretty neat idea :)
20