https://xitter.cf/t.co/fkZM93M2BL https://xitter.cf/t.co/WkwQdiSXr6

Joined November 2021
datacenters.m19182.dev/ bastante slop pero la voy manteniendo al día en españa, hay al menos un market intelligence provider que ya saca data de aquí;) video unrelated
1
2
112
y si veis algún error o algo que me falte decidme 👀
12
im sorry but if claude doesn't refuse to work on your research project it's not cool enough. i don't make the rules
21
14
5
402
27,761
🙂‍↕️🙂‍↕️ @maisaAI_
4
12
777
sometimes pca to rgb really hits
3
24
1,271
There is an (unpatchable?) supply-chain zero-day in PyTorch autograd that allows a regularization term to use the model's weights as RAM and SGD as its CPU clock to run any computation during training.
88
216
120
2,397
165,740
consciousness-as-side-effect-of-information-processing believers in shambles
Replying to @doristsao
6/ Remarkably, when we reconstructed the image using neural activity from one of the states, it looked exactly like the conscious face, and when we did the same from the other half of time, it looked exactly like the suppressed face. This means that just from neural activity, we can reconstruct in rich detail not only something that you consciously perceive, but also something that escapes your awareness.
19
29
830
57,205
as cool as they are, HFT firms are probably net negative for socienty. they are playing a zero-sum game where the losers are in part retail, and price discovery doesn't seem like enought of an excuse for this.
2
1
105
the problem is not only the speed arms race, that something like frequent batch auction could solve, but also the opportunity cost of all that talent and resources. hfts do a lot of things thought and some of them are good, is just that the aggregate seems negative to me
1
1
83
White boy swag peaked on the early 90s, it only goes downhill from here…
1
91
Colaiuta might be the best to ever do it, casually playing the most difficult patterns while perma smoking
2
39
DINOv2 has never seen a caption, and Qwen3 has never seen an image. We still aligned their embedding spaces without a single image-caption pair. It even works when the images and the captions come from different datasets. Project page: dominik-schnaus.github.io/un… ⬇️
85
359
85
2,408
243,303
It kind of smells like we want an arXiv for fast iteration, community driven science Oriented around projects not user accounts like X Something with a nice search, good algo, markdown, git, and social aspect who's building this
4
3
17
723
creo que en la historia hay mucho margen para que ocurra algo parecido a las matemáticas con la IA. cualquiera con la suficiente curiosidad puede hacer descubrimientos originales estos días resolví un reto del padre Feijoo de 1729 +
2
1
6
255
desafió a sus lectores a encontrar líneas traducidas de revistas francesas en sus libros (estaba metido en una polémica al respecto). casi tres siglos después, encontré unos 10 casos que cumplen sus condiciones! blog.m19182.dev/writings/El-…
1
1
4
306
me encantaría que más gente explorase este terreno, hay muchos misterios históricos cuya respuesta está enterrada en archivos, algunos de ellos ya digitalizados. y aprendí un montón de literatura del siglo XVIII de paso...
1
2
64
good times, used to have rotate on 5 google accounts to experiment with styleGAN
Remember when google collab would give you gpu time for free?
1
1
107
In 1956, Cybernetics was renamed "AI" by logic programming experts. Now Trump wants to rename AI again. Let's revive the old name! Modern AI is about neural nets & deep learning, closer to Cybernetics than to 1956 AI. Plus, Cybernetics sounds better than Artificial Intelligence.
93
150
58
1,194
168,682