Aller au contenu
debya
Français
Se connecter Inscription gratuite
Voir sur Bluesky Suivre ce compte

Technical Debt Jubilee

@ianwsmith.bsky.social did:plc:dsgwsrxejun75rv5cy5ouisy

Low-volume provider of high quality, bespoke skeets. Data center, HPC, networking nerd. My ADHD makes me vacillate between filling this in with a mini resumé and things like 'aspiring llama rodeo clown'. Leftist, but not *that* kind.

Sur Bluesky depuis le

Abonnés
264
Abonnements
531
Posts
1 794
Jours dans les classements
0

Résumé

Technical Debt Jubilee (@ianwsmith.bsky.social) est sur Bluesky depuis le 7 février 2024. Le compte a 264 abonnés, suit 531 comptes et a publié 1 794 posts.

Présence dans les classements

Les meilleurs jours où les posts de ce compte sont entrés dans le top quotidien de Debya.

Ce compte n’est encore apparu dans aucun classement quotidien.

Posts aimés récemment

J’aime publics du réseau Bluesky.

  • Dominik Schnaus@schnaus.bsky.social

    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/unpaired-rosetta ⬇️

    Voir le post
  • r-buller.bsky.social@r-buller.bsky.social

    For an intro to the Platonic Representation Hypothesis: Would a vision model, and a text model, INDEPENDENTLY trained, converge on the same understanding of the world? Can you demonstrate "same understanding" empirically and measurably? Can you use that to join and translate? Yes, yes, & yes!

    Voir le post
  • alphaXiv@alphaxiv.org

    Yann LeCun and Randall Balestriero just Proved the Platonic Representation Hypothesis and found a better way to achieve it. Their text only JEPA got the same image-text alignment with 42x fewer training samples than NTP, using the same text data! Read more about it: www.alphaxiv.org/abs/2610.39302

    Voir le post
  • Zach Nation@reverius42.bsky.social

    Here’s an opinion that’s going to make everybody mad: local LLMs are like cars and hosted LLMs are like trains

    Voir le post
  • asura@asura.dev

    It's a very dumb thing to say Logistically centralizing all compute makes sense on paper, but so does taking everyone's car and house away to more efficiently align resources. Same. We are not in a natural scarcity of hardware and afaik Sama even bought unfinished chips just to increase demand

    Voir le post

Connectez-vous pour en voir plus

Comptes bloqués

Sur Bluesky, les blocages sont publics.