@nasqret Wiles took 7 years and 108 pages to conquer Fermat, yet still hit a logic gap that cost another year to patch. The sheer endurance required is massively underrated ...
Je te rejoins sur le contournement mais pas sur le forcément pareil pour le texte .. anthropic utilise une variante de SynthID Text et rapporte n'avoir observé aucun impact pratique sur la qualité. Car le watermark n’est pas une dégradation ajoutée après génération : il intervient dans le sampling …
@sama NS is impressive...
Collatz is the problem Erdős called "absolutely hopeless" the one where "mathematics may not be ready."
Any genuine progress.. That's the eval I want to see.
For my first post, I’m sharing a letter @NVIDIA signed on why open models matter.
AI will transform every industry, power every company, and be built by every country.
Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty.
The world needs both frontier closed models and frontier open models.
https://t.co/AUKzoQ5Ikb
@JFPuget I vaguely remember seeing some posts from you about EU–US politics. It’s a topic where people tend to be highly polarized and hold very strong opinions …
@avsm@kc_srk it represented the culmination of eight years of R&D... Success now depends on the ecosystem adapting to the new concurrent GC and memory model
@CUDAHandbook O(1) is optimal assumes free memory access and large N, but in embedded systems arrays often beat hash maps due to zero setup, contiguous memory and also cheaper integer compares than hash computation ...
@ID_AA_Carmack Fun fact: This launched in 2021, shortly after Microsoft acquired ZeniMax.. The community joked that the acquisition was necessary just to secure the legal naming rights for "Quake Mode" 😃
@nasqret Wiles took 7 years and 108 pages to conquer Fermat, yet still hit a logic gap that cost another year to patch. The sheer endurance required is massively underrated ...