Even my laptop can generate text at a much higher rate than me. It's not clear how much the compute costs of the necessary improvements in inference can be compensated by better algorithms, but it may well turn out that H100s are seriously superhuman in practical applications?
COMPUTING
-
Openness and Computational Irreducibility: Epistemic Equivalence
By
–
Openness seems epistemically indistinguishable from computational irreducibility?
-
Service hits rate limits, team working on fixes
By
–
We just hit rate limits! Sorry about that, working on fixing
-
API Rate Limit Exceeded: Service Reaching Maximum Capacity
By
–
Everyone! So this already blew up and we're hitting the rate limit! Reaching out to get us unlocked.
-
Defining Purely Algorithmic Systems: Penrose and Computational Theory
By
–
How do you define a 'purely algorithmic system'? In the sense that Penrose uses it, it's equivalent to computational (constructive math, operations on representations via successive state transitions)
-
Google secretly used an AI to optimize itself internally
By
–
Google a gardé ça secret pendant des mois.
— VISION IA (@vision_ia) 27 mai 2025
Une IA en interne – nom de code Alpha Evolve – a optimisé ses puces, restructuré ses serveurs et influencé ses décisions stratégiques.
Pas une démo. Une IA déjà déployée, déjà active.
L'IA qui s’auto-améliorent.
Vous réalisez ce que… pic.twitter.com/mIzFHU0pZRGoogle kept this secret for months. An internal AI – codenamed Alpha Evolve – optimized its chips, restructured its servers, and influenced its strategic decisions. Not a demo. An AI already deployed, already active.
The self-improving AI. Do you realize what -
Pioneer 10 Challenge: Building the Future of Edge AI
By
–
The Pioneer 10 Challenge is live – and we’re looking for innovators ready to build the future of edge AI. Are you working on solutions in security, retail, robotics, healthcare, or any other edge domain? This is your opportunity to turn ideas into impact using the Metis AI
-
Distributed Algorithm Optimization: Speed Theory vs Empirical Reality
By
–
Not sure, just putting it out there as an option people might not have thought of. It’s faster theoretically, but I’m not sure this is true empirically, specially in distributed settings. I’m thinking about it, but maybe others can do the thinking better and faster here For
-
NVIDIA Unveils DGX Park: A Personal Desktop AI Supercomputer
By
–
🇫🇷 Traduit – Jensen Huang
— VISION IA (@vision_ia) 27 mai 2025
NVIDIA lance la DGX PARK : un supercalculateur d’IA personnel. 1 pétaflop. 128 Go. À poser sur votre bureau.
L’équivalent de la DGX-1 de 2016 (136 kg) — réduit à la taille d’un PC grand public.
C’est la fin du cloud pour les développeurs… et le début… https://t.co/KVZYqDoA27 pic.twitter.com/eAtYWhOym6NVIDIA launches the DGX PARK: a personal AI supercomputer. 1 petaflop. 128GB. To place on your desk. The equivalent of the 2016 DGX-1 (136 kg) — shrunk down to the size of a consumer PC. This is the end of the cloud for developers… and the beginning