Qwen 3.5: small models with impressive performancetwitter.com6 points·by moondistance·4 mesi fa·0 comments
Language Models Are Injective and Hence Invertiblearxiv.org1 points·by moondistance·9 mesi fa·2 comments
DeepSeek-R1 at 3,872 tokens / second on a single Nvidia HGX H200blogs.nvidia.com13 points·by moondistance·anno scorso·1 comments
ByteDance Doubao-1.5-pro matches GPT 4o benchmarks at 50x cheapertwitter.com4 points·by moondistance·anno scorso·0 comments
US-China Commission top recommendation: Manhattan project for race to AGI [pdf]uscc.gov4 points·by moondistance·2 anni fa·0 comments
moondistance·11 mesi fa·discussHow does this compare with Exowatt? Assume fresnel lenses are more efficient.
moondistance·anno scorso·discussChris Crawford is also famous for the Dragon Speech :) https://youtu.be/CBrj4S24074?si=Ph12RpW8BKsh8-qS
moondistance·anno scorso·discussYes, big time, but there continues to be lots of progress.Most importantly, models are maturing, and this means less custom optimization is required.
moondistance·anno scorso·discuss(this is also why AMD popped 10% at open yesterday - this is a new development and talks from their 2025 "Advancing AI" event were published late last week + over the weekend)
moondistance·anno scorso·discussYes, for many applications.Meta, OpenAI, Crusoe, and xAI recently announced large purchases of MI300 chips for inference.MI400, which will be available next year, also looks to be at least on par with Nvidia's roadmap.
moondistance·anno scorso·discussWhere do you find/buy land? How do you vet purchases? Can you point to a few websites, etc.? Thanks!
moondistance·2 anni fa·discussHappy New Year!I know so few of you but feel like you’re my best friends. Love to all.
moondistance·2 anni fa·discussMicrosoft is negotiating equity in OpenAI as part of the switch to a for-profit. Non-zero chance this is a negotiation flex.