
Perplexity published "web search" arena results for Sonar models. – Close top score with Gemini 2.5 Pro on search-related tasks
– Higher "Search depth" than anyone else

By
–

Perplexity published "web search" arena results for Sonar models. – Close top score with Gemini 2.5 Pro on search-related tasks
– Higher "Search depth" than anyone else
By
–
The GPU form factor is better today with current gate densities and memory architectures, but that wasn’t an option back in the day. A stronger case for vector architectures back then might even have held up progress on our current architectures!

By
–
Search Arena revealed three factors strongly correlating with human preference: • Longer responses
• Higher citation counts
• Citations from community sources Sonar models had substantially higher search depth, citing 2-3x more sources than equivalent Gemini models.
By
–
If you could afford a Cray, collecting 1.2 million digital images for ImageNet would not be a gating factor. DQN or AlphaGo didn’t need any data. The ChatGPT moment would have needed more data and compute, but the data would still be the lesser challenge.
By
–
In the future soon, we will be able to communicate with many intelligent animal species – can't wait to better understand what my dogis saying! Congrats to the DolphinGemma team building on our Gemma models – the most powerful single GPU/TPU open source models out there!

By
–
There is an alternate reality where Cray took their vector supercomputers, ditched FP64 calculations, and went with one FP32 pipe and a BF16 tensor core pipe. The same instruction set, memory architecture, and vector registers would have made a sweet deep learning machine, in
By
–
> Deep understanding of reality is intrinsically dual use. is a wonderful sentence (and broader insight).
By
–
This scenario is interesting. No mention of robotics but I really like the notes about the difficulty of training for “research taste.”
By
–
Une réflexion du PDG de Google. Traduction : C’est incroyable de penser que cela fait déjà 100 ans depuis la découverte de la mécanique quantique, qui a fondamentalement changé notre compréhension de la nature. Appliquer ces mêmes principes à l’informatique est peut-être le

By
–
What does real intelligence in AI look like? In his latest piece, http://
H2O.ai CEO @Sri Satish Ambati explores how new benchmarks like ARC focus on academic reasoning, while the General AI Assistants benchmark (GAIA) goes further—measuring how AI performs complex,