У кого охлаждение интереса к GenAI, а у кого и сид раунды на миллиард:

https://www.reuters.com/technology/artificial-intelligence/openai-co-founder-sutskevers-new-safety-focused-ai-startup-ssi-raises-1-billion-2024-09-04/

И в продолжение банкета. Не могу это развидеть

https://www.facebook.com/share/v/NeWF1pCNzJb8fW3f/

Here are my slides from today's talk at Datafest Yerevan.

The talk was about non-transformer architectures, e.g., good old MLPs, CNNs, RNNs, and brand-new SSMs. It may be too dense with too many model names, but I think it may be useful as a reference for further exploration.

https://docs.google.com/presentation/d/19jpt6sSScUb1yKnlO3a47SsMRIL7UmqQZKkuADyI7nM/edit#slide=id.g2f6fb83b821_0_15

Интересная статистика сбоев в распределенной системе для обучения большой LLM. Из работы про Llama 3 (https://ai.meta.com/research/publications/the-llama-3-herd-of-models/)

Ризонинг подвезли!

https://openai.com/index/learning-to-reason-with-llms/