Introducing an approach to directly ground video generation models to policy execution without needing any action labels!
https://t.co/diDMvefjPa
Our approach uses a generic goal-conditioned exploration procedure to learn a policy that works across robots / embodiments!
Google announces Leave No Context Behind
Efficient Infinite Context Transformers with Infini-attention
This work introduces an efficient method to scale Transformer-based Large Language Models (LLMs) to infinitely long inputs with bounded memory and computation. A key
Researcher: “ChatGPT, write this paper for me.”
Reviewer: “ChatGPT, review this paper for me.”
Editor: “ChatGPT, assess the reviewers comments for me.”
#PhDVoice@PostdocVoice
Highly recommend this talk on EDM. There are a lot of diffusion tutorials online. Different ones cater to different levels of people. This one is particularly useful if you already have basic understanding about diffusion.
https://t.co/SiUSJkSiAW
object permanence in the new openai video model is pretty wild
the weird logo and text are untouched after the woman's head slowly blocks the whole sign