#163 - Llama 3, Grok-1.5 Vision, new Atlas robot, RHO-1, Medium ban
Last Week in AI
00:00
Advancements in Embodied Question Answering
This chapter explores the groundbreaking initiative from Meta's FAIR lab focusing on embodied question answering using foundation models and introduces the Open Vocalibri benchmark dataset designed for robots. It discusses the challenges of different EQA approaches and highlights innovations like the IDFX2 vision language model and selective language modeling techniques to improve AI efficiency. Furthermore, it examines novel training methods for large language models, the performance of mixture of experts in neural models, and new AI hardware developments that could redefine technology in this area.
Transcript
Play full episode
