AMD's MI300 and the Future of Accelerated Compute
Oxide and Friends
00:00
Challenges of Training Large Models and Power Considerations
This chapter discusses the challenges faced in training large models and the need to consider power consumption. It explores the resource requirements for training billion or trillion parameter models and highlights the recent breakthroughs in efficient and smaller parameter models. The chapter also touches on the power bill of training models, comparisons to human power consumption, and the potential of AI in technical documentation and fact-checking.
Transcript
Play full episode
