@Daniel Kokotajlo, what are the implications for the rewritten AI-2027 compute forecast? Does the loss of the HBM requirement mean that the main bottleneck on AI compute growth is advanced packaging? That agents can be scaled far more efficiently than you expected? Or were the new architectures priced into the forecast?
Thanks for the comment. Would absolutely love feedback from Daniel. I think that advanced packaging, specifically 3D packaging, certainly has the potential for a high impact on memory bandwidth as well as power dissipation. With that said, I think the efficiency problem in AI hardware is being approached from several angles, where HW/SW co-design and novel deployment configurations also dictate how the compute changes over time.
@Daniel Kokotajlo, what are the implications for the rewritten AI-2027 compute forecast? Does the loss of the HBM requirement mean that the main bottleneck on AI compute growth is advanced packaging? That agents can be scaled far more efficiently than you expected? Or were the new architectures priced into the forecast?
Thanks for the comment. Would absolutely love feedback from Daniel. I think that advanced packaging, specifically 3D packaging, certainly has the potential for a high impact on memory bandwidth as well as power dissipation. With that said, I think the efficiency problem in AI hardware is being approached from several angles, where HW/SW co-design and novel deployment configurations also dictate how the compute changes over time.