I have this pang too, but for a different reason: everything is moving to the edge, training and inference. Who needs a complicated heterogeneous setup like this when you’re running everything locally?
The labs are betting on there being more applications for stronger models and that many existing applications will continue to need stronger models to keep up with competitors.
Yep. I keep thinking "this is good enough for almost anything" and then I remember how my Pentium 75 seemed like it was good enough for almost anything I could think of at the time. And it was!
The US controls most of the capital in the industry, including most of the compute used for training. And also, NVIDIA, arguably the most important actor in the industry
The US military can come and separate me from my MacBook and Qwen if it likes. I dare it. Sadly however, I suspect they’re busy with other matters at the moment.
reply