2 ms·
The ban makes sense given ML compute is dual-use, but I wonder how hard it would be to train say, a missile targeting system, on AWS. I wonder if any nationstat
by IceHegel 4y ago
The ban makes sense given ML compute is dual-use, but I wonder how hard it would be to train say, a missile targeting system, on AWS. I wonder if any nationstate would take the risk of using the public cloud for classified work as a way around sanctions.
- cjbgkagh 4y agoI don’t think it’s dual use for ML, I think it’s more about edge compute. Someone correct me if I’m wrong but I think radar resolution is compute bound so an order of magnitude or two could defeat stealth.