It might lead to us being able to get AI datacenter level GPUs to run our own models on for relatively cheaply as ewaste. Wouldn’t that be cool?
It might lead to us being able to get AI datacenter level GPUs to run our own models on for relatively cheaply as ewaste. Wouldn’t that be cool?
That won’t happen. The chips themselves are useful for loads of industrial applications and the infrastructure itself is nontrivial.
Relatively cheap would still be tens of thousands of dollars.