Hacker News
new
past
comments
ask
show
jobs
points
by
largbae
6 hours ago
|
comments
by
throwaway384034
6 hours ago
|
next
[-]
Sounds really bad for latency. Seems better to reason in the abstract or produce a mathematical model.
reply
by
WarmWash
4 hours ago
|
parent
|
next
[-]
Once a model gets good enough, etch it directly onto silicon and get 50,000 tokens per second.
reply
by
largbae
6 hours ago
|
parent
|
prev
|
[-]
Aye basically use it to make synthetic training data.
reply
by
throwaway384034
5 hours ago
|
parent
|
[-]
Training in a physics engine like MuJoCo with deterministic control seems more reliable.
reply
by
rheisen_
6 hours ago
|
prev
|
[-]
That’s.. actually insane. Just imagining a Claude operating a robot arm with a tool call that leveraged something like this to “imagine” scenarios before acting.
My god we are so cooked.
reply