Hacker News
new
|
past
|
comments
|
ask
|
show
|
jobs
|
submit
login
doubtfuluser
on May 12, 2022
|
parent
|
context
|
favorite
| on:
DeepMind: A Generalist Agent
Well… straightforward in a way, yes. But the scale of learning is huge especially with this diverse set of tasks. Not totally unexpected, but certainly not clear that it would work with current networks and sizes.
f38zf5vdt
on May 12, 2022
[–]
Right, exactly. Something that seemed like it should work but no one had ever tried it.
Guidelines
|
FAQ
|
Lists
|
API
|
Security
|
Legal
|
Apply to YC
|
Contact
Search:
