|
|
|
|
|
by TaylorAlexander
1214 days ago
|
|
I agree with you, but to me the obvious answer is that this is unfinished research. An LLM is obviously going to be a useful part of a future information processing system, but it is not a terribly useful information processing system on its own. So invest in more research, secure rights to the future capabilities, and release something in the future that actually does what its supposed to do. I am listening to a podcast with Timnit Gebru now who is talking about coming up with tests you think your system should pass, just like running tests against your code. So if you think it can be used to suggest vacation plans, it had better do a good job giving you correct information. Otherwise you're just releasing something half baked, and it is hard for me to see the point in that. |
|