Wrong. See this: https://arxiv.org/abs/1612.03969 Look at Table 2: Row 7 (counting) and 14 (time reasoning) both get 0 error rates (this is on bAbI tasks). When you are fighting a rear-guard battle, it's best to know when you adversary overtook your rear 3 years ago.
Key problem with systems like GPT-2 is not that they dont deal with quantities (as @Ylecun suggests below), it is they don't develop robust representations of *how events unfold over time*
Clearest w number, but true in many cases, and it's part of why the quantity cases fail: https://twitter.com/ylecun/status/1188746007330222082 …pic.twitter.com/H2xICn37iZ
-
-
-
Run your system on my examples?
- 1 more reply
New conversation -
-
-
Unless I'm seriously misunderstanding the purpose of GPT2 (which I probably am), I was under the impression that it was just a text generator, like an advanced version of repeatedly hitting the predictive text feature on your phone. Personally, I have yet to see any use for it.
-
It's not terribly useful, but it's a good measure of where things are today. Because it has done a bit better on Winograd Schemas, some have inferred that it has some degree of common sense; I am trying to articulate where improvements are most urgently needed.
- 2 more replies
New conversation -
-
-
No. All of you are way overstating what can even be known at all. This is easy for a computer to make. What’s your theory to explain why it exists?pic.twitter.com/PZjdmrsMEI
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.