False! I wrote academic book in 2001 & journal article 1998 that foreshadowed all current issues & supplied significant empirical evidence that deep learning researchers still haven’t addressed; reraised in arXiv+NIPS. @LakeBrenden has replicated the work. Field has not engaged.https://twitter.com/Zergylord/status/1066344681330671616 …
-
The 1998 and 2001 pieces addressed the psych community (which had been engaged for decades). Your recent arXiv paper has over 100 citations -- it engaged the community, but as it was position/research paper its hard for that engagement to take the form of followup work.
Also published pointers to that work in Science in 1999, and many people then in neural net community took notice (and wrote replies), but people forget the issue because they didn’t have an answer and it didn’t fit their narrative. Problem still key....
-
-
It's such a catch 22, AI folks want you to give them a benchmark to hit, if you did, they'd hit it with a specialized DL solution and claim that you're wrong, but miss the entire point of the benchmark.
-
Then define a benchmark over a broader set of tasks, or define metrics based or data efficiency, or *something*. Asking for empirical evaluation isn't biasing the game towards DL.
- 6 more replies
New conversation -
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.