Gary Marcus

@Ylecun Now that you have joined the symbol-manipulating club, I challenge you to read my arxiv article Deep Learning: Critical Appraisal carefully and tell me what I actually say there that you disagree with. It might be a lot less than you think.https://twitter.com/ylecun/status/1066418378070507520 …

What I don't quite understand is where your (possibly waning) hostility towards deep learning comes from when we (indeed) don't disagree that much on the way forward, and we (now apparently) agree that deep learning is part of the solution.

My hostility is not to deep learning, it’s to deep learning alone. I have been clear about the distinction for 26 years, and pointed you numerous times to places where I have clarified this, typically in > 140 characters. [repeating inside thread, by request]

Deep learning alone can do a lot of things (in NLP, CV, etc.). tbh it seems you do a lot of pointing to cases where it alone is not sufficient & say "see, Deep Learning won't get us to true AI" when no one was saying that it will in the first place (except non-experts).

Granted, there are exceptions & various levels of hype around Deep Learning (*cough* @deeplearningai_ *cough*), but I think by far the majority of researchers & practitioners see it as a tool for powerful pattern recognition, but don't predict that it alone will get us to AGI.

That’s all I am saying, but some people don’t like me saying it, and try to undermine my credentials to keep me from saying it. Go back to 2015 and you will see a lot of people more optimistic about dl alone than I was. Maybe times have changed; that was how this thread started.

True, there was quite a bit of silly hype around DL, especially for those with less expertise (maybe mostly - even as an undergrad I knew fancy big neural nets won't get us to AGI)... My impression is it's getting better now, seems AI is on down slope of Gartner's hype cycle.

This whole thread got started when I pointed out that the narrative was changing.