Gary Marcus

Ideally, you'd be saying this through NeurIPS submissions rather than New Yorker articles. A lot of the push-back you're getting right now is due to the perception that you haven't been using the appropriate channels to influence the field.

They also publish actual research though. I'm not saying Marcus nor Google/DeepMind/OpenAI/FAIR shouldn't talk to the media or the general public. But its hard to listen to complaints that the field isn't changing when you're not directly engaging with it.

If you want to change the direction of an academic field (he does), then you should be speaking the language of academics. Present novel results using symbolic systems. DL has been incorporating more forms of prior knowledge, its unclear to me that it needs to correct course now.

Addressed this yesterday; see tweet re why I don’t build AGI in a day.

Not asking for AGI in a day, just a concrete algorithm/model/system. Its easy for you to poke holes in existing DL systems, but its currently impossible for us to poke holes in yours. I'd be happy to add a symbolic baseline to my next paper, but one has to exist.