Gary Marcus

Another jewel from @GaryMarcus !Aside from the great technical content, I think an important issue is observed: contemporary AI papers use a "pretty good" specific result to make absolutely general and outrageous claims about a much broader subject.A bit modesty would be advised.

Great job. Poor popularization of AI is a major obstacle to progress. I raised the same warning in a French article two weeks ago. We need more specialists able to cross the domains with discernment. That said, the concept of innateness is fascinating. A pleasure to read !

Really enjoying these article/papers you've been writing. Rational criticism about AI techniques is crucial in making progress towards AGI.

Thank you for making it clear in this paper that there are two distinct approaches towards building cognitive architectures. The verdict is still out as to what will prevail. My bet is of course on intuition machines being the successful approach. https://deeplearningplaybook.com/ 

You’ve been busy writing lately, Gary ;-)

I forwarded Zero results to a control theorist; response: "I've always been mystified that they use human data in closed form, finite rule games." In Go, rules are known. What's the point of human data? Bias/initialize solver to humanlike play? checkers has analytic solution...

In a developmental robotics study, before training, one might specify which elements are assumed to be “innate”, which are previously learned, and which are learned through environment.