3 ms·
It's Gary Marcus, again, as always and as ever, criticizing other people's work as "machines that manipulate data but aren't really intelligent." He's been on
by cs702 4y ago
It's Gary Marcus, again, as always and as ever, criticizing other people's work as "machines that manipulate data but aren't really intelligent."
He's been on HN many times before, always criticizing the same things:
https://hn.algolia.com/?dateRange=all&page=0&prefix=true&query=gary%20marcus&sort=byPopularity&type=all https://hn.algolia.com/?dateRange=all&page=0&prefix=true&que...
As far as I know, all he's ever done is criticize, without ever delving into the mathematical details.
To understand those who disagree with him, read "The Bitter Lesson" by Rich Sutton:
http://incompleteideas.net/IncIdeas/BitterLesson.html http://incompleteideas.net/IncIdeas/BitterLesson.html
--
EDITS: Modified and rearranged sentences to reflect more accurately what I meant to write the first time around.
- area51org 4y agoDoes he need to develop an AI system at all in order to have a valid opinion, though?
- cs702 4y agoNo, he doesn't. I updated my comment to reflect what I actually meant to say.
- Der_Einzige 4y agoHonestly? Yes, given the amount of trust reporters put into his statements.
- sizzle 4y agoDo movie critics need to produce box office films before criticizing movies?
- dimmuborgir 4y agoHis book "The Algebraic Mind" goes into great detail about connectionism (neural networks), symbolic systems, limits of connectionism and proposals to integrate neural networks with symbolic systems (hybrid systems). The "deep learning alone is enough" camp (especially LeCun) has abused him for years but now slowly coming to the realization that we need to feed neural networks with explicit inductive biases to attain AGI which is exactly what Marcus has been saying since the 90s. LeCun, for some reason, refuses to call these explicit biases as symbols and that's the only disagreement between Marcus and LeCun these days.