"A quick tour of what you missed at the NeurIPS 2020 AI conference"

[dmm]

www.zdnet.com/google-amp/article/a-quick-tour-of-what-you-missed-at-the-neurips-2020-ai-conference/

Some of the things mentioned there are available without registration, e.g.

neurips.cc/virtual/2020/public/tutorial_877466ffd21fe26dd1b3366330b7b560.html

'Also Monday, a tutorial on reasoning brought together three complementary presentations by Francois Chollet and Christian Szegedy of Google, and Melanie Mitchell, the Davis Professor at the Santa Fe Institute.

Chollet made the case that abstraction is key to intelligence. [...]

[...]

"If you want to know how AI will operate in the future," said Chollet, "I think you should just look at how human engineers operate today."

[...] '

Comments

[juan_gandhi]

Abstraction capability is pretty rare in humans.

[dmm]

I think it's graded; basically, a continuous spectrum... People learn to use a complicated public transportation system in a city, with demands for navigation and planning, that already requires quite a bit of absraction.

I think even GPT-3 does some abstraction.

(But I should listen to what he is saying there; I've only read a ZDNet write-up so far.

I know that in the past I've agreed with some of his writings, and disagreed sharply with some of his other writings.)