7 years of AlexNet paper

[dmm]

It was presented on Dec 4, 2012 at NIPS.

en.wikipedia.org/wiki/AlexNet

To the extent that a single event can be considered a start of the "Deep Learning Revolution", this one is it.

Это, конечно, некоторая условность:

www.eff.org/ai/metrics

Но, тем ни менее, "технологическая сингулярность" в декабре 2012-го года не произошла, но "произошла" "Deep Learning Revolution".

Можно про это думать, как про "фазовый переход второго рода" (вроде возникновения сверхтекучей компоненты в жидком гелии), вместо "фазового перехода первого рода" (вроде таяния льда)...

Comments

[dmm]

And then Charles Stross, who is his political opposite, has these deep things to say today (starting with a non-trivial thesis that a corporation is a slow procedural very old-fashioned AI, which is the thesis he was pushing for some time now):

https://www.antipope.org/charlie/blog-static/2019/12/artificial-intelligence-threat.html

(He does talk "conventional, muggle nonsense" about early history of neural nets - I did expects better of him, given his reputation and track record.

Otherwise, some of his remarks are very interesting; however, he is not too up-to-date with the trends in the AI community (e.g. with the current push towards more explainable AI). So it's a mixed bag.)