Ali Rahimi's talk at NIPS(NIPS 2017 Test-of-time award presentation)

Channel:ObservantVids · Length:23:19 · Published:2017-12-06 · Views:78207

Watch on YouTube

Summary

Ali Rahimi's NIPS 2017 Test-of-Time award talk, in which he argues that modern machine learning has become alchemy, working without being rigorously understood. He calls for more rigor and understanding in the field.

Why HN recommends it

HN points to it as the sharpest, most-quoted critique of deep learning practice, and cites it whenever ML reliability or understanding is debated.

For a more cutting and insightful critique, watch Ali Rahimi's short talk at NIPS 2017 — HN, 2018

Machine learning & deep learning · Scientific method & science communication

Subtitles (excerpt)

good morning I think we are ready for the next talk if this this actually works so not every nips has a test of time award so it's a test of time even to get a test a time award this year has one and the criteria that we decided or the program she had decided on for for awarding it was ten year back nips papers so we formed a little committee and I must say that it was not easy all the ones that did not get their test of time award don't feel too bad this was not an easy decision but we did arrive we did arrive at a paper that we definitely felt was was was really really good and we're happy about it so just let me introduce it in the author's own word random features I trick to speed up supervised learning algorithms so they can operate on big that as millions of data point data says traditional learning algorithms work too hard because they optimized parameters that don't action need to be optimized instead one can randomize over some of the parameters and quickly optimize over the rest thanks to the concentration of measure of phenomenon this still provides excellent accuracy this recipe makes for accurate learning algorithms that run extremely fast and are very easy to implement so the paper

On Hacker News

This article is a little too glib in my opinion, preferring citations and statements to substance and explanations. For a more cutting and insightful critique, watch Ali Rahimi's short talk at NIPS 2017 (where he was presenting for a paper that won the "Test of time" award, for standing out in value a decade after publication). The standing ovation he received at the end indicate that his comment… — ssivark, 2018-02
Ali Rahimi's NIPS 2017 Test of Time award talk. "Artificial intelligence is not like electricity (quoting Andrew Ng), but alchemy" [VIDEO] — vowelless, 2018-01
>In no way can I believe that the corporation doesn’t have a good understanding of exactly how it can manipulate the AI systems it leverages to achieve its goals. See this talk which compares modern AI work to alchemy (i.e. it works, unreliably, and we don't always know why): [VIDEO] — ShredKazoo, 2023-02
I'm also an AI researcher and I'm not hopeful that the hype will die down anytime soon. People have been touting insignificant results through shoddy science for a long time now. They've noticed it works well enough because current ML is still pretty much alchemy (Ali Rahimi's 2017 NIPS Test of Time [1] talk still resonates today) so people rarely spend the effort to effectively refute the bogus … — nsagent, 2025-03

Related videos