[tyrick]

The subtitle of the article is "What Machine Learning Teaches Us About Ourselves"; This is backwards. Brain sciences inform ML (In fact, ML techniques are often coined after the biological counterpart). A result or finding in ML does not necessarily, or at all, imply anything for neuroscience.

Artificial neural networks do not teach us about biological neural networks, or 'Neuronal Networks', a term reluctantly used by a close neuroscientist for contradistinction. We don't need Google's cat research, but Hubel and Wiesel's cat research.

Let's see: Cheap reference to Kant, check. Vague parallel to the Sapir-Whorf hypothesis, check.

The 'intriguing' mapping that involves 3 ML terms is desperate.

This article appearing on the front page of HN shows how delusional some of today's ML lovers are with respect to neuroscience, the discipline that actually studies human brains.

[return0]

I wouldn't be so dismissive. The last time neuroscience informed neural networks was in the 1940s

[tyrick]

Frederick Jelinek, a researcher in natural language processing, has a funny quote, "Every time I fire a linguist, the performance of the speech recognizer goes up."

In general, I think a neuroscientist would be a distraction to any ML team. I don't mean to say that neuroscience is what drives ML insight, but if asked to pick which field influences the other most, my choice is clear.