At NeurIPS 2020, researchers proposed faster, more efficient alternatives to backpropagation

Recommendable, but the headline clearly over promises! Most of the papers cited in this article are not yet accepted at any major conference.

By the way, if you want to know why the prestigious, long standing NIPS conference was renamed to NeurIPS in 2018, see my blog post here.

However, it is well worth to question one of the major methods of machine learning, i.e. backpropagation.

The article cites an old quote by Geoffrey Hinton:
"... In 2017, Geoffrey Hinton, a researcher at the University of Toronto and Google’s AI research division and a winner of the Association for Computing Machinery’s Turing Award, told ... in an interview that he was “deeply suspicious” of deep learning. “My view is throw it all away and start again,” he said. “I don’t think that’s how the brain works.”

I will say that the human brain has evolved evolutionary over millions of years. If we develop some form of synthetic intelligence, humans may be able to develop this much faster than nature ever could and it may not be similar to natural intelligence and it may even be superior in the end.

At NeurIPS 2020, researchers proposed faster, more efficient alternatives to backpropagation | VentureBeat