AI researchers allege that machine learning is alchemy

[♦ nivek ♦]

From Science 

AI researchers allege that machine learning is alchemy

 

Quote

 

By Matthew Hutson

May. 3, 2018 , 11:15 AM

 

Ali Rahimi, a researcher in artificial intelligence (AI) at Google in San Francisco, California, took a swipe at his field last December—and received a 40-second ovation for it. Speaking at an AI conference, Rahimi charged that machine learning algorithms, in which computers learn through trial and error, have become a form of "alchemy."

 

Researchers, he said, do not know why some algorithms work and others don't, nor do they have rigorous criteria for choosing one AI architecture over another. Now, in a paper presented on 30 April at the International Conference on Learning Representations in Vancouver, Canada, Rahimi and his collaborators document examples of what they see as the alchemy problem and offer prescriptions for bolstering AI's rigor.

"There's an anguish in the field," Rahimi says. "Many of us feel like we're operating on an alien technology."

 

The issue is distinct from AI's reproducibility problem, in which researchers can't replicate each other's results because of inconsistent experimental and publication practices. It also differs from the "black box" or "interpretability" problem in machine learning: the difficulty of explaining how a particular AI has come to its conclusions. As Rahimi puts it, "I'm trying to draw a distinction between a machine learning system that's a black box and an entire field that's become a black box."

 

 

..snipped

[Bhim]

He has a good point. I use machine learning a lot, and while the algorithms themselves are pretty well understood by most practitioners, it's true that there generally isn't a rigorous way to pick one over another for any given use case.

 

Having said that, "alchemy" may not have been the best choice of words given that this will be consumed by the general public. Actual alchemy doesn't work. Machine learning, on the other hand, is in fact a mathematically valid means of gaining insight into data or performing predictions when given data inputs. Calling it alchemy suggests (to some people) that the algorithms in question can't be rigorously stated, and that's not the case.

[pantheory]

I agree that Alchemy is not the appropriate word to describe the system of using sometimes indiscernible algorithms to teach AI entities. This could be called the Engineers approach to accomplishment. Try a great number of things and find out the one(s) that seem to work.  On the other hand the scientific approach might be stated as following theory by application. If an application is contrary to theory and/ or logic then drop it, "it's a form of Alchemy."  In reality the merits of both systems are debatable, but in the end each system has to be reconciled with the other for the advancement of both science and technology.