Working Up

by Dwayne Phillips

New deep learning tools have turned a technology field into a general-purpose programming language.

I worked with neural networks 25 years ago (yes, I am that old). Recently, big, big neural networks have returned to favor as deep learning. Clever techniques have reduced this to a programming language instead of an area of technology.

For example,

FORTRAN: crunch numbers in banking, image processing, signal processing, etc. There are many different things you can do with the language.

Deep Learning: “learn the computer” to recognize things in images, understand speech, recognize faces, understand text, play chess, etc. There are many different things you can do once your have your deep learning system created. Point it a large data sets, and viola.

Perhaps those who are at the forefront of deep learning made a mistake in making it so easy for everyone else to use.