Neuroscience & Mind Icon Neuroscience & Mind

Robert Marks on the Lovelace Test

Lovelace test

A postscript here to a couple of posts by Sarah Chaffee. She notes today the views of Robert Marks, author of Introduction to Evolutionary Informatics, on the limits of what can be expected, for good or bad, from AI robots. In an earlier post, she considered the “genius” of a computer program, AIVA, that composes classical music.

Serendipitously, Dr. Marks himself now comes along and participates in a new ID the Future podcast conversation with Robert Crowther. They discuss, among other subjects, computer-generated music. Marks offers a really interesting insight.

AIVA can combine musicals styles — that of, say, Bach and Beethoven, if you feed it enough of those two composers’ works. What such a program can’t do is innovate, says Dr. Marks. It can’t strike out in a new direction of its own, put Bach together with Beethoven and come up with…Stravinsky. Such a leap would be uncomputable, therefore permanently beyond the reach of even the most cleverly designed artificial intelligence.

Marks explains the Lovelace test which, unlike the better-known Turing test, focuses precisely on this hard limit to what computer algorithms can do. AI cannot, in this sense, truly create. That indicates an impassable border for AI, not the only one. Beyond lies the unique realm of the human, no matter what addled things Stephen Hawking may say about computers “replacing” us. Listen to the podcast or download it here.

Image: Igor Stravinsky (1918), by Robert Delaunay [Public domain], via Wikimedia Commons.