I’m reading Melanie Mitchell’s excellent and very interesting new book about Artificial Intelligence. What I didn’t expect was how deeply she would lead the reader in exploration of what intelligence is — and the poor grasp we have on the term. Let me explain with a couple of examples she offers in her prologue.

Consider AI and Creativity

Mitchell tells the story about how, in the 1990s, a musician trained a computer to write works in “in the style of classical composers such as Bach and Chopin”. No one had thought it could be done — much less Douglas Hofstadter (author of Gödel, Escher, Bach). But the computer created some quite passable pieces. How passable?

Mitchell recounts a story from Hofstadter of a presentation he gave at the Eastman School of Music. He discussed the computer with his audience — an audience which included music theory and music composition faculty.

Then he had a pianist play both an obscure (not commonly recognized) mazurka by Chopin himself and a mazurka composed by the computer. The experts were asked to guess which piece Chopin himself wrote.

The faculty voted the computer generated piece to be the one written by Chopin.

This shook Hofstadter, Mitchell observes.

I was perplexed reading this — especially given my past as a musician and my deep involvement in visual arts. Then I realized it’s not as shocking as it first appears.

Chopin’s human brilliance wasn’t making 100 pieces that are all are “Chopin-like”. Rather, before Chopin there was no music like Chopin’s. He created an entirely new music based on his training, history, and his own ear, skills and instincts. Let me suggest his human intelligence was creating music that was both (a) new and (b) uniquely, personally his own.

A computer doesn’t do that (at this time). In the example Mitchell gives, the AI copied (in a very sophisticated way) what Chopin invented.

Here’s where I think there’s an additional clue:  It shouldn’t surprise us that a computer piece might be “more like Chopin” than a Chopin piece. Computers are excellent at looking at a bunch of pieces a human created and finding the stereotypical things in them. Human’s aren’t nearly as good at that process.

In other words, the computer made a brilliant caricature of Chopin. In fact, it’s not surprising that it sounded more like Chopin than Chopin did.

And a lot of people might be fine listening to that music (which is sad because it would put a lot of composers and artists out of business).

What this story should do is force us to seek far more deeply for that which is human.

A thought that occurs to me is that throughout his career, Chopin continued to develop and discover new things. In each of his works he invented something that a computer would add to the set of all things “like a Chopin”. His search is what makes a true Chopin work incredibly human.

I’ll even suggest that the computer piece likely was missing the human “errors” which arise in any composer’s works — and perhaps it’s in the things which aren’t “Chopin-llke” where we’ll find interesting hints at truly human intelligence.

We Don’t Know What Intelligence Is

Playing chess was seen as the epitome of human intelligence. Chess players (especially) believed they had a uniquely human skill and there were relatively few humans who possessed those skills.

Looking back now, it appears what they did have was be the ability to “look forward” in disciplined ways to anticipate different moves and how they would play out. We might find (if we looked quite carefully) that the chess players we admired most had become that way because they had this very unique skill for humans:  The ability to follow branching paths into the future to see where a choice today might lead within the limited boundaries of a chess board.

Before computers, that was an incredible skill which humanity was right to honor. That skill let people do things which appeared magical to the rest of humanity.

Except, computers DID arrive. And it turned out computers are better, faster, and more methodical at following those paths. In fact, that skill is exactly what computers are good at — automating repetitive searching tasks with many steps.

Ok. I’m probably over-simplifying a chess playing computer. But as I read Mitchell I thought “we shouldn’t have been surprised computers conquered chess — we defined intelligence wrong”.

Perhaps this also gives us insight into the silliness modern humanity calls an “IQ test”. The last time I took a real IQ test (not the Facebook fakes) it was filled with things where an agile, detail oriented mind excelled and a complex, thick, brilliantly insightful mind would fail. As I’ve lived life, though, speed and holding detail in mind seem to be less and less critical to “intelligence.” Human intelligence seems far thicker, more complex, and far, far more interesting than remember 11 digit numbers effectively.

So What IS Intelligence? I Don’t Think Words Can Define It.

We shouldn’t expect to ever be able to say “this is intelligence.” Intelligence is a thing which resides outside the ability of humans to express whether in words or music or art or poetry. We can dance around it — but never zero in to say “This is it! THIS is intelligence.”

Should we  be threatened, then, when it turns out a computer can do something well when we previously thought only humans could do it? I don’t think so. There will always be room for humanity. In fact, Mitchell points out how frustrated AI teams are confronting some things that are incredibly easy for people. She offers the example of the incredible struggle computers have with “simple” tasks like facial recognition — things humans do without “thinking” about them and without effort.

Why Should Business Care?

Businesses love to convince themselves they’ve discovered something no one ever saw before. Businesses build incredible mythology about themselves.

And business, in both mythology and defining their own strengths, are very often wrong in ways which lead to their failure.

Businesses need to learn to bring to the fore their skills which are truly human. In the future, it will not be enough to be a business which knows how to play chess well. It will not be enough to be a business which invents a theme and repeats it over and over. We can already see among tech companies it’s far too easy to copy those advantages which are merely short term.

That Leads Us to the Complex

As Rick Nason has observed, the truly human work is the work of complexity. All else will eventually be somewhat easy to copy with AI.

And that defines the ultimate question for business considering AI. It’s nowhere near enough for a business to learn to use AI. In the future, the key to business success will be identifying what it does that AI never could.

And that should give some pause to all businesses. There’s little amid business history to suggest we’ll be very good at knowing what we do that is so truly human.

So we’d better start now — with minds open to discover that what we USED to think was the key to human intelligence might not be after all.

©2020 Doug Garnett — All Rights Reserved

Through my company Protonik LLC based in Portland Oregon, I advise a select group of clients to drive success with better marketing of new and innovative products. I also work with clients attempting to bring new life to Shelf Potatoes or take their existing products to new markets. And I teach marketing and advertising in the Business School at Portland State University. You can read more about Protonik’s services and my unusual background (math, aerospace, supercomputers, consumer goods & national TV ads) at www.Protonik.net.