There’s this great story that the jazz pianist Herbie Hancock tells about playing in the Miles Davis Quintet in Stuttgart when he was young:
“We started and everything was fine, and I remember that we were playing... Tony Williams was playing drums, Ron Carter bass, Wayne Shorter saxophone. And it was a really hot night, the music was tight, it was powerful, it was innovative and fun. We were having a lot of fun and... right in the middle of Miles' solo, when he was playing one of his amazing solos, and I'm trying, you know, I'm in there and I'm playing, right in the middle of his solo I played the wrong chord—a chord that just sounded completely wrong. It sounded like a big mistake and I did this and I went like this: I put my hands around my ears and Miles paused for a second and then he played some notes that made my chord right. He made it correct, which astounded me. I couldn't believe what I heard. He was able to make something that was wrong into something that was right, with the choice of notes that he made, and that feeling that he had.”
Jazz fans will tell you that this is an illustration of what Miles meant when he said “the next note determines whether the note you played was wrong”.
Herbie’s chord was “wrong” in the key they were playing in, but it was “right” in the key that Miles shifted the whole song to by playing the notes he played. (My classical musician son had to explain to me that keys may be different but they can share the same notes. So while in the context, say, of the three chords that Herbie played, that last one was not in the key and therefore “wrong”; in the context of the note that Herbie played before it and the note Miles played after it, it was “right” – it was just in a different key.)
Which I think has a lot to do with how we should look at AI and what creatives can do with it.
AI, as I understand how it is currently constituted, uses really complex math to give us “the next most likely answer” based on the information it has at hand. Think of this as giving us “the next most likely note within the key we’ve been playing”. In that sense, it can’t give you a wrong note – it can only give you the notes based on the notes you’ve told it you’re playing. (What about “Hallucinations” – those freakish aberrations that luddites and critics of AI like to point to with derision and glee, you ask? Hey, if that stuff is in the data you’re feeding into the math, then it’s in the key, as it were. So It’s not “wrong” – even if it’s horrifying)
So AI can’t give you a wrong note, as it were - only humans can (at least right now). And this unfortunately is where a lot of people stop when they’re talking about AI. They say “AI will be able to do everything we humans can do, faster and better and error-free (barring hallucinations, of course), so it will take our jobs and consign us all to the dustbin of history.” Well, those of us not in charge of deciding whose jobs AI can do faster and better, I suppose.
But it is a mistake to stop there. Just like it was a mistake for Herbie Hancock to stop when he hit the wrong chord. Because AI also can’t make that wrong note into something remarkable, like Miles did. Like every musician does when she changes keys. Because changing keys (metaphorically speaking) is not in the math. It’s in us. It’s in our humanity. Hell, AI can’t even make the wrong note that creates the opportunity for something wonderful. It can only give you the next most likely note in the key. Which is fine. It’s just not wonderful.
Later in that same story, Herbie says:
“…the only way we can grow is to have a mind that is open enough to be able to accept situations, to be able to experience situations as they are and turn them in to medicine. Turn poison into medicine. Take whatever situation you have and make something constructive happen with it.”
The mistake right now is that people keep looking at AI as if it will help us play more right notes faster. That it will make us more “efficient” – because this is always the expectation of new technology. But invariably it doesn’t make us faster. Quite the opposite. It slows us down because it gives us more options. Options that often reveal paths we never knew existed. And I think that’s the opportunity of AI – for us to figure out how, not just to make the wrong notes right, but to make the wrong notes into something powerful and innovative and fun.
AI – as it is right now – doesn’t do that. But humans, as they are right now, do.
So it’s up to you. What note are you going to play next?