Recently our director, computer engineer Robert J. Marks, was on the Mark Davis Show in Dallas (660 AM) discussing whether computers can be creative. That’s one of the topics of his recent book, Non-Computable You (Discovery Institute Press, 2022) and his chapter in Minding the Brain (Discovery Institute Press, 2023).

Host Mark Davis worried about computers taking over creativity and robbing humans of a chance to develop it.

Here’s the podcast: (27:10 min) and transcript.

Mark Davis: The AI threat, if there is one, is really creative. It’s writing term papers for people, it’s writing songs, it’s doing things that are usually hatched from the human mind and the human heart, isn’t that different?

But Marks took issue with that:

Robert J. Marks: Well, I would push back with you, Mark. I would say that AI is not creative. In fact, Noam Chomsky said that AI, especially these ChatGPT sort of things, is actually digital plagiarism.

And we see this manifest now all over the place. There’s all sorts of these lawsuits about people going in and suing these AI companies because they used all this copyrighted material in order to come up with their generative AI. The New York Times is suing the people that generate ChatGPT because they found out that they went behind New York Times paywall and used a lot of those articles to train the ChatGPT. Getty Images has a big lawsuit now about generative AI that generates images and they have incredible evidence to that. So it’s going to be interesting to see how this plays out.

So I would push back on that and as a creative, I would go to a guy named Selmer Bringsjord. He’s at Rensselaer Polytechnic. He said, “AI is going to be creative if it does something beyond the intent or the explanation of the programmer.” And we have seen no cases where a program has done more than the intent of the programmers or — if it goofs up — the “explanation” of the programmers.

ChatGPT as a songwriter?

Mark Davis had, earlier in the program, gotten a ChatGPT, a large language program, to write a song. Marks listened to it and commented.

Robert J. Marks: I went back and listened to your podcast and listened to the song. It was pretty good, but it was kind of derivative. I used to be in a garage band. I know enough about chord structures to be dangerous. Yeah, chord structure was kind of nice.

But let’s look at, for example, training artificial intelligence to generate the music of Bach. What would you do? You give it all of the baroque music from Bach and Handel and things like that. And what does it generate? It generates something that sounds like Bach. Will it ever generate jazz? Will it ever generate something that sounds like Wagner? No.

In order for it to generate something that sounds like Wagner, you have to train it on something that generates that… You have to train it on Wagner music. And you ask it specifically if I remember from the podcast, you ask it to generate some soft jazz. And so it did that. It went to that database from which it learned and generated some soft jazz. And the lyrics were pretty good. I mean, it is kind of cool.

That, presumably, is what Chomsky means by digital plagiarism. There is no new input, only remixing.

Will better AI change that?

Mark Davis: But isn’t it true that the AI of a year ago is not the AI of today and the AI of today is not the AI of even this Christmas? It is constantly getting better.

Robert J. Marks: It is constantly getting better. But one of the premises of my book is that there are limitations to computation. Just like there’s limitations to physics. You can’t generate or invent a perpetual motion machine. You can’t go near the speed of light because your mask goes to infinity. There are just some things you can’t do. Mathematics has some things you can’t do. You can’t trisect an angle with a compass and a straight edge.

Better AI, he pointed out, will still depend entirely on the input.

Mark Davis: I myself asked AI how many genders are there? And it gave me the most ridiculous gobbledygook about, well, gender is fluid. It might be a social construct … What is the nature and the future of bias when it comes from an AI brain?

Robert J. Marks: Well, you’re exactly right. If one of these models is trained on literature that says the world is flat and there’s no contradiction, it’s going to come out and it’s going to say the world is flat. The AI without bias is like water without wet. You cannot generate AI without generating bias.

Mark Davis: Which proves something, which goes to say that all these references to cold inhuman calculation, ultimately everything AI does is go in search of something that some human being said or wrote sometime. Correct?

Robert J. Marks: Absolutely. And that goes back to creativity. Yes.

That’s something to keep in mind if you worry that AI will take your job. It’s worth asking, is there any aspect of your job that requires creative intelligence? If so, don’t rage against the machine; work on developing that part!

You may also wish to read: Thinking machines? Has the Lovelace test been passed? Surprising results do not equate to creativity. Is there such a thing as machine creativity?