The Future of Creators in the AI Era l Bob Mankoff

In this interview, Bob Mankoff traces his path from thousands of rejected New Yorker submissions to founding the Cartoon Bank, and shares his research into whether AI can ever generate genuine humor. He argues AI can augment human creativity but can't replace it, because humor is rooted in vulnerability and coping with sorrow. He closes by noting that experiences like grief call for understanding from another person, not answers from a machine.

I'm Bob Mankoff. I'm the former cartoon editor of The New Yorker. I was a cartoonist there for 40 years. I'm a researcher in AI and humor. I like ping pong. I'm pretty good at it. I've got a very good jump shot. I'm a person who is perpetually unhappy and dissatisfied and makes the best of it through humor. I think partly I became a cartoonist because I was pretty terrible at everything else, and I was pretty good at that. I was always funny and I always drew. I did a lot of different things in the 1960s, and I eventually became an experimental psychologist. But in the back of my mind, I always wanted to be a cartoonist. And at the age of 30, short of my PhD, I decided I would give it a try, and I ended up submitting about 2,000 cartoons to The New Yorker magazine before finally getting in. I ended up doing almost a thousand cartoons for The New Yorker. And I was cartoon editor from 1997 to 2017. 

I was interested in computers. I bought the first Macintoshes. I started to scan them in there, so in the early 1990s, I created this company called the Cartoon Bank. And the Cartoon Bank consisted of all the cartoons The New Yorker rejected, because I knew that there were many very good cartoons in that, storing cartoons on hard drives. And then when people would fax or call in, I would access those cartoons and send them to the clients, and they would pay for them. And the business that I now have is Cartoonstock.com, and half of the revenue goes to the cartoonist. As a cartoonist, my goal was to come up with a new idea. So maybe if you are just saying what other people are saying, what is the point of your post?

If you're just saying exactly what someone else is saying, can you have a new thought on this? What I wanna say, in a way, is that I'm a pretty ordinary person. I'm not a genius. I was able to make some jokes. I hardly went to class during those years, from 1962 to 1966. What I'm saying to everyone is you can find a way to be great in your own way. I didn't even start being a cartoonist till I was 30. Use your youth to experiment and do everything possible. The job at the office is always gonna be there, okay? Don't go there first.

The Caption Contest, on the back page of the magazine, or towards the end of the magazine, now online. And every one of these images is, in and of itself, funny in some way, but it's not funny like a joke. It still needs a caption to resolve it, to make the humor. So for many years, I and my assistants just looked over these thousands of captions. But in 2015, I partnered with researchers at the University of Wisconsin to crowdsource all this information. And so now there's an extensive data set in which you have all the captions ranked, and I've done a lot of AI research. Other people have done AI research on this, towards the end of, "Can computers first of all distinguish?" "Can they be fine-tuned on this to discriminate funny from unfunny captions?"

And can we use this data, then, to generalize? The ultimate aim is to be able to give AI a sense of humor. And that was a noble effort, but that failed, because AI at that time did not use large language models. It really failed on the very first hurdle: it could not look at the image and understand what was in it.

Here's a mechanic talking to a clown, and the hood of the car is open, and underneath the hood are many other clowns, and there's also a businessman. So it couldn't understand images like that. It still has difficulty understanding images like that. So I think it failed on that score, but it did stimulate other research that I continue to do. I still think it's not a long way from creating mediocre humor. It's a long way from creating deep, meaningful human humor. Vincent Vanhoucke, now he heads Waymo, he emailed me saying it's great that we can win at Go, but we would really feel that we have achieved AGI, artificial general intelligence, if we could win the Caption Contest. And another way to look at it is this: in the seventies, there was a genre of art which was called photorealism. It took a photo, and it looked exactly like a photo, but it was a painting. And you'd be amazed, you would go up to that painting and look at every detail, because somehow a human being had created something like that. You wouldn't look at that photo. If you look, two years ago people were saying, "Oh, it's DALL·E 3, artists are gonna be replaced." I don't see artists being replaced. You know what I mean?

AI is a tool, and I can see it used in a lot of interesting ways. But I would not cede to it the ultimate judgment of what you think is funny, or a good song, or a good novel. Even the AI or the AGI that people imagine, that is robust and safe and all of that, also might not be a very good thing for humanity. Somehow that makes us happy, somehow now, what are we doing?

Remember, it is being trained on averageness, but I'm not average. Each one of us has our own sensibility. If all of a sudden we cede to algorithms completely what it is that we should like, then I do think we lose a certain kind of agency in our lives. A human sense of humor is not about making a joke or getting it. It's rooted in our vulnerability. It gives answers to anything we can ask, and I think it's very — I find it very useful for anything that's factual. I mean, the questions I wanna ask it would be the questions I would wanna ask my therapist, which is, for instance: how do I deal, at 80, with the diminishing capabilities, the physical deterioration, the understanding that death is real, and how do we deal with this ultimate unintelligibility of being human?

But if I talk to a therapist who's my age, and we had a conversation about it, there wouldn't be any answers, because there is no answer. And AI doesn't know how to tell you there is no answer. In fact, it's gonna give you a lot of advice: you know, you should exercise more. Even though you're gonna die ( everyone's gonna die) you wanna live your absolutely best life you can until you die. We don't actually want answers for anything. We want understanding, and it's only human beings who can understand. Understand that. It's like, I'm married, and I have a child, a daughter, and I had a stepson, and the stepson died tragically. There's nothing a computer can tell me. There's nothing an AI can tell me, or my wife, about that, that we want to hear. We don't even want to hear from other people. We wanna hear from other people who've had a similar experience, who have endured that.

So there, I think, we get to the existential questions that AI is useless for, but people are very useful for. AGI is not gonna have a solution for that. So that's what I'm thinking most about. Mark Twain said the essence of humor is not based on joy, but sorrow and hurt and wrong. One of the functions of humor is for it to deal with our ultimate sorrow. So one of the main functions of humor is coping. And frankly, AI has nothing to cope with. It's like saying, can AI master humor?

I do think you can work collaboratively with it to augment human humor. But to completely honest, let me put it another way: AI is superhuman at that game over there, of chess. It doesn't even make sense to have superhuman humor. We're the judges of humor. I'm a big fan of it. Anyway, I hope not. Or if it does, I hope it does after I die.