avatarChaz Hutton

Summary

An artist explores the capabilities of ChatGPT in creating comic ideas, then draws them, to assess the AI's potential in replacing human creativity in cartooning.

Abstract

The article discusses the author's experimentation with ChatGPT, using the AI to generate comic strip concepts. Despite initial skepticism about AI's ability to match human creativity, especially in humor, the author finds that while the AI's comic descriptions show potential, the humor often falls flat or relies on clichés. The author draws the AI-generated concepts, ranging from procrastination jokes to surreal scenarios, and observes that while the AI can produce serviceable content, it struggles with nuanced humor and wordless comics. The experiment leads to reflections on the limitations of AI in creative fields and the unexpected emotional responses it can provoke, even though the AI lacks consciousness.

Opinions

  • The author initially doubts AI's ability to create humorous and engaging comic content but acknowledges the AI's decent effort in generating relatable themes.
  • AI-generated dialogue in comics is seen as verbose with little comedic impact, suggesting a lack of understanding of comedic timing and effectiveness.
  • There is a concern about the ethical implications of using AI art tools, particularly regarding the sourcing of visual data and the potential for infringing on artists' work.
  • The author notes that as requests become more specific, the AI's responses, while sometimes missing the mark, show an impressive ability to incorporate detailed instructions.
  • The AI's attempts at humor are likened to "Dad jokes," implying a certain level of basic, unrefined humor.
  • The author experiences a range of emotions towards the AI, from competitiveness and a desire to "beat" it, to sympathy for its limitations, particularly in creating wordless or surreal comics.
  • The article concludes with a mixed view on AI's role in creative industries: while AI can assist in idea generation, it is far from mastering the subtleties of human creativity and humor.

I Asked ChatGPT to Create Comics, Then I Drew Them.

By now you’ve likely heard of ChatGPT, the open AI chatbot that seemingly does everything from being able to write code, write essays for high school students, content for websites, or even explain the James Webb Telescope while pretending to be Bodhi from Point Break.

Predictably, the articles claiming AI was finally good enough to take everyone’s jobs all came out, quickly followed by all the articles claiming that actually, no, it probably won’t.

Which, as always, makes me start to wonder how far away AI is from doing what I do. In short, how good is this souless collection of 1’s and 0’s at drawing fairly basic and occasionally funny cartoons.

At first glance, it was good news…

Of course, not being able to draw should never stop you from going ahead and drawing comics, so I decided to ask the chatbot for a description of a comic instead.

I’ll be honest, this was a pretty solid effort. Combining some relatable procrastination humour with the idiom your Uncle would use to describe your Arts Degree is actually quite well done. So, given the chatbot’s lack of drawing ability, I figured I’d draw it out for it.

(Quick Caveat: There was an opportunity here to utilise an art AI to then create visuals for these comics, however there’s currently a lot of concern about where the visual data for those AI artworks are coming from and I’d rather not steal someone else’s work while in the process of attempting to destroy my own.)

For another comic option using the same request, the Chatbot decided to add dogs, which cartoon-wise are a real classic and always a great way to go. Interestingly, the fact it’s a single panel comic didn’t stop the chatbot from adding in quite a lot of dialogue for not a great deal of comedic effect.

So, Game On. From here I figured the more specific I made the requests, the more interesting the outcomes might get…

I mean… at this point they’re still not quite hitting the mark, the humour level is christmas bonbon cracker level at best. (although extra points for correcting my typo and carrying on like it didn’t happen)

I also like how it proclaimed it was making everything I asked for, except for it being funny, which makes me wonder if it’s lacking comedic confidence.

So, from here I figured it’s just a case of even more specifics…

In this case I feel like the bot threw in genrification to satisfy my ‘cynical’ brief, even though it doesn’t totally make sense. (to be honest, I was expecting an Ozymandias reference in the caption, that’s comedic no-brainer there)

You can also ask the chatbot to try again, and this is where things got a little weird for me, because I couldn’t bring myself to ask it for more work without saying ‘please’… to a soulless, non-sentient AI. Interesting.

So in my opinion, this is pretty good. The Humour level has advanced up to low-level Dad joke, and it’s also relying heavily on the visual aspect for the joke to work, which I find impressive.

Of course, now that I can see AI taking over my job in the very near future, my intent quickly shifted from vague inquery to a relentless effort to crush this AI’s soul (or, lack thereof) with either more inane specifics or send it off the crative deep end with some good old surrealism.

Four thoughts occured to me with this one, in the following order:

  1. In a panic, and seemingly overloaded with details, the AI had simply repurposed its old comic and had clumsisly shoehorned in the extra information requested. (hey, we’ve all been there)
  2. This made me feel like I was finally beating the AI.
  3. I then realised I was now batteling against an AI at a game that I alone had created, and one it was both unaware of, and certainly not trying to win.
  4. I must ignore this existential confusion and carry on beating it regardless.

To my mind, one of the most difficult single panel comics to create is one that has no words at all; just an image showing a scene that contains all the elements of the joke within. This could be something like Sysaphus standing next to a boulder looking at his Fitbit for instance.

But what would the bot manage to do with that kind of devestating and dibilitating creative constraint…

Our bot is struggling a bit here and now I feel a little sorry for it. On the other hand, I love the fact I didn’t request for this one to be surreal and yet, it’s gone ahead and created a properly surreal and vaguely poignant comic, even if it is a result of being completely confused.

It’s also having trouble with my ‘no words’ request.

So, because I felt sorry for it I figured it was worth giving it another shot…

Oh dear. Oh dear me. I don’t know if AI’s are capable of projection but I gotta say, this really does feel like the work of an artist that is absolutly over it and starting to question why they even took on this weirdly demanding human client in the first place.

At this point I’d given myself quite a bit of drawing to do and so left the chatbot alone for the next couple of days while I drew up all its creations. I returned to it for one last drawing to finish off the article, giving it another shot at the elusive single panel no-worder…

Hate to admit it.

It’s not too bad.

Funny
Comics
AI
Art
Technology
Recommended from ReadMedium