In the midst of the 2023 Writers Guild of America strike (I wrote about this here), there's a topic that's becoming impossible to ignore: the role of AI in creative fields, specifically comedy writing.
I was thinking the same thing. I do think that’ll change in time as more specific LLMs are built plus as humans learn to collaborate. Most comedy writers start with a premise and then tweak words until they get something they think will work anyway.
I am 100% convinced that AI humor will match human humor some time in the not-too-distant future. I've thought a lot about creativity, too, and I'm pretty confident that many, many more domains previously thought to be "human only" are going to be going the way of the outhouse.
A Doctoral candidate walks into a bar and says, "ouch, my nose!"
A second Doctoral candidate walks into the bar a couple days later, and say,"ouch, my forehead!"
A third Doctoral candidate walks up to the bar a couple days later, and says, "someone ought to tell those MAGA Reumpublicans that this is not what is meant by 'raising the bar' on education.
I don't think we're there yet, but I think we'll all be surprised at how quickly stuff like "improv" will actually work for LLMs. GPT4 is incredibly fast. It doesn't need much time to come up with a response that might take most humans 100x longer to come up with. The originality that improv necessitates comes from the randomness setting; LLMs are deliberately configured to be more "random" and less accurate in some ways, so people get more creativity.
Creativity and accuracy are diametrically opposed, sort of.
My personal guess is something like that with aggregator services allowing consumers to pick and choose what sort of stuff they watch akin to the way cable services used to work. Just a different delivery platform. Perhaps AI interface to fine tune finding stuff? Probably. Yeah, it’s dense.
As you know, I could expound for a long time about the WGA strike. While I do support writers getting paid more, I’m not sure the WGA is presenting their side honestly. Most of it is a long time coming. But the strike is ultimately about an industry that has changed dramatically over the last 5-10 years and it will only change more, particularly in the overall business model. Keep in mind that only 50% of WGA constituents actually get employed with any regularity and rely on residuals on a regular basis to keep the dream alive of hitting it big. Residual payments are actually higher than they’ve ever been but they are much less frequent than the old model.
I feel like AI is just a target to get people riled up. I’m not the only one to think that most writers will want to use AI. I honestly feel like the WGA is asking for the wrong things.
I agree. I also think that the writers have a point, but so does the free market, in a sense. What I mean is that all of a sudden, thousands of new potential writers are going to crop up thanks to being able to create so much faster. It's obviously not limited to writing, but all of these creative types aren't going to take ideas away from TV and from the world; the opposite is true; they're going to bring an enormous amount of novel creative energy to pretty much every single platform. Hollywood is no exception, and wile protectionism makes sense in certain (very short term) instances, at the end of the day, having a lot more artists working, and not some chosen elite group (chosen because of their connections in many cases) is a net positive.
Also, we are going to get even more great TV. The golden age of television is going to continue.
There’s a lot of layers to it definitely. I am also curious how many people outside of LA and NYC even really care. The streaming services have enough content already in the can to last a very long time. Only the big free broadcast networks like ABC, CBS,NBC have anything to be concerned about. And honestly those are dying anyway. Shit is changing. I wish there was more anticipation of that but the WGA has a history of missing every big technological shift in entertainment unfortunately. I think the changes will only make it easier for really unique stuff to make it to air. I’m not sure it’ll make as much money though.
Yes and no. Streaming services have a LOT to worry about. Right now it's a race to the bottom, with only 2 or 3 big winners looking to actually make a profit. They're all losing money right now except for Netflix.
Also: if I could only bottle up your loathing of the WGA, I could sell that hate all over the world and we could both retire like hogs in the fathouse.
Lol yeah I have a ton of resentment towards the WGA as an organization but not the writers themselves.
To your point about streaming services, yes they have plenty to worry about but it’s not new scripts. The streaming industry as I see it will be different in the next 5 years. 5 years ago it was pretty much just only Hulu, Netflix, and Sling. Maybe Crackle? Two of those aren’t even big players anymore. I’d expect things to be wildly different sooner than later. That’s what the WGA isn’t getting and not telling their constituents.
I think you've hit the n-AI-l on the head: Generative AI lacks the subtle cues and cultural context to truly shine as a comedy writer.
The best AI jokes tend to be "Dad Jokes" precisely because they are about puns and wordplay. Words and their meaning are something that large language models are explicitly trained on. But it's much harder to get these models to truly understand what makes us humans tick.
But I wouldn't be surprised if the formula of comedy can be taught quite successfully to more niche AI models. Still, they are likely to become great mimics or parrots of comedy rather than offer genuinely fresh takes.
My experiment with "AI Jest Daily" is supposed to be completely hands-off, but I still find myself "helping" Bing (which I use to generate most captions) by offering a bit of context and potential curious connections between things I see in the picture. This sometimes nudges it towards a joke that's closer to being funny than it otherwise would be. (This also supports your idea of a collaborative effort.)
What I'm excited to see is whether I start noticing improvements in how AI handles comedy as new models are released in the future (provided I stick with AI Jest Daily for that long). Then the site can also serve to trace the evolution of AI humor over time.
Thanks for sharing your thoughts on this and for the shoutout. I appreciate it!
Well said, Daniel! Let's do this again some time soon, and I want you to consider writing a little for the piece as well. You write well, and your observations are on-point.
Also, I agree that AI can be a good writer’s assistant. Hence, AI writing websites like Sudowrite.com. They recently announced their new software, Story Engine, that can write a whole novel with prompting from an offer.
AI being funny is like Data from Star Trek trying to be funny. They both stumble on it once in a while.
Yes! But unlike Data, LLMs are on an exponentially improving trajectory. Hopefully they get it right before the 24th century!
I was thinking the same thing. I do think that’ll change in time as more specific LLMs are built plus as humans learn to collaborate. Most comedy writers start with a premise and then tweak words until they get something they think will work anyway.
I am 100% convinced that AI humor will match human humor some time in the not-too-distant future. I've thought a lot about creativity, too, and I'm pretty confident that many, many more domains previously thought to be "human only" are going to be going the way of the outhouse.
A Doctoral candidate walks into a bar and says, "ouch, my nose!"
A second Doctoral candidate walks into the bar a couple days later, and say,"ouch, my forehead!"
A third Doctoral candidate walks up to the bar a couple days later, and says, "someone ought to tell those MAGA Reumpublicans that this is not what is meant by 'raising the bar' on education.
Top that, AI! ...But watch out for that bar.
If you trained a LLM on comedy skits, eventually it would come up with something funny. But, how do you get a LLM to do improv?
I don't think we're there yet, but I think we'll all be surprised at how quickly stuff like "improv" will actually work for LLMs. GPT4 is incredibly fast. It doesn't need much time to come up with a response that might take most humans 100x longer to come up with. The originality that improv necessitates comes from the randomness setting; LLMs are deliberately configured to be more "random" and less accurate in some ways, so people get more creativity.
Creativity and accuracy are diametrically opposed, sort of.
My personal guess is something like that with aggregator services allowing consumers to pick and choose what sort of stuff they watch akin to the way cable services used to work. Just a different delivery platform. Perhaps AI interface to fine tune finding stuff? Probably. Yeah, it’s dense.
As you know, I could expound for a long time about the WGA strike. While I do support writers getting paid more, I’m not sure the WGA is presenting their side honestly. Most of it is a long time coming. But the strike is ultimately about an industry that has changed dramatically over the last 5-10 years and it will only change more, particularly in the overall business model. Keep in mind that only 50% of WGA constituents actually get employed with any regularity and rely on residuals on a regular basis to keep the dream alive of hitting it big. Residual payments are actually higher than they’ve ever been but they are much less frequent than the old model.
I feel like AI is just a target to get people riled up. I’m not the only one to think that most writers will want to use AI. I honestly feel like the WGA is asking for the wrong things.
I agree. I also think that the writers have a point, but so does the free market, in a sense. What I mean is that all of a sudden, thousands of new potential writers are going to crop up thanks to being able to create so much faster. It's obviously not limited to writing, but all of these creative types aren't going to take ideas away from TV and from the world; the opposite is true; they're going to bring an enormous amount of novel creative energy to pretty much every single platform. Hollywood is no exception, and wile protectionism makes sense in certain (very short term) instances, at the end of the day, having a lot more artists working, and not some chosen elite group (chosen because of their connections in many cases) is a net positive.
Also, we are going to get even more great TV. The golden age of television is going to continue.
There’s a lot of layers to it definitely. I am also curious how many people outside of LA and NYC even really care. The streaming services have enough content already in the can to last a very long time. Only the big free broadcast networks like ABC, CBS,NBC have anything to be concerned about. And honestly those are dying anyway. Shit is changing. I wish there was more anticipation of that but the WGA has a history of missing every big technological shift in entertainment unfortunately. I think the changes will only make it easier for really unique stuff to make it to air. I’m not sure it’ll make as much money though.
Yes and no. Streaming services have a LOT to worry about. Right now it's a race to the bottom, with only 2 or 3 big winners looking to actually make a profit. They're all losing money right now except for Netflix.
Also: if I could only bottle up your loathing of the WGA, I could sell that hate all over the world and we could both retire like hogs in the fathouse.
Lol yeah I have a ton of resentment towards the WGA as an organization but not the writers themselves.
To your point about streaming services, yes they have plenty to worry about but it’s not new scripts. The streaming industry as I see it will be different in the next 5 years. 5 years ago it was pretty much just only Hulu, Netflix, and Sling. Maybe Crackle? Two of those aren’t even big players anymore. I’d expect things to be wildly different sooner than later. That’s what the WGA isn’t getting and not telling their constituents.
I expect several of them to be consolidated into bigger streamers, like with WBD recently.
Let's keep talking about this space. I feel another collaboration coming on.
I think you've hit the n-AI-l on the head: Generative AI lacks the subtle cues and cultural context to truly shine as a comedy writer.
The best AI jokes tend to be "Dad Jokes" precisely because they are about puns and wordplay. Words and their meaning are something that large language models are explicitly trained on. But it's much harder to get these models to truly understand what makes us humans tick.
But I wouldn't be surprised if the formula of comedy can be taught quite successfully to more niche AI models. Still, they are likely to become great mimics or parrots of comedy rather than offer genuinely fresh takes.
My experiment with "AI Jest Daily" is supposed to be completely hands-off, but I still find myself "helping" Bing (which I use to generate most captions) by offering a bit of context and potential curious connections between things I see in the picture. This sometimes nudges it towards a joke that's closer to being funny than it otherwise would be. (This also supports your idea of a collaborative effort.)
What I'm excited to see is whether I start noticing improvements in how AI handles comedy as new models are released in the future (provided I stick with AI Jest Daily for that long). Then the site can also serve to trace the evolution of AI humor over time.
Thanks for sharing your thoughts on this and for the shoutout. I appreciate it!
Well said, Daniel! Let's do this again some time soon, and I want you to consider writing a little for the piece as well. You write well, and your observations are on-point.
Also, I agree that AI can be a good writer’s assistant. Hence, AI writing websites like Sudowrite.com. They recently announced their new software, Story Engine, that can write a whole novel with prompting from an offer.