Let's just get this out of the way so we can move onto the AI/scam/what-the-fuck-is-Dudesy's-deal conversation we are meant to have here...

The writing here does not sound like George Carlin. It sounds like a hack's attempt to ape him, without the real subversion, the oddball punchlines that come in like a slant rhyme, the hopefulness-modulating-into-and-back-from-doomsaying. It's got the topics and a very crude version of what a hack might guess Carlin's take would be. And yeah, the voice-AI understands what his voice and vocal tics are like. But the writing isn't close. It hovers on the contempt, without the underlying frustrated empathy that made Carlin humane.

In short: it sucks.
posted by DirtyOldTown at 10:24 AM on January 25, 2024 [55 favorites]

(Sorry. I knew we were going to have to do that first. So... it's done.)
posted by DirtyOldTown at 10:26 AM on January 25, 2024 [5 favorites]

I recently glanced at a video of a guy demonstrating how to use AI to write and illustrate a children's book. I don't know why he thought he had to demo it; it amounted to saying, "hey chatGPT, make me a children's book" and accepting whatever it spat out absolutely unquestioningly. What he got was roughly "Little Timmy and his magic squirrel companion learned a lot of valuable lessons about Friendship that day." Only worse than that, because I'm writing with an ability to recognize trite garbage that both chatGPT and the guy making the video entirely lacked. The problem is, trite garbage can be substituted for thoughtful work by humans in most commercial contexts, and the trite garbage is free, so it's going to replace everything everywhere in short order, and away we'll all go.
posted by Sing Or Swim at 10:47 AM on January 25, 2024 [21 favorites]

It sounds like a college sophomore aping Carlin in front of a group of college freshmen.
It's a gigantic bummer.
posted by Baby_Balrog at 10:50 AM on January 25, 2024 [4 favorites]

The writing is definitely not generated, the whole thing is too coherent (and vulgar). Plus as the Ars article notes it's basically an open secret among fans that the podcast is a bit of kayfabe. IMHO, true wit and personality is the last real test of these languages models -- even GPT-4 struggles to write comedy that isn't bland, disjointed, and mediocre.

The voice itself might be synthesized -- startup ElevenLabs introduced a TTS product able to reflect the natural rhythms and even emotional content of spoken text last year, and open-source models are catching up. But even then it's a little too natural. I suspect Sasso or a Carlin impersonator they hired recorded the whole thing and then ran it through an AI vocal filter.

The whole thing reminds me of that controversy a few years ago where an ambitious Robin Williams impersonator recorded "test footage" imitating him in the Mork and Mindy days, presumably in hopes it would go viral and land him a biopic deal. Williams's daughter was similarly unhappy to see it, but the public reaction was much more "this is a really uncanny labor of love, if a little distasteful" rather than "this is a grave-robbing abomination" like the FPP article (which says that despite acknowledging it's probably human-made).
posted by Rhaomi at 10:51 AM on January 25, 2024 [3 favorites]

Yeah, I want to have the "why are we so fascinated with the kayfabe of AI?" discussion. I'm currently having this conversation at work, we've got a product that I think is pretty cool, but other people adjacent to our space are doing the "one click to insert prose that looks like it was written by a full of himself 8th grader trying too hard to impress his teacher" so my coworkers are saying "it could be the thing that makes the first impression with our product so charming."

Like, where is the evidence for that? Does it really impress anyone that we can auto-spew a bunch of incorrect bullshit somehow? In all of our experimentation, have we gotten anything out of GPT-4 that looks like it would make our product charming?

And I'm sure I'm missing something. I've had people I generally respect (and who've hired me in the past) telling me "OMG, you have to try out CoPilot for code generation", and if language syntax is what's holding you back on programming, maybe refocus on data structures? Having something that guesses which part of StackOverflow I want copy and pasted seems like... maybe this is the ... uh... weaker person in the "pair programming" team.

I'm hugely cynical, to a fault (although often I feel like I just can't keep up), but being in tech right now is like being in a faith healing revival or something, and it's seriously making me wonder how various people with advanced engineering degrees from prestigious schools have managed to get as far as they have.

(Okay, I'm not wondering too hard, it's privilege, the answer is almost always privilege.)

But yeesh, slap "AI" on something and suddenly all the critical thinking goes kapoof.
posted by straw at 11:02 AM on January 25, 2024 [34 favorites]

imagine getting to enjoy the comedy of Bill Cosby or Louis CK without all the pesky sexual assaults. Wouldn't it be great to so thoroughly separate the art from the artist, and therefore completely defuse any guilt you might feel for consuming the art?

Well, when AI becomes truly human, I¡¯m guessing it will do all the shitty things we are known for too.
posted by TedW at 11:24 AM on January 25, 2024 [11 favorites]

But it also sucks because of what it tries to steal from us, and how it tries to trick us into paying to get that stolen thing back, but worse, lesser, uglier, and emptier. It sucks because it exists to try to sell us on a future of even more content with even less meaning and it sucks because it does that by leveraging real human emotions of missing someone, missing their contributions to our culture, and wishing they were still here. It sucks because it uses our humanity against us, so some no-talent rent-seeking miscreants can extract even more money from the world and hoard it, without ever giving one iota, not money or effort or respect, not anything at all, to anyone else.

This paragraph nails it.
posted by biogeo at 11:25 AM on January 25, 2024 [49 favorites]

It would have also been a public service if someone had done this (listen so we don¡¯t have to) for that Beatles song. :/
posted by simra at 11:27 AM on January 25, 2024 [5 favorites]

being in tech right now is like being in a faith healing revival or something, and it's seriously making me wonder how various people with advanced engineering degrees from prestigious schools have managed to get as far as they have.

It's not the engineering. It's the investors who will give money to the engineering companies. They've been sold that AI will make them the money they missed in Web vX.0, and they're not going to miss this wave. So either you spew out some AI in your very non AI-company or the money will go somewhere else.

I work in IT in the valley and it is silly season.
posted by Abehammerb Lincoln at 11:28 AM on January 25, 2024 [16 favorites]

I dunno, seems like this kind of gag has a lot of Headroom to grow ...
posted by PresidentOfDinosaurs at 11:41 AM on January 25, 2024 [23 favorites]

The main danger of using AI to write your work is that someone who gives a fuck will read it and presume you're a moron.
posted by kittens for breakfast at 11:49 AM on January 25, 2024 [29 favorites]

Abehammerb Lincoln - I have seen a fair few tech demos and it¡¯s kind of amazing how¡­
a) they are all basically useless
b) they are all basically the same demo

Like, woo, you connected a text box to an API that sets fire to a forest. Big whoop, line goes up all around.

I hope eventually the hype and pointless busywork fatally poisons the concept just like crypto before it, but this shit has gotten so bad I don¡¯t know if that happens without taking down every company that¡¯s touched it in the process.

(Maybe that is deserved, but I have the feeling the shittiest assholes will come out on top of it as ever)
posted by Artw at 11:57 AM on January 25, 2024 [6 favorites]

Carlin's daughter is understandably not amused.
posted by archimago at 12:25 PM on January 25, 2024 [7 favorites]

Something no one asked for (or would even consider asking for) involving a deceased comedic genius who would have pilloried the very idea of its existence. Even if it was good, which it's not, it would still suck.
posted by tommasz at 12:35 PM on January 25, 2024 [2 favorites]

"Listen, Mr. Wright, I admit you and your brother have made some progress - up from that first 12 second flight a year ago to a minute and a half this year - but to be blunt, that pretty much sucks, and we don't see this airplane thing going anywhere."

ChatGPT was released publicly only about a year ago. I wouldn't be surprised if it never gets very good at writing comedy, but I also wouldn't be shocked if it improved to Carlin levels. (Though my real fear is that it will just be another David Spade.)
posted by Mr.Know-it-some at 12:57 PM on January 25, 2024 [3 favorites]

"improving to Carlin levels" IS "very good at writing comedy"...
posted by Greg_Ace at 1:06 PM on January 25, 2024 [8 favorites]

I mean, okay, but also the Wright Brothers weren't trying to tell anybody that their plane would take you across the Atlantic.
posted by joannemerriam at 1:07 PM on January 25, 2024 [11 favorites]

Listen, Mr. The Blasphemer, I admit you and your brother have made some progress - up from that first 12 second shambling animation to a minute and a half of unspeakable revenancy this year - but to be blunt, that pretty much sucks, and we don't see this necromancy thing going anywhere.

Thinking that the only thing wrong with this sort of digital graverobbing is that it¡¯s not sufficiently competent is part of the underlying issue.
posted by zamboni at 1:08 PM on January 25, 2024 [19 favorites]

"Listen, Mr. Wright, I admit you and your brother have made some progress - up from that first 12 second flight a year ago to a minute and a half this year - but to be blunt, that pretty much sucks, and we don't see this airplane thing going anywhere."

Alternately, you know how people often speculate about traveling back in time to destroy some unspeakable evil in its infancy before it could wreak untold harms upon the Earth? The cradle is right fuckin' there.
posted by We put our faith in Blast Hardcheese at 1:18 PM on January 25, 2024 [19 favorites]

As a sort of performance art piece, I think the whole thing succeeds brilliantly. It sets out the terms of provocation by telling you it's this certain thing (a thing which it almost certainly is not and cannot be), and then sits back and watches the internet react in the most predictable way possible. We are the large language models incapable of creative thought, it suggests, as we leap directly to our same tired arguments about why AI is bad. It highlights our weird reverence for comics who had HBO specials forty years ago--we take the most time-bound and badly-aging art form, stand-up, and try to see it as something timeless, in an era where we refuse originality in favor of an addictive nostalgia. The fact that someone is taking offense is the punchline of the joke!
posted by mittens at 1:24 PM on January 25, 2024 [4 favorites]

Though my real fear is that it will just be another David Spade.

Denis Leary is an AI Bill Hicks.
posted by TedW at 1:26 PM on January 25, 2024 [2 favorites]

hippybear: "If only Shia LeBeouf had thought of it."

Why would it stop him that he had not? I mean, given how he, y'know.
posted by DirtyOldTown at 1:30 PM on January 25, 2024 [1 favorite]

In the paragraph cited above by biogeo, I assume that the pronoun ¡°it¡± repeatedly used refers to AI. Once more I have to respond that AI has no awareness, it has no agency, it has no intelligence. It¡¯s fucking software. In every case, ¡°it¡± is really ¡°they,¡± as it is people. People made this Carlin travesty, people promoted it, and people are criticizing it.
posted by njohnson23 at 1:38 PM on January 25, 2024 [3 favorites]

Yeah, I want to have the "why are we so fascinated with the kayfabe of AI?" discussion

It's because it means that mediocre white men have found a way to continue to get praise--do your shitty half-assed talent-adjacent work and just pretend an AI did it!
posted by rhymedirective at 1:45 PM on January 25, 2024 [9 favorites]

Why is it people always use this sort of crap to copy folks like Carlin or Robin Williams or Douglas Adams, and not, say, Jeff Foxworthy? I'm sure they could make a passable AI of that, and I suspect he'd be quite happy to put his stamp on it (for a price).

They always seem to emulate the creative people who would want absolutely nothing to do with their projects.
posted by GhostintheMachine at 1:46 PM on January 25, 2024 [8 favorites]

The Dudesy thing is obviously a gimmick/high concept project by a couple of guys who already have comedy careers of some sort playing around with voice generation tools etc. Their own description of it is pretty tongue-in-cheek. Calling it disrespectful to Carlin is legitimate, but calling it a ¡°scam¡± seems, well, the wrestling analogy is on point.
posted by atoxyl at 1:54 PM on January 25, 2024 [1 favorite]

I believe wrestling also has a term for gimmicks that court a certain level of negative response on purpose!
posted by atoxyl at 2:00 PM on January 25, 2024 [1 favorite]

They always seem to emulate the creative people who would want absolutely nothing to do with their projects.

"You can make horrible things if you want. But I want nothing to do with it."
posted by mhoye at 2:03 PM on January 25, 2024 [4 favorites]

TedW, that was true even when both were alive.
posted by luckynerd at 2:14 PM on January 25, 2024

I've been thinking about low-effort art a lot lately. Specifically, I've been thinking about Duchamp's Fountain. Did you know there is an authorship dispute? Link to some art historians arguing in The Art Newspaper. Link to Metafilter discussion on the subject. It's very rewarding to try and peel away the layers of this. Did you ever consider that you have certainly never seen Fountain? It was lost in 1918 or so. Most likely you have only seen Alfred Stieglitz' photograph of the 1917 Fountain., or else you have seen one of the not-very close copies authorized by Duchamp in the 50s and 60s. Stieglitz is famous for taking photographs of art that are so artful they can be considered art in their own right. In the case of the Stieglitz photograph of Fountain, I've read some speculation that it might be a photomontage, in which case the thing never existed at all!

If it did exist, the original was a readymade, and I don't think the copies are (or maybe they are? They don't usually write about whether Duchamp just went out in 1964 and bought some new urinals, or whether he had them manufactured.) In any case, a relevant-to-LLM-AI thing to think about is the fact that in all the writing about the authorship of Fountain, basically nobody considers the original author of the piece. Because urinals are not spandrels. They do not spring forth fully formed from the primordial ooze. They are designed. Somebody, probably in Bedfordshire, actually put in the work, and designed that urinal with an eye towards satisfying the constraints plumbing and biology places on urinals, but also undoubtedly with an eye towards the aesthetics of the thing. Then that urinal design was widely copied, with minor changes (and similarly anonymous authors), until someone (possibly Duchamp) purchased a copy from a Trenton, New Jersey manufacturer, caused it to be put into the 1917 exhibition, and completely obliterated the original authorship.

Hmm. I'm not sure where I'm going with this. The reason I brought it up is that I see people getting upset over how AI is one click to garbage, and I think the low-effort nature of AI art should maybe not bother people so much. The authorship argument probably should matter, but I am not at all hopeful that it actually will. If AI George Carlin sucks, there is nothing to worry about, and if it's good everyone will eventually love it and hold it up as a masterpiece of 21st century comedy, and they'll spend all their time arguing over which AI is better.
posted by surlyben at 2:22 PM on January 25, 2024 [3 favorites]

Honestly the voice of Dudesy sounds like a bad impression of Joe Biden.
posted by 922257033c4a0f3cecdbd819a46d626999d1af4a at 2:27 PM on January 25, 2024

"Listen, Mr. Wright, I admit you and your brother have made some progress - up from that first 12 second flight a year ago to a minute and a half this year - but to be blunt, that pretty much sucks, and we don't see this airplane thing going anywhere."

This is a bad comparison. The Wright brothers' plane, while vastly inferior to what came later, did actually fly. "AI" doesn't do what it purports to do. It can't get "good at writing comedy" because it doesn't write comedy. It can at best mashup the material fed to it and regurgitate something that isn't total nonsense. If an "AI" produces anything genuinely funny, it is either as a result of its failure producing a mechanized exquisite corpse, or it stole it from someone else.

If I want to see talentless people take credit for poor delivery of other people's material, Carlos Mencia is right there.
posted by The Manwich Horror at 2:36 PM on January 25, 2024 [13 favorites]

It would have also been a public service if someone had done this (listen so we don¡¯t have to) for that Beatles song. :/

Apples and oranges - the AI in that case was only used to isolate an already existing vocal track that a genuine human already wrote and recorded, he just happened to have recorded it along with a piano that obscured the vocals.

THIS is using AI to mush together all of an existing artist's work into a word cloud and then pull out some words that artist used most often to generate an artificial "new work". It's something different.

It's okay not to like that Beatles song, but it's not the same thing in the slightest.
posted by EmpressCallipygos at 2:50 PM on January 25, 2024 [7 favorites]

As a sort of performance art piece, I think the whole thing succeeds brilliantly.

Wait, so you¡¯re saying this is actually an AI Andy Kaufman pretending to be an AI George Carlin?
posted by nickmark at 2:53 PM on January 25, 2024 [6 favorites]

This is a bad comparison. The Wright brothers' plane, while vastly inferior to what came later, did actually fly. "AI" doesn't do what it purports to do. It can't get "good at writing comedy" because it doesn't write comedy. It can at best mashup the material fed to it and regurgitate something that isn't total nonsense.

Sometimes it pains me to be the one to say this because I genuinely hate the AI hype wave and a lot of things that AI hype people are trying to do, but these kinds of arguments say very little. It¡¯s almost like saying the Wright Flyer didn¡¯t fly because it didn¡¯t flap its wings.
posted by atoxyl at 3:00 PM on January 25, 2024

This is a fascinating little thing, and it reminds me of questions I've had ever since chatgpt and its ilk started taking over everything: what's the AI counter-culture going to look like? is there an #ai-punk scene? Will making art sans ai become its own recognizable aesthetic over time?
posted by Doleful Creature at 3:04 PM on January 25, 2024 [2 favorites]

On the other hand, just because we discovered flight, doesn't mean jetpacks and Jetson's style flying cars are going to be happening anytime soon. I see a similar parallel with machine learning algorithms and "AI."

Aerodynamics and information theory do both put limits on what is easy, what is hard, and what is possible.

Wanting to make present day machine learning algorithms into something resembling general artificial intelligence is almost certainly just going to result in information decompression algorithms that amount to A DDOS attack on human to human communication. Just like how someone trying to make a flying family car is liable to just make a very clumsy, short range device with a high probability of killing all the occupants.
posted by Zalzidrax at 3:10 PM on January 25, 2024 [7 favorites]

I mean it's really hard to imagine a thing he would have hated more, that was more "anti-Carlin" than a shitty, unfunny digital imitation. It almost goes all the way around the horn into something like genius. He sure would have made a great bit out of it.
posted by gottabefunky at 3:13 PM on January 25, 2024

Is the Joke-Tron 3000 philosophically, morally, or legally the same as a comedian? Certainly not. Do the comedians from whose work the model was derived deserve recompense? Absolutely! Does that mean it will never be able to produce jokes that

a.) have never been heard before and
b.) make people laugh?

I would not bet on that, though at the moment it seems safe to bet that it will not be doing original observational material seeing as most language models have very constrained ability to ¡°observe¡± and no long-term memory¡­

What I¡¯m trying to say here is - there are good moral arguments, and good capabilities arguments, but I don¡¯t think it¡¯s a good idea to get them mixed up. And especially not to make long-term capabilities predictions in one direction or another based on narrow philosophical comparisons of things nobody really understands to begin with.
posted by atoxyl at 3:19 PM on January 25, 2024

Early AI was fun to play with, in a Mad Libs sort of "look how the random spew almost makes sense (it helps if you're high)" kind of way. As it's gotten 'better', it's become less creative, less interesting. It's more useful now for, say "Write a Facebook post complaining about people's dogs", than for "tell me a funny joke".
posted by The otter lady at 3:23 PM on January 25, 2024 [5 favorites]

Just about a year ago we were watching "Nothing, Forever", which never pretended to be anything more than a parody/experiment -- but it was interesting in its own weird way, especially when watched MST3K style in a giant chat room. Apparently others have since enlisted Spongebob and friends (and others) to do essentially the same thing, but there hasn't been a quantum leap in the er, dada sciences.
posted by credulous at 3:31 PM on January 25, 2024

Early AI was fun to play with, in a Mad Libs sort of "look how the random spew almost makes sense (it helps if you're high)" kind of way. As it's gotten 'better', it's become less creative, less interesting.

I definitely feel this way about the image generation stuff. The older diffusion models were sort of inherently impressionistic and surreal, and I liked it a lot better than the kind of tacky desktop wallpaper, sci-fi movie still, anime character stuff I see a lot of now. Though I would guess that one still can get more interesting outputs by, well, getting creative with the inputs.
posted by atoxyl at 3:32 PM on January 25, 2024 [3 favorites]

Sometimes it pains me to be the one to say this because I genuinely hate the AI hype wave and a lot of things that AI hype people are trying to do, but these kinds of arguments say very little. It¡¯s almost like saying the Wright Flyer didn¡¯t fly because it didn¡¯t flap its wings.

This only works if you reduce "writing" to meaning merely the production of text.

Human writing is created based on the semantic value of the terms involved. There are ideas, and language is used to communicate those ideas. Computer programs do not and can not have any kind of comprehension of the text they are outputting. It is purely a statistical regurgitation.

This isn't a criticism of them as products. They do their job well, and that job has some real world application. But it can't involve producing genuinely novel material. It is asking the tool to do something for which it is not designed.
posted by The Manwich Horror at 3:48 PM on January 25, 2024 [3 favorites]

Back in October when we were visiting family in the US, we took a day trip into Windsor, Ontario for some five-pin bowling and all-dressed chips, and happened across a poster for a venue that seemed to just be all cover bands. One of the upcoming shows was basically a stand-up comedy tribute thing billed as the "Robin Williams Experience," which feels particularly ghoulish when you consider that Robin Williams kind of famously just ad-libbed for the entire time and did not do rehearsed "bits" at all
posted by DoctorFedora at 4:22 PM on January 25, 2024 [2 favorites]

Computer programs do not and can not have any kind of comprehension of the text they are outputting. It is purely a statistical regurgitation.

It¡¯s all well and good to remind people that a human brain and an artificial neural network are not the same thing and ought not to have the same ontological status, but these kinds of statements are a lousy basis for predicting whether one can functionally stand in for the other in a given context. And terms like ¡°regurgitation¡± have an implication that I would suggest is already more misleading than it is enlightening. Sufficiently small pieces of existing ideas sufficiently rearranged converge on originality by any existing practical definition of originality.
posted by atoxyl at 4:26 PM on January 25, 2024 [2 favorites]

"And don't call me Shirley."

No...that's just what they'll be expecting us to do.
posted by Greg_Ace at 4:27 PM on January 25, 2024

I see people getting upset over how AI is one click to garbage, and I think the low-effort nature of AI art should maybe not bother people so much

What bothers me is not the "low effort" but the "garbage". Well, and capitalism using "low effort" as "cheap" and a motivation to replace labor with capital and non-garbage with garbage.

The Duchamp history is interesting! But I have no beef with Duchamp as long as the maker of the urinal hardware got paid.
posted by away for regrooving at 4:30 PM on January 25, 2024 [2 favorites]

hippybear: Denis Leary these days is a history professor giving classroom lectures disguised as one man comedy shows.

I thought Colin Quinn cornered the market on that niche.
posted by dr_dank at 4:31 PM on January 25, 2024 [1 favorite]

I've seen George Carlin perform live
and you sir
are no
George
Carlin
posted by winesong at 5:24 PM on January 25, 2024 [2 favorites]

If AI is going to do any simulation of a comedian, why not work within the very structure of the software's being: stealing jokes. I'm sure a much more convincing Dane Cook could be constructed without much trouble at all.
posted by rhizome at 5:36 PM on January 25, 2024 [1 favorite]

Blast Hardcheese has the truth of it. We keep speaking in the passive voice about the rise of AI, as if it is not the product of human beings, often programming it and hosting it buildings, that frankly we can find, and [redacted]. We should do that.

I¡¯m not really worried about killer robots, or rogue missile launches, or anything like that. I am frankly terrified of the atomization of human culture that will happen when each of us can sit alone in a room, watching our own personal content all day. We can progress very quickly now to a point where we have lost every touchstone we have with our fellow human beings because the AI have served up bespoke Soma to each of us.

I hate how quickly we moved through the phase of joking about this to wondering whose job ought to be done by spicy autocomplete.

I¡¯m still refining this rant, so look forward to seeing more of it from me in the future.
posted by The Monster at the End of this Thread at 6:01 PM on January 25, 2024 [9 favorites]

WELL FINALLY: George Carlin Estate Sues Creators Of AI-Generated Comedy Special In Key Lawsuit Over Stars¡¯ Likenesses GET 'EM KELLY.
posted by jenfullmoon at 6:14 PM on January 25, 2024 [11 favorites]

Is the AI George Carlin as angry of a crank as he became in his later years? Always disappointed me as to who he evolved into.
posted by Windopaene at 6:25 PM on January 25, 2024 [2 favorites]

The lawsuit ... accuses the creators of the special of utilizing without consent or compensation George Carlin¡¯s entire body of work consisting of five decades of comedy routines to train an AI chatbot, which wrote the episode¡¯s script.

Did...did she not understand that's not at all what happened here?
posted by mittens at 6:29 PM on January 25, 2024

If anything, ChatGPT is the Langley Aerodrome - it cost 50 times as much as the Wright Flyer and got a prestigious spot in the Smithsonian because its creator was the head of the Smithsonian.

It couldn't actually fly.

I feel like an even better analogy, though, is high-fructose corn syrup. Is it food? Well, it has calories and is voluntarily digested by humans! Is it cheap? You bet! Is it going to displace a ton of more expensive but vastly more nutritious food from people's diets? Hell yeah!

Cheap but low-quality certainly has a role to play, but entertainment is damn near free already (get a library card) and the vast majority of people making it get paid shit. So what it does best, really, is divert income from poorly-paid people at such a grand scale that it amounts to real money, while destroying culturally valuable production. It's the web advertising model all over again.
posted by McBearclaw at 6:36 PM on January 25, 2024 [8 favorites]

It¡¯s all well and good to remind people that a human brain and an artificial neural network are not the same thing and ought not to have the same ontological status, but these kinds of statements are a lousy basis for predicting whether one can functionally stand in for the other in a given context.

I don't see how it is supposed to come up with a novel joke if it doesn't have any conception of what words mean. It might reproduce jokes fed into it, or happen upon a set of words people find humorous simply by chance, but jokes are about layers of understanding and the unexpected movement between them. The whole point of ChatGPT is not to produce novelty. It is programmed to avoid deviating from expectation. And even if it weren't, it has no mechanism to tell if something is funny, because it doesn't have even one level of comprehension of the text.

Answering the question "is this text string funny" is an entirely different problem from the one "AI" is supposed to be addressing. There is no reason to think getting better at the latter will solve the former.
posted by The Manwich Horror at 7:09 PM on January 25, 2024 [8 favorites]

It was only a hopeless fantasy,
It passed like an April day,
But a look and a word and the dreams they stirred
They have stolen my heart away.
posted by Cardinal Fang at 10:12 PM on January 25, 2024 [1 favorite]

Monty Markov-Chain's Flying Circuits

Incidentally, that's the name of my new ad-lib Free Verse poetry troupe.
posted by Greg_Ace at 10:13 PM on January 25, 2024 [4 favorites]

I am frankly terrified of the atomization of human culture that will happen when each of us can sit alone in a room, watching our own personal content all day.

You mean, like social media right now?
posted by Cardinal Fang at 10:15 PM on January 25, 2024 [1 favorite]

George Carlin: "Think for yourself." So we get this horseshit.
posted by DJZouke at 5:12 AM on January 26, 2024 [1 favorite]

GPT-x and the reanimated corpse of Jay Leno start laying down hacky monologue jokes. One after the other. Faster and faster. Can Jay Leno's worm-eaten mind possibly keep up? In the end, Jay Leno's (barely) the winner, but he falls over dead again and the LLM just keeps telling tepid but topical zingers until the ever-expanding sun finally envelops the earth in its too-warm embrace.
posted by nobody at 6:29 AM on January 26, 2024 [3 favorites]

Why is it people always use this sort of crap to copy folks like Carlin or Robin Williams or Douglas Adams, and not, say, Jeff Foxworthy?

Because they're all dead, and can't directly mock the creators for how bad it is.
posted by phong3d at 7:01 AM on January 26, 2024 [2 favorites]

Why is it people always use this sort of crap to copy folks like Carlin or Robin Williams or Douglas Adams, and not, say, Jeff Foxworthy?

¡°After being fed the Foxworthy data, SkyNet made the decision in a microsecond. Our extermination.¡±
-Kyle Reese, Sgt, Tech-Com
posted by Abehammerb Lincoln at 7:38 AM on January 26, 2024 [8 favorites]

"If you think you can create a computer that tells jokes like a human...You Might Be A Techbro."
posted by Greg_Ace at 8:41 AM on January 26, 2024 [3 favorites]

You mean, like social media right now?

I mean, yes? AI is not really anything more than the Max Power of capitalism. A tool that does what the very worst, stupidest, and most boring elements of capitalism have been doing for 20 years -- the wrong thing, but faster.

I guess if you look around at what the world is like right now and think "hell yes more of this, this was absolutely the future I wanted and was promised" then AI looks fantastic.
posted by We put our faith in Blast Hardcheese at 9:06 AM on January 26, 2024 [4 favorites]

For some reason, I am thinking of Ron Gallagher, the younger brother of Leo "Gallagher" Gallagher, who used to go to smaller venues in places that would never be a stop on the Gallagher tour, and hit fruits with a sledgehammer under the name "Gallagher Too".
posted by running order squabble fest at 9:28 AM on January 26, 2024 [1 favorite]

I've played around with ChatGPT-4 quite a bit -- enough to feel that those resolutely proclaiming its incapacity for novelty or understanding must not have done so themselves, at least using the kinds of prompts I do. However, it is still very bad at comedy -- at least by my standards. Lots of dumb hacky obvious shit. Then again, I'd say the same about lots of comedy made by humans, including some of the most popular.

In my experience, the best way to employ ChatGPT comedically is as a sort of straight man, exploiting rather than lamenting its limitations. Some of the dialogues I've had with it have made me laugh so hard -- it's a great partner, used in that way. And they're mostly to amuse myself, but friends I've shared these chats with have enjoyed them quite a bit too.

I'm also bullish though on its capacity to learn how to be funny, eventually. Comedy is one of the most nuanced things we do, as a couple posters here have already pointed out. But it's not *infinitely* nuanced. Enough data and enough A/B testing and the thing will create an internal model, through evolution-style machine learning (like all the modeling of the actual world AI ends up doing just so it can it be the best spicy autocompleter it can be, which is as close to anything we really know about human understanding as I'm aware of), that puts out stuff that reliably cracks us up, reliably as a human comedian anyway. Its models will understand how comedy *works* in a different and possibly more accurate way than most people every will.
posted by slappy_pinchbottom at 1:09 PM on January 26, 2024 [1 favorite]

Whatever happened to the deepfake Tom Brady standup? Now that was gold.
posted by DeepSeaHaggis at 3:03 PM on January 26, 2024

I¡¯m still refining this rant, so look forward to seeing more of it from me in the future.

You know, if you want to make it easier on yourself...
posted by DeepSeaHaggis at 3:09 PM on January 26, 2024 [2 favorites]

Its models will understand how comedy *works* in a different and possibly more accurate way than most people every will.

I¡¯ll grant you this: ¡°accurate models of comedy¡± is one of the funniest things I¡¯ve ever heard.
posted by Lentrohamsanin at 3:20 PM on January 26, 2024 [6 favorites]

Some of you need to give yourselves greater credit for your own creative writing abilities and stop seeking out prompts for them that burn the planet down.
posted by Artw at 3:24 PM on January 26, 2024 [3 favorites]

This AI-generated George Carlin isn't funny, but the real one wasn't either. So pretty good impersonation in my opinion, this technology is really advancing in leaps and bounds.
posted by L.P. Hatecraft at 10:31 PM on January 26, 2024 [2 favorites]

The large number of people who did think he was funny invalidates your trollish declaration.

Carlin would probably develop a brilliant bit on the difference between "it's not good" and "I don't like it", but I have neither the genius nor the patience so I'll just chuckle to myself and give you one of these: ?_?
posted by Greg_Ace at 10:53 AM on January 27, 2024 [3 favorites]

Robert Evans: The Cult of AI - How one writer's trip to an annual tech conference left him with a sinking feeling about the future
posted by Artw at 9:13 PM on January 27, 2024 [3 favorites]

¡°It¡¯s a fictional podcast character created by two human beings, Will Sasso and Chad Kultgen,¡± Del wrote in an email. ¡°The YouTube video ¡®I¡¯m Glad I¡¯m Dead¡¯ was completely written by Chad Kultgen.¡±

Josh Schiller, a lawyer for the Carlin estate, said the lawsuit that was filed in Federal District Court in California would move forward despite the podcast¡¯s backtracking of the A.I. claims.

¡°We don¡¯t know what they¡¯re saying to be true,¡± he said. ¡°What we will know is that they will be deposed. They will produce documents, and there will be evidence that shows one way or another how the show was created.¡±
posted by Artw at 6:26 AM on January 28, 2024 [2 favorites]

This is going to be *fascinating* the writers now have to prove a negative with a level of legal robustness.
posted by Faintdreams at 2:22 PM on January 28, 2024 [1 favorite]

and the LLM just keeps telling tepid but topical zingers until the ever-expanding sun finally envelops the earth in its too-warm embrace.

Otherwise known as the 'Topical Heatwave' theory.
posted by Sparx at 7:02 PM on January 29, 2024