Hey guys, I've got a lot of requests for my famous purple stuffed worm in flapjaw space - all you need is a tuning fork to do a raw blink on Hari-Kari rock
>But the “success” of Super Recipes suggests something like the opposite--weird and bad A.I. generated content is often much more arresting and engaging than the human equivalent, and the “better” A.I. gets--the more human--the less we’ll even bother looking at it.
Not sure if this is intentional "being wrong on purpose for corrective engagement" bait or not, but this seems like a misunderstanding of how AI is going to get "better." With shitting out revenue-generating content being a major productive use of AI, people are going to train models that do that. So they'll get more and more effective at "shitting out high-engagement content," which, as you note, is not particularly correlated with being "more human." "Weird and bad" AI content will just be a gross offshoot for as long as it remains a successful way to monetize attention.
Maybe the safety rails + professional/very very serious techno-gospel aspirations of the current major firms will see them try to cut down on this type of weirdo shit for a little while (??? even this seems wrong though since we're already seeing spammers make such prolific use of their products), but if they do, eventually someone will pitch to the right investor that there's ad revenue being left on the table, then make grossout AI shit their whole deal.
And with how responsive grossout AI could be to changing trends in human attention, we'll get to see some real weird shit as it interacts/feedbacks with our changing psychologies/cultures/interests. The real fun will start once we get a full generation who grows up on a content-system that's machine-learning-calibrated to arrest their attention as strongly and constantly as possible. What will they do? To themselves?? Their content??? Their eggs???? Let's find out!
I agree with all this--any kind of content trained to platform incentives is going to maximize certain kinds of weirdness, whether it's A.I.-generated or not.
But I think a lot of the conventional wisdom around A.I. suggests that the next frontier of generative A.I. is fully non-weird stuff that effectively replaces human authors, movie studios, human game developers, etc. by creating stuff so perfectly attuned to each individual's desires that it obviates all human work. Now--I agree with you that, judged by engagement time, stuff "attuned to individual desire" is likely to be kind of spiky and strange, and not like a fully A.I. generated blockbuster movie but where I'm the star. But the marketing from places like OpenAI focuses a lot on the generative models' fidelity to pre-existing forms (thinking e.g. of the videos they show off made my Sora). This makes sense because the real pitch for A.I. is as replacement for creative labor. But my central point is that the broad conventional wisdom/marketing pitch about what's next with A.I. and how it might transform cultural consumption and the internet misunderstands (1) what's appealing about art (even mass art) in general and (2) the bigger environment in which this all plays out.
I wonder if the immersion of all one's time on clickbait is behind the number of amazingly "low information" voters who actually say they don't realize trump is under multiple indictments. Not to mention having never heard of Project 2025. Maybe the Dems should start creating this kind of site, and in between the recipes or cats or Amazing Stories about celebrities they could, instead of ads, just provide a bit of, y'know, information. "DID YOU KNOW that trump is promising to put millions of immigrants into prison camps and you will be taxed to pay for it?"
That tOSU speech went off the rails super fast. It was memorable, and if that was the goal, then achieved. But in such a brutal way as to generate an astronomical number of "the fuck?" from graduates, family, faculty and (gestures broadly) everyone. Not usually a good thing.
The one bizarre recipe fails video series I remember watching was one of a bartender guy making fun of poorly-made cocktail mixing videos starring a young blonde woman named Jenee.
One was the making of an Old Fashioned. Now you can tell these videos were made by a company that just wanted to make YouTube videos and grab cash, because there was zero professionalism involved. It was shot in a black room with no studio lighting, and Jenee did not have any proper equipment to work with other than the booze. No barware, no utensils, wrong ingredients for the cocktail, but the real hilarity ensues when she doesn't have a rocks glass but mixes the drink in a tall soft drink glass.
There's only one way for Max to solve the egg horror mystery—he must come to Brazil!
Hey guys, I've got a lot of requests for my famous purple stuffed worm in flapjaw space - all you need is a tuning fork to do a raw blink on Hari-Kari rock
>But the “success” of Super Recipes suggests something like the opposite--weird and bad A.I. generated content is often much more arresting and engaging than the human equivalent, and the “better” A.I. gets--the more human--the less we’ll even bother looking at it.
Not sure if this is intentional "being wrong on purpose for corrective engagement" bait or not, but this seems like a misunderstanding of how AI is going to get "better." With shitting out revenue-generating content being a major productive use of AI, people are going to train models that do that. So they'll get more and more effective at "shitting out high-engagement content," which, as you note, is not particularly correlated with being "more human." "Weird and bad" AI content will just be a gross offshoot for as long as it remains a successful way to monetize attention.
Maybe the safety rails + professional/very very serious techno-gospel aspirations of the current major firms will see them try to cut down on this type of weirdo shit for a little while (??? even this seems wrong though since we're already seeing spammers make such prolific use of their products), but if they do, eventually someone will pitch to the right investor that there's ad revenue being left on the table, then make grossout AI shit their whole deal.
And with how responsive grossout AI could be to changing trends in human attention, we'll get to see some real weird shit as it interacts/feedbacks with our changing psychologies/cultures/interests. The real fun will start once we get a full generation who grows up on a content-system that's machine-learning-calibrated to arrest their attention as strongly and constantly as possible. What will they do? To themselves?? Their content??? Their eggs???? Let's find out!
I agree with all this--any kind of content trained to platform incentives is going to maximize certain kinds of weirdness, whether it's A.I.-generated or not.
But I think a lot of the conventional wisdom around A.I. suggests that the next frontier of generative A.I. is fully non-weird stuff that effectively replaces human authors, movie studios, human game developers, etc. by creating stuff so perfectly attuned to each individual's desires that it obviates all human work. Now--I agree with you that, judged by engagement time, stuff "attuned to individual desire" is likely to be kind of spiky and strange, and not like a fully A.I. generated blockbuster movie but where I'm the star. But the marketing from places like OpenAI focuses a lot on the generative models' fidelity to pre-existing forms (thinking e.g. of the videos they show off made my Sora). This makes sense because the real pitch for A.I. is as replacement for creative labor. But my central point is that the broad conventional wisdom/marketing pitch about what's next with A.I. and how it might transform cultural consumption and the internet misunderstands (1) what's appealing about art (even mass art) in general and (2) the bigger environment in which this all plays out.
Ohh, lol I was just misreading you as suggesting that the AI marketing was likely to be the actual content trajectory. Thanks for clarifying!
I wonder if the immersion of all one's time on clickbait is behind the number of amazingly "low information" voters who actually say they don't realize trump is under multiple indictments. Not to mention having never heard of Project 2025. Maybe the Dems should start creating this kind of site, and in between the recipes or cats or Amazing Stories about celebrities they could, instead of ads, just provide a bit of, y'know, information. "DID YOU KNOW that trump is promising to put millions of immigrants into prison camps and you will be taxed to pay for it?"
Indeed: Come to Brazil! [2]
That tOSU speech went off the rails super fast. It was memorable, and if that was the goal, then achieved. But in such a brutal way as to generate an astronomical number of "the fuck?" from graduates, family, faculty and (gestures broadly) everyone. Not usually a good thing.
The one bizarre recipe fails video series I remember watching was one of a bartender guy making fun of poorly-made cocktail mixing videos starring a young blonde woman named Jenee.
One was the making of an Old Fashioned. Now you can tell these videos were made by a company that just wanted to make YouTube videos and grab cash, because there was zero professionalism involved. It was shot in a black room with no studio lighting, and Jenee did not have any proper equipment to work with other than the booze. No barware, no utensils, wrong ingredients for the cocktail, but the real hilarity ensues when she doesn't have a rocks glass but mixes the drink in a tall soft drink glass.
Wild stuff this week!
Max Meyer is not a venture capitalist, he’s just a young boy.
Maybe AI is seeing how many humans die in grease fires to gauge the next phase - gardening tips.