My hope is that people's addiction to prestige will win out and chase them away from the AI internet, because an AI-infested internet is just not as good as "real" internet at conferring status on those who win at being online. We've seen this with people falling away from Twitter, for example. People will never fully be offline, but when everyone just assumes that all your followers/likes/subscribers are not real (and also that you yourself aren't real), then the same vanity that drew people to the internet will hopefully drive them away. Then we go back to having more local cultures and scenes.
An anecdote from military history might offer an instructive parallel. Early firearms weren't superior to the longbows they ousted from the battlefields of Europe. Quite the contrary, that first generation of guns were all around inferior weapons. They were less accurate, had less effective range, and a far slower rate of fire than the old fashioned longbow. A skilled archer was deadlier than any early gunner.
How then did firearms ever manage to supplant the bow in the first place? Because archers required intensive specialized training. They had to practice extensively almost from the time they learned to stand to be any good. When they were killed, they could be difficult to replace. Meanwhile, any idiot could become proficient with a gun within a few hours. Gunners replaced archers not because they were better, but because they were cheaper.
The analogy with internet content creators seems exact. An educated flesh and blood human might still draw a better picture or write a superior Sport's Illustrated profile than a bot. But the human needs a salary and benefits. AI costs next to nothing. It doesn't even require appreciation. Humans, like archers, will be priced out of the market.
Guns, of course, eventually evolved into more lethal weapons than bows ever were. Bots, I fear, will become better artists and writers than humans could hope to be. With the accelerated pace of technological change, it might happen sooner than we think. I'm 53 and in poor health, but I'm still afraid I'll live to see it.
(Suggested reading: War Before Civilization: The Myth of the Peaceful Savage, by Lawrence H. Keeley)
I agree with your general framing of the notion of semantic pollution, Erik, and also that it is important, and that it is a tragedy of the commons. However, this was already happening before the AI-advances of the past few years. Bizarre mass-produced kids videos have been on Youtube for at least 7 years, probably longer, and so before they could have been AI-generated. The business logic and consequences of the older content farms are similar. Now AI is making it cheaper to generate the content. But was "content generation cost" ever the bottleneck to profit for these operations? I'm not sure. Overall, it seems like more of an acceleration of an existing problem right now. But I share your concern that the problem of semantic pollution will be qualitatively worse within a few years.
Had an interesting conversation with an approx. 20-year-old guy who told me he thinks his generation will abandon the internet altogether; such a practice seems highly rational and sensible if the alternative is to be exposed to the kinds of garbage discussed here.
The positive spin, here, is that the cultural pollution has been happening for at least ten or fifteen years, but now, with the tidal wave of bilge AI content washing over everything, people are finally starting to see the problem. And the solution, as before, will be a combination of human curation and trust markets. The good news is that the more people become aware of the problem, the more profitable the solution will be.
It’s damming that YouTube Kids doesn’t allow comments, for obvious reasons—a loophole identified by the inhuman content farms—leaving out any direct way to signal parents of the subpar content quality.
We’re on a rapid decline. The upcoming generation, well beyond their toddler years, faced the lockdown education system headfirst unprepared. Now dissociated from their peers and unable to read/write adequately, indicated by the slew who’ve taken their pleas online to voice the growing epidemic.
If utopia is ahead, who maintains it other than the machines the predecessors will leave behind?
I think AI generated content is just the logical consequence of the attention economy. If you make money by getting eyeballs on your content, then finding a way how to produce it cheaply, fast, and at massive scale will become the main objective. In that sense, we haven't landed anywhere surprising - we're just cruising on the same predictable trajectory, like a comet through space.
While I fully agree with this critique of AI-generated content and share your sense of impending doom for the Internet, your use of “the tragedy of the commons” to build you argument took me aback. Economist Elinor Ostrum debunked this idea (which was never based on data) when Hardin first published his theory in 1968. Her work on the topic finally* won the Nobel Prize in Economics in 2009, making her the first woman to receive the prize. This “tragedy” has no basis in observed fact or data modeling— it’s the theory of one biologist that got traction because it aligned with neo-liberal anti-communism and a “dog-eat-dog” interpretation of human nature that tacitly endorses unbridled capitalism. We MUST stop using it to build arguments, even good ones, and let it whither away . Read a fairly good takedown of this non-tragedy: https://aeon.co/essays/the-tragedy-of-the-commons-is-a-false-and-dangerous-myth
“Might not actual human-generated cultural content normally contain cognitive micro-nutrients (like cohesive plots and sentences, detailed complexity, reasons for transitions, an overall gestalt, etc) that the human mind actually needs?”
Well said, Erik. Some of the children's video problem appeared several years ago with the Elsagate phenomenon. I'm not sure we've ever fully determined how much of that was AI generated, and how much was just garbage skits/animated shows from places with cheap labor. Regardless, Elsagate demonstrated that inert, unsophisticated child viewers could be a highly lucrative audience for torrents of low-quality but shocking visual content. Beloved characters like Spiderman and Elsa giving each other injections, playing with feces, weird pregnancy stuff --- the fusion of schizophrenia with Saturday morning cartoons.
And it turns out that children are not the only audience for this swill. I can't wait until it gets piped straight into our VR headsets. Or, what the hell, right into our Neuralinks.
I really got it for the first time a few days ago when I went on YouTube to get into a music hole. I do that every few months: I grab a beverage, play a song I like and see where the algorithm takes me — YouTube has served for many years as a source of new music and fascinating reviews by experts that have led me to expert producers’ explainers on why pop music pops, singing techniques of various a cappella groups, a Siberian metal band, and a bevy of new independent artists I’d never have found directly. This last session, I gave up after about ten minutes. More than half of the videos YouTube fed me were AI-produced schlock, mostly consisting of duets between various dead people. I’m so sad one of my favorite pastimes is ruined, but even more, sad for all those artists whose work is being buried in an avalanche of dreck.
Somehow in someway, we need to grow a community of people in every free country where the ethos is to pay for, or only try to engage with, creators who do not use generative AI. That’s what I try to do, a large part because as a creator myself, I believe using generative AI lacks integrity. I pay for your newsletter Erik, not only because it’s brilliant, but it seems like you’re not cheating me with your creations, that you’re not using AI. Now, as you said in your piece, there exists no decent AI detection software so what needs to happen is development of faith and trust in those you believe are not using AI - and an adjacent community of investigators who seeks to expose creators who use AI. Anyway great article, once more you freaked me out lol.
I was skeptical: As a lazy parent, I know there are millions of lousy toddler videos on youtube. (Even Ryan's world is bad and most is worse.) Without AI. But, I stand corrected: This AI stuff manages to creep me out. Scott Alexander was right about the future of movies by AI: "The dumbest possible way to do this is to ask GPT-4 to write a summary (“write the summary of a plot for a detective mystery story”), then ask it to convert the summary into a 100-point outline, then convert that into 100 minutes of a 100-minute movie, then ask Sora to generate each one-minute block." - But that 'future' is now - for the youngest. - I hope the market will soon offer easy ways to block that BS out. If not: tough luck for the kids of less-than-perfect-parents.
The last thing the world needed was more crap content, and yet somehow that’s the great invention of the last five years: a machine that prints crap.
My hope is that people's addiction to prestige will win out and chase them away from the AI internet, because an AI-infested internet is just not as good as "real" internet at conferring status on those who win at being online. We've seen this with people falling away from Twitter, for example. People will never fully be offline, but when everyone just assumes that all your followers/likes/subscribers are not real (and also that you yourself aren't real), then the same vanity that drew people to the internet will hopefully drive them away. Then we go back to having more local cultures and scenes.
An anecdote from military history might offer an instructive parallel. Early firearms weren't superior to the longbows they ousted from the battlefields of Europe. Quite the contrary, that first generation of guns were all around inferior weapons. They were less accurate, had less effective range, and a far slower rate of fire than the old fashioned longbow. A skilled archer was deadlier than any early gunner.
How then did firearms ever manage to supplant the bow in the first place? Because archers required intensive specialized training. They had to practice extensively almost from the time they learned to stand to be any good. When they were killed, they could be difficult to replace. Meanwhile, any idiot could become proficient with a gun within a few hours. Gunners replaced archers not because they were better, but because they were cheaper.
The analogy with internet content creators seems exact. An educated flesh and blood human might still draw a better picture or write a superior Sport's Illustrated profile than a bot. But the human needs a salary and benefits. AI costs next to nothing. It doesn't even require appreciation. Humans, like archers, will be priced out of the market.
Guns, of course, eventually evolved into more lethal weapons than bows ever were. Bots, I fear, will become better artists and writers than humans could hope to be. With the accelerated pace of technological change, it might happen sooner than we think. I'm 53 and in poor health, but I'm still afraid I'll live to see it.
(Suggested reading: War Before Civilization: The Myth of the Peaceful Savage, by Lawrence H. Keeley)
I agree with your general framing of the notion of semantic pollution, Erik, and also that it is important, and that it is a tragedy of the commons. However, this was already happening before the AI-advances of the past few years. Bizarre mass-produced kids videos have been on Youtube for at least 7 years, probably longer, and so before they could have been AI-generated. The business logic and consequences of the older content farms are similar. Now AI is making it cheaper to generate the content. But was "content generation cost" ever the bottleneck to profit for these operations? I'm not sure. Overall, it seems like more of an acceleration of an existing problem right now. But I share your concern that the problem of semantic pollution will be qualitatively worse within a few years.
Had an interesting conversation with an approx. 20-year-old guy who told me he thinks his generation will abandon the internet altogether; such a practice seems highly rational and sensible if the alternative is to be exposed to the kinds of garbage discussed here.
The positive spin, here, is that the cultural pollution has been happening for at least ten or fifteen years, but now, with the tidal wave of bilge AI content washing over everything, people are finally starting to see the problem. And the solution, as before, will be a combination of human curation and trust markets. The good news is that the more people become aware of the problem, the more profitable the solution will be.
It’s damming that YouTube Kids doesn’t allow comments, for obvious reasons—a loophole identified by the inhuman content farms—leaving out any direct way to signal parents of the subpar content quality.
We’re on a rapid decline. The upcoming generation, well beyond their toddler years, faced the lockdown education system headfirst unprepared. Now dissociated from their peers and unable to read/write adequately, indicated by the slew who’ve taken their pleas online to voice the growing epidemic.
If utopia is ahead, who maintains it other than the machines the predecessors will leave behind?
I think AI generated content is just the logical consequence of the attention economy. If you make money by getting eyeballs on your content, then finding a way how to produce it cheaply, fast, and at massive scale will become the main objective. In that sense, we haven't landed anywhere surprising - we're just cruising on the same predictable trajectory, like a comet through space.
While I fully agree with this critique of AI-generated content and share your sense of impending doom for the Internet, your use of “the tragedy of the commons” to build you argument took me aback. Economist Elinor Ostrum debunked this idea (which was never based on data) when Hardin first published his theory in 1968. Her work on the topic finally* won the Nobel Prize in Economics in 2009, making her the first woman to receive the prize. This “tragedy” has no basis in observed fact or data modeling— it’s the theory of one biologist that got traction because it aligned with neo-liberal anti-communism and a “dog-eat-dog” interpretation of human nature that tacitly endorses unbridled capitalism. We MUST stop using it to build arguments, even good ones, and let it whither away . Read a fairly good takedown of this non-tragedy: https://aeon.co/essays/the-tragedy-of-the-commons-is-a-false-and-dangerous-myth
I'm surprised nobody has commented on the obvious yet, which is maybe don't let your toddlers use screens.
The bar is so low, it's in hell.
This, “cognitive micronutrients” yes. Exactly.
“Might not actual human-generated cultural content normally contain cognitive micro-nutrients (like cohesive plots and sentences, detailed complexity, reasons for transitions, an overall gestalt, etc) that the human mind actually needs?”
Well said, Erik. Some of the children's video problem appeared several years ago with the Elsagate phenomenon. I'm not sure we've ever fully determined how much of that was AI generated, and how much was just garbage skits/animated shows from places with cheap labor. Regardless, Elsagate demonstrated that inert, unsophisticated child viewers could be a highly lucrative audience for torrents of low-quality but shocking visual content. Beloved characters like Spiderman and Elsa giving each other injections, playing with feces, weird pregnancy stuff --- the fusion of schizophrenia with Saturday morning cartoons.
And it turns out that children are not the only audience for this swill. I can't wait until it gets piped straight into our VR headsets. Or, what the hell, right into our Neuralinks.
Nightmare.
I really got it for the first time a few days ago when I went on YouTube to get into a music hole. I do that every few months: I grab a beverage, play a song I like and see where the algorithm takes me — YouTube has served for many years as a source of new music and fascinating reviews by experts that have led me to expert producers’ explainers on why pop music pops, singing techniques of various a cappella groups, a Siberian metal band, and a bevy of new independent artists I’d never have found directly. This last session, I gave up after about ten minutes. More than half of the videos YouTube fed me were AI-produced schlock, mostly consisting of duets between various dead people. I’m so sad one of my favorite pastimes is ruined, but even more, sad for all those artists whose work is being buried in an avalanche of dreck.
Somehow in someway, we need to grow a community of people in every free country where the ethos is to pay for, or only try to engage with, creators who do not use generative AI. That’s what I try to do, a large part because as a creator myself, I believe using generative AI lacks integrity. I pay for your newsletter Erik, not only because it’s brilliant, but it seems like you’re not cheating me with your creations, that you’re not using AI. Now, as you said in your piece, there exists no decent AI detection software so what needs to happen is development of faith and trust in those you believe are not using AI - and an adjacent community of investigators who seeks to expose creators who use AI. Anyway great article, once more you freaked me out lol.
Chur,
The Delinquent Academic
I was skeptical: As a lazy parent, I know there are millions of lousy toddler videos on youtube. (Even Ryan's world is bad and most is worse.) Without AI. But, I stand corrected: This AI stuff manages to creep me out. Scott Alexander was right about the future of movies by AI: "The dumbest possible way to do this is to ask GPT-4 to write a summary (“write the summary of a plot for a detective mystery story”), then ask it to convert the summary into a 100-point outline, then convert that into 100 minutes of a 100-minute movie, then ask Sora to generate each one-minute block." - But that 'future' is now - for the youngest. - I hope the market will soon offer easy ways to block that BS out. If not: tough luck for the kids of less-than-perfect-parents.