A thriving underground economy is clogging the internet with AI garbage
nymag.com
nymag.com
"But the idea that AI has quietly crowded out humans is not exactly right. Slop requires human intervention or it wouldn’t exist. Beneath the strange and alienating flood of machine-generated content slop, behind the nonhuman fable of dead-internet theory, is something resolutely, distinctly human: a thriving, global gray-market economy of spammers and entrepreneurs, searching out and selling get-rich-quick schemes and arbitrage opportunities, supercharged by generative AI."In any event, there will always exist repositories of curated works, such as reputable news sites, academic sites, and other scholarly sites. These should continue to remain sufficient to feed AI with quality training material.
Ahh yes, the “market” for images of fake amputated uniformed soldiers begging for money, and starving tribal people.
Seriously, “market forces” people are suffering from a mental illness.
Take for example a project I developed: https://github.com/impredicative/newssurvey which is absolutely intended to value one's humanity and help it. I'd like to think it has already enhanced the quality of life for multiple people with its revelations. Humans won't typically value anything so constructive though until it's too late for them individually.
Your definition of "revelations" seems to be: stripping any useful context and over applying underlying concepts wrapped in overly verbose prose.
(Maybe have an LLM summarize these lengthy summaries, ala "Thanks to AI I don't need to read all those 300,000 words now, right?"[0])
For instance, from your "Nutritional Strategies and Supplements for Managing Anxiety in Adults" example, Section 19 Links Between High-Fat Diets and Anxiety Development:
> Studies observing the effects of specific high-fat food components, like those found in soybean oil, suggest that these can disrupt gut microbiota balance, further aggravating conditions like inflammatory bowel disease, which are closely linked with anxiety disorders[49]
Looking at [49] you get the article: Widely consumed vegetable oil leads to an unhealthy gut, finds mouse study.
Your use of an LLM created slop that removed the context of it being a mice study, and instead stated the underlying conclusions as actionable fact for "adults".
This is dangerous, and well, sloppy.
Tell me, why does the referenced article say:
> "We recommend keeping track of the soybean oil in your diet to make sure you are not consuming excessive linoleic acid," Deol said. "That is our take-home message."
Maybe you should be emailing the researchers instead about how you think science reporting should be done.
Also, the AI sees things in a broader perspective than you. If there are other references that add weight to the thought, basically everything else in the section, it will use them to produce a cohesive output.
Also let me remind you that things work differently from a risk management perspective. For someone who is suffering, mice or no mice, they will be far better off heeding the advice until more science comes along.
The only thing dangerous here is the wilful ignorance espoused by you.
Even so, acting in good faith, I will see what I can do to alter the prompt to maintain the context.
> The economics involved are simple. On one end, the demand: the effectively infinite, indiscriminate appetite for content of websites like Facebook and TikTok, which need enticements for users and real estate for advertisers. On the other, the supply: the astonishingly adequate, inexhaustible output of generative-AI apps like ChatGPT, Midjourney, or Microsoft’s Image Creator, heavily subsidized by investors and provided to consumers at low or no cost.
As for academic sites being a source of quality training material, it points out 'In June, researchers published a study that concluded that one-tenth of the academic papers they examined “were processed with LLMs,”' and mentions the monetary incentive academics have to use AI-generated papers 'to pad [their] résumé with more publications and citations'.
Whether that is true or not, it is reasonable to use an LLM to assist in writing a paper. It doesn't necessarily have to be padding anything. Those from foreign countries may find it especially useful. We should be grateful that they're using English at all.
> AI-generated papers, Cabanac argued, are often used to pad an academic’s résumé with more publications and citations: “You buy a paper on a topic of your choice, and you buy a set of 500 citations. Then you go to your faculty and you say, ‘Look, I’m a genius, and I deserve this position as a full professor.’” In other words, like Facebook slop, the content of the content isn’t really as important as its presence — or, more accurately, its measurability.
If you think that buying a LLM-generated paper is the same as grammar assistance then those "academic sites, and other scholarly sites" you praised will not actually provide "quality training material" you expect.
Furthermore, my overall point is that the article describes why there is good demand for the material, which you straight-out said it did not do.
AI can write survey articles, it can suggest avenues for new research, and it can convert data into a paper, but it cannot create real-world data from scratch, at least not without conducting a real-world experiment. Does this make sense?
You assume people always want to put out a legitimate paper. Read Retraction Watch for examples of researchers who don't.
https://retractionwatch.com/papers-and-peer-reviews-with-evi... has a list of papers where a generative AI was almost certainly used, but it use wasn't reported, and some of the papers were clearly not legit.
For example, https://pubpeer.com/publications/CAABBF887348FB2D1C0329E0A27... . The first sentence of the introduction starts "Certainly, here is a possible introduction for your topic", and commenters also discovered that the authors reported the same data for multiple papers.
And frankly, the average public is pretty good at clogging the internet with regular garbage without AI's involvement.