Train Models on Hostility Oriented Slop and You Get Happiness? Nope, Nastiness on Steroids
November 10, 2025
This essay is the work of a dumb dinobaby. No smart software required.
Informed discourse, factual reports, and the type of rhetoric loved by Miss Spurling in 1959 can have a positive effect. I wasn’t sure at the time if I wanted to say Whoa! Nelly to my particular style of writing and speaking. She did her best to convert me from a somewhat weird 13 year old into a more civilized creature. She failed I fear.
A young student is stunned by the criticism of his approach to a report. An “F”. Failure. The solution is not to listen. Some AI vendors take the same approach. Thanks, Venice.ai, good enough.
When I read “Study: AI Models Trained on Clickbait Slop Result In AI Brain Rot, Hostility,” I thought about Miss Spurling and her endless supply of red pencils. Algorithms, it seems, have some of the characteristics of an immature young person. Feed that young person some unusual content, and you get some wild and crazy outputs
The write up reports:
To see how these [large language] models would “behave” after subsisting on a diet of clickbait sewage, the researchers cobbled together a sample of one million X posts and then trained four different LLMs on varying mixtures of control data (long form, good faith, real articles and content) and junk data (lazy, engagement chasing, superficial clickbait) to see how it would affect performance. Their conclusion isn’t too surprising; the more junk data that is fed into an AI model, the lower quality its outputs become and the more “hostile” and erratic the model is …
But here’s the interesting point:
They also found that after being fed a bunch of ex-Twitter slop, the models didn’t just get “dumber”, they were (shocking, I know) far more likely to take on many of the nastier “personality traits” that now dominate the right wing troll platform …
The write up makes a point about the wizards creating smart software; to wit:
The problem with AI generally is a decidedly human one: the terrible, unethical, and greedy people currently in charge of it’s implementation (again, see media, insurance, countless others) — folks who have cultivated some unrealistic delusions about AI competency and efficiency (see this recent Stanford study on how rushed AI adoption in the workforce often makes people less efficient).
I am not sure that the highly educated experts at Google-type AI companies would agree. I did not agree with Miss Spurling. On may points, she was correct. Adolescent thinking produces some unusual humans as well as interesting smart software. I particularly like some of the newer use cases; for instance, driving some people wacky or appealing to the underbelly of human behavior.
Net net: Scale up, shut up, and give up.
Stephen E Arnold, November 10, 2025
Comments
Got something to say?
-
- Subscribe to Beyond Search
Feature archive
News archive
-
