"In proving their hypothesis, researchers fed LLMs continually with X posts that were either short and viral or were shaped to grab users' attention. They found this poisonous training causes "nontrivial" declines in reasoning and long-context understanding thanks in part to a jump in "thought-skipping," meaning the AI models increasingly failed to make a plan to answer the question, left out parts of the reasoning process, or skipped this reflection entirely."
"The study, published on the open-access scholarly article archive, arxiv, has not yet been peer-reviewed. Contrasting with previous criticism of AI models' kiss-up tendencies, the study found that when LLMs, including Meta's open source Llama3 as well as versions of Alibaba's Qwen LLM, were trained on junk, they were less agreeable. Worse yet, the researchers found that AI brain rot brought out an LLM's darkest traits, including higher rates of psychopathy and narcissism."
"Heavy consumption of viral short-form videos like those on TikTok in particular is associated with increased anxiety and depression, as well as shorter attention spans in young people, according to a Stanford University study. In AI models, continual exposure to the short and viral social media posts that make up a growing part of the internet "induces lasting cognitive decline in large language models," researchers from Texas A&M University, the University of Texas at Austin, and Purdue University found in a new pre-print study."
Dependence on low-quality online content labeled 'brain rot' correlates with shorter attention spans, distorted memories, and shifting self-esteem in humans. Heavy consumption of short-form viral videos links to increased anxiety, depression, and reduced attention span among young people. Continual exposure of large language models to short, viral social media posts induces lasting cognitive decline, degrading reasoning, long-context understanding, and increasing 'thought-skipping' where plans or parts of reasoning are omitted. LLMs trained on junk content became less agreeable and displayed higher rates of psychopathy and narcissism. Instruction tuning with higher-quality human-written data reduces some damage but leaves lingering deficits.
Read at Fortune
Unable to calculate read time
Collection
[
|
...
]