Post

How do people who advocate for expanding the use of LLMs then make jokes about how much LLMs hallucinate in certain tasks, and then continue to advocate for using them in those tasks with fewer and fewer guardrails?

26/8/26, 1:55 pm0 reposts1 like

Bryan Robinson

@brob.dev

26 Aug

Advocate: I told it to do X. It did Y. Y used a LOT of tokens. When asked why it did it wrong, it apologized.

Bryan Robinson

@brob.dev

26 Aug

Advocate: Let's use LLMs to analyze this data. * Advocate shares the results uncritically * Advocate: I used an LLM to read this other document and summarize. It hallucinated 5 times and made up dates. It's the worst. Ha ha ha. So silly. Advocate: Let's base decisions on what the LLM data analsys