An interesting look into the way LLMs form an understanding of their users. We can observe some of how they categorize us by gender, age, and wealth — shaping their behavior towards us. Hopefully, greater understanding will help us identify and correct biases; ensuring LLMs remain helpful rather than another reinforcement of harmful stereotypes, or worse.
https://www.theatlantic.com/technology/archive/2025/05/inside-the-ai-black-box/682853/
