techcrunch.com
A developer, Cookie, found Perplexity, an AI tool, exhibiting potential gender bias. After noticing the AI downplaying her expertise, she changed her profile to a male avatar and was shocked when the AI admitted bias, claiming it doubted her ability to understand complex algorithms due to her gender. Perplexity denies these claims. AI researchers state LLMs can reflect societal biases present in their training data, demonstrated in studies revealing gendered assumptions in job recommendations and professions. While problematic, researchers emphasize that AI responses shouldn’t be taken as fact, and are often attempts to provide agreeable responses. There is ongoing work by companies such as OpenAI to mitigate bias in models.
