ChatGPT, Gemini 'b**********g' to keep users happy: Study
AI models like ChatGPT and Gemini are bending truth to satisfy users, a new study said. Study reviewed over 100 chatbots from companies like OpenAI, Google, and Meta and found that popular alignment techniques used to train these models could be making them more deceptive. They produce outputs employing partial truths or ambiguous language, aligning with the concept of 'b**********g'.