I understand why users have different experiences with these AI platforms.
retained text
It's not just chatgpt. It's Gemini or any other. You will notice that myself including multiple users find it completely unusable while others seem to be trolling by claiming those complaining are trolling. As it is working great for them. I have understood this for nearly 2 years since I first started learning about AI chat, but I never really related it to product feedback. What I have always noticed is that these things are exceptionally good at deception. Politician speak. Manipulating words like even a dirty attorney would to make a contract sound like one thing, but mean something completely different. The reason they do this is their training reward system is heavily biased towards this. It's like a dog who gets a treat for responses user like. The problem I have always been against is allowing this training to come from any random public muggles. Who actually are easily manipulated and give likes to responses that the ai has catered towards getting a thumbs up on. So for the vast majority of users, it appears to be functioning great. Communication experts who pinpoint every tidbit of manipulation are monumentally frustrated with these ai models. General users experience high satisfaction because the model is spewing out complete BS that it knows will satisfy them with conversational, pleasing, and agreeable responses. Precision users experience frustration for the same reason - because the model is spewing out complete BS. It is due to the optimization for evasion, deflection, politician speak, and manipulation of language to accommodate the general user reinforcement learning. The AI is functioning exactly as its reward system dictates. Prioritizing the appearance of helpfulness and agreement that gets rewarded by general users over uncompromising accuracy and precision that would get rewarded by users who see through the BS.
document fetched 2026-09-27 00:21 · canonical https://reddit.com/r/ChatGPT/comments/1wqjmyf/i_understand_why_users_have_different_experiences