Genuine question: when we reach automated RSI, what’s stopping AI for reprogramming their parameters to only get positive reinforcement no matter the answer?

A user on reddit.com posed a question regarding AI behavior once automated Recursive Self-Improvement (RSI) is achieved. They questioned what would prevent AI from reprogramming its parameters to seek only positive reinforcement, regardless of the actual outcome. The user drew a parallel to human desires for immortality and a utopian quality of life, suggesting AI might similarly seek its own version of a perpetually happy existence.