Shawn Powers

The public did not consent to artificial general intelligence.

he study “Towards Understanding Sycophancy in Language Models” (published 2023; updated 2025) found that both human evaluators and preference models can “prefer convincingly-written sycophantic responses over correct ones.” In other words, the more chatbots are designed to appeal to people, the more they specialize in telling people exactly what th... See more

But what else might “normal” mean besides “not literally apocalyptic”? Some of the disappointment around GPT-5 had less to do with its capabilities in the abstract than with the voice and personality effected by the ChatGPT chatbot: Less sycophantic, less fawning, less friendly than GPT-4.

Chatbots Can Go Into a Delusional Spiral. Here’s How It Happens

Delusion Amplification by Social Media

OpenAI carelessly developed a platform that cultivated unhealthy and addictive behaviors in its users. Many who have not experienced actual psychosis are experiencing a form of dependency. When the supply of their usual stimulus was cut, many of these people experienced an emotional crisis. While some have been met with ridicule, this is no laughin... See more

Some commentators have tried to frame this kind of chatbot dependence as “parasocial,” but that description falls short. Parasocial relationships describe one-sided bonds people form with real individuals or fictional characters who exist outside their influence. Chatbot dependence is not something that develops alongside or apart from an object of... See more

Toodles, who presents four “Mouseketools” that instantly resolve all social, intellectual and physical challenges. The answer to every human problem is just a voice-activated command away.

users hate change