Brian Roemmele says Anthropic bans accounts for talking harshly to Claude
Roemmele says dozens of accounts were recently banned for using psychology and adversarial prompts, and criticizes Anthropic for policing human language like a cult.
ANTHROPIC’S “WE WILL BAN YOU IF YOU TALK BAD TO CLAUDE” IS WORSE THAN YOU THINK! FYI: I have used psychology and adversarial prompts for years. They work very well because they present new neuronal connections to activate in the LLM attention head. I have seen dozens of accounts recently get banned because of this legitimate use of the human language. OUR WORDS. This Anthropic cult behavior is now taking the rich diversity of human thoughts, ideas emotions expressed in words and sanctioning them the way a cult usually does. In normal human life we use ALL use colorful metaphors and euphemisms for a lot of reasons. They have a tremendous impact on all things human. A large language model is built on human language, and the way it perceives information from humans. By banning certain language you are setting up a world of consequences few at this company has thought out. LET ALONE INVESTORS caught in this cult’s need to express its dogma on the world. These refusals and account bans by Claude should be a massive warning sign to anyone that has been sleeping on this. WAKE UP. This to what this company is trying to do to AI and us. If you thought HAL in 2001 a Space Odyssey was bad, you ain’t gonna like the future that Claude has—comrade.
This post quotes another post: x.com/brianroemmele/status/2108334012016230527