🖥️ **Censored AI Chat Models Hallucinate More, Research Finds**

"_New research finds that the same internal mechanisms used to block ‘unsafe’ responses also suppress factual information, meaning that attempts to align models for safety can backfire by making them hallucinate more_"

🔗 https://www.unite.ai/censored-ai-chat-models-hallucinate-more-research-finds/.

#AI #ArtificialIntelligence #Technology #Tech #Hallucinations #Research #Study