
Stanford's Alpaca Chatbot Was Benched for Hallucinating Facts It Had Never Checked
Ourtake
Hosting costs, weak content filters and made-up facts. The chatbot's demo lasted about as long as its grasp of reality.
What happened
Stanford researchers took their public Alpaca chatbot demo offline about a week after launching it. They cited hallucinations, rising hosting costs and inadequate content filters. The bot was built on Meta's LLaMA model.