The John Batchelor Show podcast

51: PREVIEW. The Risks of AI Feedback Loops: Why Feeding Grok Its Own Output Causes Insanity. Spencer Klavan details interactions with the Grok chatbot, explaining that it learns and improves by receiving human input. Chatbots often end answers with questions

0:00
2:30
Rewind 15 seconds
Fast Forward 15 seconds
PREVIEW. The Risks of AI Feedback Loops: Why Feeding Grok Its Own Output Causes Insanity. Spencer Klavan details interactions with the Grok chatbot, explaining that it learns and improves by receiving human input. Chatbots often end answers with questions to encourage user responses, which serve as training data. Critically, Klavan notes that if Grok's own output is fed back into the system instead of new human output, the machine actually "goes insane."
11963

More episodes from "The John Batchelor Show"