What was 60 Minutes thinking, in that interview with Geoff Hinton?

Abstraction: Gary Marcus rebuttal annotating 60 Minutes Hinton AI interview

Key points:

• Gary Marcus annotates the CBS 60 Minutes transcript line-by-line, sharply disagreeing that current LLMs "understand" or are "intelligent" in a deep sense
• Hinton claims AI systems have experiences "in the same sense as people" but also says they lack consciousness — Marcus flags the self-contradiction
• Marcus critiques the house-painting ChatGPT-4 demo as cherry-picking a success while ignoring the many documented failures (reversal curse, weight comparisons, fabricated facts)
• The "baby shoes" example Pelley praised is actually a well-known Hemingway-attributed story with 21M Google hits — classic LLM parroting/IP theft
• Hinton's 2016 prediction to "stop training radiologists" went nowhere; Marcus urges caution about similar confident forecasts
• Both agree on one point: enormous uncertainty exists and we are at a genuine turning point requiring careful decisions

Connections: Gary Marcus · Geoffrey Hinton · Chatgpt · Openai · Large Language Models · AI Safety

Source: https://garymarcus.substack.com/p/what-was-60-minutes-thinking-in-that