Gary Marcus: GenAI's inability to follow instructions is the core trust problem

GaryMarcus · x · 2026-09-05

Responding to a challenge about his stance, Gary Marcus doubles down: he has spent years arguing GenAI is unreliable and can't be trusted, and that failure to follow instructions is the core issue — he likened ChatGPT to a bull in a china shop back in 2022: powerful but reckless and hard to control.

Original post →

More from AGI Musings

AGI Musings channel →