TLDR

Rohan Paul posted a quote from Chamath Palihapitiya at the Stanford AI Club. Chamath stated that long-horizon tasks do not work and called them a joke. He added that current systems also fail at complex problems. Gary Marcus replied to the post with agreement. Marcus later quoted the same post and wrote that he had been right. The posts record Chamath's assessment of present AI limits and his mention of an impending hype cycle.