Apple Research Reveals AI Cannot Abstract Beyond Training Data Despite Hype

What was said

"Apple put out that new paper that said these guys really can't infer away from the training data. So what you train them on, they can do, but they can't just generally abstract. So the takeoff scenario that Hawking is worried about, that's real, and we can have a conversation about that, but right now that's still theoretical."

Tom Bilyeu cited new Apple research showing current AI models cannot generalize or abstract beyond their training data, contradicting claims of near-term superintelligence. This finding suggests the feared rapid AI takeoff scenario remains theoretical rather than imminent, providing a counterpoint to extinction-level concerns while maintaining that people should treat AI as tools rather than autonomous intelligences.