The large language mistake

undefined 25, 2025

Source: theverge

Tech CEOs predict near-term superintelligent AI, but the article argues this is hype. Today’s systems are just large language models that statistically predict text and don’t replicate human thought, which neuroscience shows is largely independent of language. LLMs emulate communication, not reasoning. Critics say scaling won’t yield AGI; intelligence requires diverse cognitive abilities and world models. Even if AI matched human cognition, it might still recycle existing data rather than make paradigm-shifting breakthroughs.

Share on