From big data to human-level artificial intelligence
summary
Argues that cheap parallel computation and massive datasets plus advances in deep learning still fall short of human-level cognition, claiming current systems match narrow tasks like Go or speech but cannot match a three‑year‑old at language. Illustrates gaps with Rosie vs Roomba and a robot demo that hides humans in the background sped up 60×, stresses that deep learning = pattern recognition with phenomena like an Oprah Winfrey detector yet lacks the symbolic, reasoning and compositional tools needed for general intelligence.
AI-written summary · about
