AlphaGo’s Creator Says AI May Be on the Wrong Track

The scientist who helped build AlphaGo now warns that much of artificial intelligence may be chasing the wrong idea.

David Silver, widely known for his work on game-playing AI, has emerged with a new billion-dollar company built around a different ambition: creating AI “superlearners.” Reports indicate the effort centers on systems that learn more like adaptable problem-solvers than today’s dominant models, which often rely on massive troves of human-generated data. That stance matters because it challenges the direction that has come to define the current AI boom.

At the center of Silver’s argument sits a basic question: should AI mostly absorb patterns from existing human knowledge, or should it learn by exploring, adapting, and improving on its own? Sources suggest Silver sees the second path as more powerful in the long run. That view aligns with the reinforcement learning tradition that helped produce landmark systems in games, where machines advanced by trial, error, and feedback rather than simple imitation.

The debate is no longer just about how fast AI improves; it is about whether the industry has mistaken scale for intelligence.

This is more than an academic disagreement. The industry has poured enormous money and computing power into large models that predict, summarize, and generate with striking fluency. Silver’s new venture suggests a rival thesis: that raw scale may hit limits, and that the next breakthrough could come from systems that learn more efficiently and independently. If that bet proves right, it could reshape how companies spend capital, how researchers define progress, and which technical ideas gain momentum next.

What happens next will matter well beyond one startup. If Silver can show that “superlearners” outperform today’s AI in meaningful ways, he could push the field toward a new competitive race over how machines learn, not just how much data they consume. If not, his critique will still sharpen a question hanging over the industry: whether the current path leads to deeper intelligence, or just bigger machines that sound smart.