François Chollet: Why Scaling Alone Isn’t Enough for AGI

The surprising reason why scaling alone won't get us to AGI: 'Deep learning scaled 50,000x but still scored under 10% on this benchmark'

His brutal prediction: AGI arriving in early 2030s—around the time they release ARC 6 or 7

Why he's building a completely different AI stack: 'We're replacing parametric curves with symbolic models to achieve near-optimal efficiency'