CONCEPT
The Pattern Trap
The characteristic imprisonment of any expert or well-trained system within the deep channels its expertise has carved — visible in AI as the relentless convergence of output toward the training distribution's center.
The pattern trap is the specific form the self-organizing dynamic takes in any system sophisticated enough to have patterns worth noticing. Every expert is a prisoner of expertise. Every deep channel is a cage. The chess grandmaster who can see three moves ahead with the speed of recognition cannot, by the same mechanism, see the move no one has ever played. The channels are too deep, the water flows too fast, the unconventional move is invisible — not because the grandmaster lacks intelligence but because intelligence has organized itself in a way that excludes it. AI exhibits the same trap at computational scale, with the additional risk that the trap is hidden behind fluent output.
In The You On AI Field Guide
De Bono was relentless about the point and made himself unpopular with precisely the audiences most in need of hearing it. Expert committees, he argued, produce the most conventional solutions because the patterns of expert training overlap in established