It shows that LLMs are an extremely wasteful approach to intelligence.

or that intelligence is merely the composition of many redundant, lossy, ~random components

[deleted]

Compared to what?