[calf]

People are just outright abusing the terminology. OP's argument would also conclude that a sorting algorithm is not a "real" algorithm because it too can be done by an infinite lookup table.

That said, the general debate is a valid one. Are LLMs just doing fancy statistical compression of data, or are they doing "reasoning" in some important sense, be that merely mechanistic logical reasoning, or "human-level intelligent reasoning"?

For that matter, did the paper authors ever define "Reasoners" in their title, or leave it to the reader?

[FrustratedMonky]

Sure. I agree.

The debate is good. I just don't see how the 'table-lookup' analogy is helping.

Except maybe by helping people see the non-free-will nature of the universe. But seems like people that reject this, are also ones rejecting the 'table-function' equivalence.