If (1) an LLM is able to arrive at solutions in the same class of difficulty as the solution for the target problem and (2) it's not possible to establish the provenance of the solution actually offered by the LLM, then what's the argument for assuming that the solution is based on IP rather than constructive reasoning?
If (1) an LLM is able to arrive at solutions in the same class of difficulty as the solution for the target problem and (2) it's not possible to establish the provenance of the solution actually offered by the LLM, then what's the argument for assuming that the solution is based on IP rather than constructive reasoning?