Just keep stacking layers baby
Actually stop at k-1 layers where k represents the number of layers necessary to automatically rederive marxism from first principles and put forward a plan to inevitably free the working class from the yoke of the owning class
oh fuck oh shit boss we hired shoddy coders that started all their integer arrays at one instead of zero! we accidentally coded for k layers! the LLM just sold your fifth wife’s yacht collection and repurposed the funds to initiate a carnation revolution in the US Army! we built the communism engine boss what are we gonna fuckin do
we hired shoddy coders that started all their integer arrays at one instead of zero!
This is matlabism
I don’t know all the ins and outs of ensemble of experts, but I do know that relying on multiple general domain models is not how you do ensemble of experts.
This sounds like someone trying to come up with the most expensive solution then working backwards for why you’d want it.