Can we stop already? we've demonstrated sufficiently that all known architectures lack the required indicative bias for what the openAI charlatans are Hucking.

Can we stop already? we've demonstrated sufficiently that all known architectures lack the required indicative bias for what the openAI charlatans are Hucking. Sure if you scale a dense feed forward neural network enough it can do vision just fine, but the lack of inductive bias means the model becomes so inefficient that you will _never_ be able to extract any net gain from it. So it is with transformers. Fantastic ontological cartographers, but not cognitive / logic engines like promised.