Anyone else feel like this idea of commercializing GPT-3 is bound to go nowhere as the research community figures out how to replicate the same capabilities in

Anyone else feel like this idea of commercializing GPT-3 is bound to go nowhere as the research community figures out how to replicate the same capabilities in smaller cheaper open models within a few months or even a year? Not to mention, what commercial applications actually require the model to be so few-shot? (see eg this recent paper that achieves similar results with a bit more data https://arxiv.org/pdf/2009.07118.pdf https://arxiv.org/pdf/2009.07118.pdf)