Well, diffusion networks seemed to work because they increased the amount of data by training so many examples of added noise and it’s removal. Some similar app

Well, diffusion networks seemed to work because they increased the amount of data by training so many examples of added noise and it’s removal. Some similar approach might be possible with text, too. Or, i guess they will talk to itself, like alpha go played itself.