@nigini I suspect that might be hard to achieve given that the commodity models do not do this at present. But can you think of ways to easily bootstrap fair data sourcing without having to do large-scale model training?
@ntnsndr Very true @ntnsndr. At least for now, it is unreasonable to set a training pipeline. I think, though, that to make chatbots really useful, they need to be fine-tuned to certain purposes (and maybe that is a first opportunity.) Another high-level service that would be valuable and could use high-quality-fairly-sourced data would be versions of KOI Pond. ;)