@devinprater @eric Yes if we had a specifically designed and trained model then the mistake part becomes a non-issue. As things stand now, this'll be great until things become inaccurat e enough so as to make the responses useless. See: model collapse.
Matt Campbell (toot.cafe)