@mozilla@mozilla.social I appreciate the effort to make the web more accessible, but:
1. How do we know this model was trained ethically? Literally every human text LLM I've seen was trained on data that was used without permission of the author.
2. What quantitative testing have you done to ensure that the results of this model are correct? A model that produces incorrect descriptions a significant percent of the time is worse than not having one at all.