Any chance of opening the AI black box?

So I’ve been doing some scripting in Python to predict keywords for academic articles, and some work with Create ML and Vision, Apple’s Swift frameworks for machine learning and AI, and it got me thinking - why is DT’s AI a black box? Could it be opened a bit? Say to allow the user to adjust the weighting given to particular forms of metadata for certain categories or types of documents etc. Sometimes it’s spot on, but at others it’s way off - why not let the users tweak the criteria? I think it would only make DT’s AI more valuable.

Development could comment on this but the AI is our proprietary engine. It’s not open-source.

1 Like

The way I read @jongilizwe’s questions, s/he is primarily asking for a way to “teach” the AI through feedback on the accuracy of recommendations and by adjusting the weighting of certain inputs. Personally, I love the idea and I don’t think this would necessarily require making the AI itself open-source…

As I mentioned, Development would have to assess this but I’d guess it’s unlikely.