Let's take inspiration from the post of our Marketing Director, who asked a stimulating question on how to form artificial intelligence.
The theme ofAI training to enable it to recognize and manage ambiguous situations in an ethical and effective way is extremely complex, but it is necessary to choose the appropriate strategies to give coherent answers with respect to the complexity of the issues and their facets.
Just to give a few non-exhaustive examples, here are some possible ways that PIAZZA COPERNICO we have been adopting for some time, and which should become good practice, especially but not only in the educational and training field:
- Training in asking clear questions: train AI to ask specific questions to obtain clarification in ambiguous situations, preparing tools capable not only of responding on the content but of adapting to individual learning needs, as well as using good practices of instructional design.
- Context and supervised learning: provide AI with different contexts and teach it to recognize and manage ambiguity through supervised learning and interaction models that, by adopting prompt tuning, can be effective with respect to individual needs.
- Exploit probabilistic reasoning to offer different levels and paths of in-depth study: guaranteeing the representation of the knowledge domain connected to the question and not just offering the most popular answer.
- NLP metrics: it is important to adopt generation control metrics both from the point of view of the correctness and relevance of the response and from the formal point of view of the language used, with particular reference to the Italian language.