Human-Feedback for AI in Industry
Résumé
Artificial Intelligence (AI) offers a wide variety of opportunities to the manufacturing industry. However, there are still gaps and challenges to be solved before it can be successfully applied, with data availability and quality being one of the critical factors. The latter highlights the necessity of developing AI systems that can continually learn (from one or more domains) over a lifetime, starting from limited sets of data. This work presents research done on human reinforced learning approaches on small training data sets of open dynamic environments. Beginning this way allows the development of AI models able to learn over time, while taking advantage of a data driven approach along with a knowledge-based approach considering human-feedback as a key enabler.