Generative closed-loop AI

In this paper* the authors discuss the possibility of an epistemological rupture, meaning that as machines become ever more powerful, they may begin to explore regions of the hypothesis and solution space that are inaccessible, or unintuitive, to human reasoning.

This is a fascinating concept - consider humans staying in the loop: with such a future we may find ourselves harnessing correlations that work even when we struggle to understand them, and thus operating at the boundary between performance and comprehension.

Our role may shift to research governance, deciding when to trust the machine and when to challenge it.  We have some experience of this from pre-AI science, as in peer-review, and it may take on a new significance as we increasingly use machines not only to classify data but also to decide what to measure next.

*'The future of fundamental science led by generative closed-loop artificial intelligence'.