[goodmachine]

In order for that not to happen (uninterpretable ML models) some research on symbolic distillation, aka symbolic regression

https://arxiv.org/abs/2006.11287

https://www.science.org/doi/10.1126/sciadv.aay2631