This is an unacceptable outcome. You can’t have safe AI if you can’t understand or explain decisions. We reject this premise at Mind AI, and we’re developing a different approach with a patented new data structure.

We do expect AI to come up with answers that a human wouldn’t have conceived of, but we then need to be able to retrace the reasoning path of the AI and learn from it, so that we can incorporate those answers into future research/decisions, instead of blindly handing over the keys to AI.