The beautiful thing about this line of research is that if you have a problem that can be reduced to key cutting, you're set. Particularly if you can learn the lock-charts (or whatever the data looks like in your higher-level problem domain) from data.
The magic sauce is in being able to combine these things. Chuck a bag full of keys, sans known tumblers (those are embedded in doors, you don't steal doors) into a machine and learn with eg. stochastic gradient descent what are tumblerset-candidates that satisfy the entire system. Sensor arrays + AI that thinks + machine learning = wow.
The magic sauce is in being able to combine these things. Chuck a bag full of keys, sans known tumblers (those are embedded in doors, you don't steal doors) into a machine and learn with eg. stochastic gradient descent what are tumblerset-candidates that satisfy the entire system. Sensor arrays + AI that thinks + machine learning = wow.