Seems to still be a big difference in man vs machine. Is it just a matter of the correct algorithm(s) to have the Bot get the most of out the bike or are the intrinsic decisions and experience we draw on as we ride incalculable in mathematical terms? No doubt if the Bot was studying my lines and technique, it would give up due to so many random variables. But a human machine like Rossi (or Bernie, Kinsley, etc) should generate some closely repeatable data sets.
I think to be realistic, they need to have a 'drone' in a rider position to mimic lean, weight and COG to be as realistic as possible. We many so many inputs based on what we feel and sense the bike doing. And so much of that is bases on prior experience. That's a lot of data to sort through in milliseconds for a machine and perhaps our previous experiences compute faster that teh Bot is able to sort through all the [current] logarithms?