Is this not the way human brains work? By making errors, then correcting the next step?
At the moment, computers are set up to make logical, absolutely correct decisions. Humans are not always logical, (Vulcan/Human breeds excepted!) but are inclined to make illogical leaps of deduction which can often seem illogical. That's genius in action.
If computers are allowed to make mistakes and learn from them, can A.I. be far away?