Mechanical intelligence is extremely limited in the early stages of robotic technology. The best PL 5 robots have processors only as advanced as computers, and they are little better than remotes. If a situation falls outside the conditions for which the PL 5 robot was programmed, the robot doesn't know what to do and sees no reason to take any actions at all.
At
Progress Level 6, robots step closer to achieving true artificial intelligence with the invention of the first commercially viable neural networks: "learning" computers. Designed to mimic how an organic brain processes and stores information, the neural network allows the robot to analyze the data it receives from its
sensors and make autonomous decisions based upon that data. In other words, a neural network allows a robot to think.
However, true artificial intelligence does not arrive until
Progress Level 7. While neural networks allow robots to learn and think, artificial intelligence allows robots to plan and be
creative. Further, the AI attaches appropriate significance to what it learns; not only can it create but also it can decide for itself whether doing so is a good idea. In effect, artificial intelligence allows a robot to simulate humanoid behavior (for better or for worse) without being programmed to do so. It learns by observation and deduction, not unlike a human child learns to behave as the adults he knows.