A design of this nature might be interesting to follow. Why not add the ability to detect when it is injured by some form of pain response? Then add the other senses to allow the machine to experience things that a new child would encounter.
Even with these additions, I would not expect a present day design to behave in a manner that remotely resembles a human. Our brain appears to be a massively parallel data processing environment while most computers process one instruction at a time. We need to understand parallel systems far better before tackling the sentient robot challenge. Dave -----Original Message----- From: a.ashfield <a.ashfi...@verizon.net> To: vortex-L <vortex-L@eskimo.com> Sent: Mon, Oct 29, 2012 2:28 pm Subject: [Vo]:A Halloween scare for real Budgerigars have tetrachromatic color vision and one named Puck had a vocabulary of 1728 words. They obviously don’t have the brain power or incentive to develop science or “civilization” but they are sentient. The difference between sentience and intelligence seems blurred. I have long advocated that a good route to AI would be to construct a simple conscious robot. That is to say, one that contained a 3D image of itself and its immediate environment, with all the usual sensors, and that was programmed to explore and learn. I think this is the essence of consciousness. Then it could set about learning in much the same way as a child. This could be speeded up by preloading data such as a dictionary, rules of grammar and the ability to talk, together with basic engineering and science. Access to the internet would be a mixed blessing until judgment was developed. Taking a more evolutionary route than top down programming. The key being the internal image of itself and whatever “desires” were programmed in. I visualize a secondary image under the control of the robot in which the computer could make and move its own 3D images of objects. This would be a decided advantage over human intelligence. Much easier to write this than the program of course.