The debate over the chances to create artificial intelligence is ongoing. Many people have doubts based on the four reasonings as described by Dennett. Then there is a second level of controversy created by people like Dreyfus, Toulmin and Penrose and Dennett as described above. Nevertheless new efforts have started to construct an artificial robot; the most famous of them called COG. An attemp to create a (semi-)conscious robot by Rodney Brooks and Andrea Stein of MIT. The ideas used in the construction are based upon the theories by Dennett and on the previous work by Brooks. COG is possibly the first serious attempt to create a humanoid. It has an upper body similar to humans: arms, hands, a moveable hands which can all move very humanlike. It has eyes that can fixate at almost human speed and ears. Every part is full of sensors and limit switches that control the movements: it gives 'pain-signals' (for example when he turns his neck too far) and sensitivity (so it won't crush your hand while you shake it). Besides all this hardware it will have several rules embedded in its software. Some of them are extremely practical in nature, for example visual face recognition, signal analyzing software (in order to learn our language) and a speech synthesizer. COG will have a motivational structure: it should be human in the way it responds to wants, fears, likes and dislikes. It should have goal-registrations and preference-functions. This in order to create delight in learning, strive for novelty, abhor error and recognize process. It must be curious but careful. These should all be simple rules, like the ones that control the cellular automata in the game of life. From them, behaviour should appear. It is a bottom-top approach and basically COG will start out with an 'empty hard disk' like a human child. Progress is dependent on his ability to process information and recognize patterns outside and within itself.
Some of the criticism described in the previous chapter has been taken at heart in the idea of constructing COG. Dreyfus' claim that sensomotoric intelligence should be an important factor is a major focus in the construction of COG and in this respect, sidestepping the epistemological assumption as put by Reynolds. As a principle COG is supposed to 'learn' common sense. According to Dennett, in a similar way as children do, based uopn his consciousness order. More specifically, the design is not fully based on the biological assumption or the psychological assumption; important parts of the information are not processed symbolically but as an intrinsic feature of the senses of the robot. The rules to do so are not formal; content is important for the way the signals are processed. Rodney Brooks' approach to avoid a cognitive bottleneck seems to address a great part of the criticism put forward by Dreyfus.
'Climbing the consciousness tree' by learning from and interacting with the surrounding world also means that COG will partly be a product from that world. In a sense, the builders serve as 'parents' and pass on values and believes to the robot. It seems that some of issues raised by Stephen Toulmin are being addressed as well by the approach taken by Brooks and Dennett. However processing the incoming information is still largely algorithmic and in this sense the criticism raised by Toulmin is not circumvented.
It will have to be seen what the results will be from the current projects into AI. In any case it will not only give us valuable information about the way that our brain works (or does not work), it will have many interesting side products, like the speech analysis. In one way or another it will influence our lives tremendously.
It is very interesting to note that the ideas underlying AI and the discussion about its consequences are now also the heart of discussions in seemingly completely different fields. I would like to mention one of those here: evolution theory. To show the link a quote from Dennett:
'Darwin's idea had been born as an answer to questions in biology, but it threatened to leak out, offering answers, welcome or not, to questions in cosmology (going in one direction) and psychology (going in the other direction). If [the cause of design in biology] could be a mindless, algorithmic process of evolution, why couldn't the whole process itself be the product of evolution, and so forth all the way down? And if mindless evolution could account for the breathtakingly clever artifacts of the biosphere, how could the products of our own 'real' minds be exempt from an evolutionary explanation? Darwin's idea thus also threatened to spread all the way up, dissolving the illusion of our own authorship, our own divine spark of creativity and understanding.' This materialist reductionism point of view by people like Richard Dawkins, William Hamilton and Dennett is now under heavy discussion. There is a large opposition from well known scientists (Noam Chomsky, Roger Penrose, Jerry Fodor, John Searle and Stephen Jay Gould). They deny the capacity of natural selection to produce specifically human mental qualities like the capacity for language. They accept evolutionary naturalism, but regard it as no more than a place holder for a true explanation of specific human capacities. For various reasons they do not want to extend the Darwinian principles to the full extend. Whereas Dennett and Dawkins and other see consciousness, just like any other phenomenon, as a mere side effect of evolution (of complexity, self-organizing powers of complex systems, etc.) the opponents refuse the make this step. It is very similar to the discussions over AI and have the same basis.