One of the main problems of an artificial intelligence system is how to develop self-awareness. Current AI systems can analyze some complex problem and provide a simplification of it. They are simulating two processes of the human mind that we are going to analyze now.
– Deduction: From a general law, to provide a more particular result. This is the common way of working done by an expert system that uses laws provided by an expert and the inference engine provides a value for a certain group of input variables.
– Induction: From a set of particular cases, to provide a general law. This is the common way of working done by an MLP, for instance, that constructs a function that provides a certain output for a group of input variables related to a set of groups of known input and outputs during the training process.
These processes are useful for practical applications but there are very simple to cope with the most complex problems. AI systems cannot have the feeling that they are individual people.
It could be interesting to think about this fact, and how being aware of the self can be explained. What kind of natural processes is driving life towards intelligence?
I propose you some clues to think about how the self is characterized by some aspects that any advanced AI system should consider. I know that can result odd for many people but it can help us to make a reflection about AI:
– We know our existence because we are interacting with our environment. This is the opposite of the Cartesian concept. The need of determining that we are something out of the environment drives us to be aware of our existence instead of the own thinking. Thinking is a process got through many years of the existence of individuals in a process of natural evolution instead of the existence a consequence of the thinking.
– The aim of thinking and of the sense of self is to try to reduce the uncertainty of our existence. Universe is a hostile environment to preserve order and life is a kind of order. We know that its entropy is increasing due to the second principle of thermodynamics. Life can be seen as a system that is fighting against this natural trend, it is a process of entropy reduction that requires the counterpart of increasing the entropy of the environment to fit the second principle of thermodynamics. Intelligence is a phenomenon of life to make this process more effective.
– Intelligence is a process that increases the complexity of the systems. If we look at the evolution of live beings, intelligence is the result of a huge increase in the number of neurons and their interconnections.
– The sense of self proceeds from identity and this one from coordinated actions. Can a group of people have sense of self? Imagine a single person, if he wants to be integrated in a group of a person, his own brain is aware of the identity of this new created group and the group itself is aware of the existence through the mind of the member. If the group is converted in a couple, for instance, a marriage, the couple will have an identity (as marriage). They share actions to improve their lives and the couple has an identity. Is the couple itself aware that marriage exists through its members? We could move from this point to a family, a village, a country. This reasoning can be similar to what Hegel did, however, I do not share his opinion because when the number of members of the group increases the degree of coordination of the actions is being reduced and identity is weaker, and then the sense of self would be being lost.
– The sense of self requires a structured way of acting. From the previous point we can understand that a coordinated way of acting among the different parts of the system is totally needed. This fact implies that information must flow through the system. In any communication channel, information is degraded because there is a loss of energy during any transmission. A structured system to transfer in more efficient way this energy will produce better coordination. Nature is more efficient to provide this effect between two neighboring neurons than between two unicellular beings. This physical disposition is important for the sense of self. Back to the couple, in a marriage two people is living together in coordination; however, for a man with a girlfriend in another country it is more difficult to live in coordination.
– Although all the natural intelligence systems that we know are analogic, digital technology improves the capability of systems for AI. Digital transmission is more robust in a noisy environment. This fact lets to implement robust structures even at long distances: Neighborhood is not needed.
Summarizing, perhaps, we cannot get a system with self-awareness because we are focusing the problem in a wrong way. We would not need to replicate the schemes of human reasoning to imitate deduction or induction because these are high level processes, we need to produce systems that act in a coordinated way, that adapt themselves to the environment in order to preserve their own structure, and that can share information in an effective and efficient way. This can be related to the research on intelligent agents. When this kind of systems can control themselves the increase of their own complexity to fit well the environment (probably imitating the former high level mind processes) in order to preserve their own existence, they will be nearer to be an AI system with self-awareness than those ones we are using nowadays.