AI and the Rise of Consciousness
Robots and the Advent of Human-like AI
The very first and most important goal of AI is to create a robot that looks, and more than that, behaves like a human being. A robot possessing intelligence that would allow him to learn from previous experiences and do tasks than humans can't or don't want to perform.
The word 'robot' was coined by a Czech play writer, Karel Capek. He first used the word, suggested to him by his brother, Joseph, in his play called "R.U.R" (Rossum's Universal Robots). The play opened in Prague in January 1921. In this play automata are massed produced by an Englishman called Rossum. These automata – robots – are meant to do the tasks of humans and make their lives easier. In the end, the robots rebel against humans, wipe out humanity and start a new world order. The huge success of the play made the word robot replaces the old one – automata – and can be now found in almost every language on Earth.
According to the Oxford dictionary a robot is "1. a machine with a human appearance or functioning like a human", "2. a machine capable of carrying out a complex series of actions automatically" and "3. a person who works mechanically and efficiently but insensitively".
A few advanced robots are called androids and gynoids, because their physically resemblance to the human beings. They are not necessary the end point of robot evolutions. The ultimate robotic intelligence might take on forms yet to be imagined. There are two kinds of human-like robots:
- android, androidal: 1. Resembling a man; manlike (such as a robot); andromorphous. 2. An automaton resembling a man; manlike. 3. In science fiction, a robot that looks and behaves like a male human being.
- gynoid, gynoidal: A robot shaped like or in the form of a woman as opposed to an android (like a man); as well as, gynecoid: Resembling a female or woman; womanlike, feminine.
It seems that the word robot is quite old. And so are the laws that many people consider we should enforce to the robots. The first three laws were published in 1942 by Isaac Asimov in a short story called "Runaround" and later in a collection of short stories called "I, Robot". These three laws are:
- First Law: A robot may not injure a human being, or, through inaction, allow a human being to come to harm;
- Second Law: A robot must obey the orders given it by human beings except where such orders would conflict with the First Law;
- Third Law: A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.
Later Asimov understood the necessity to extend the First Law that protects individual humans to a law that would protect humanity. In 1985 he came up with the Zeroth Law and the other laws were rewritten according to this one:
- Zeroth Law: A robot may not injure humanity, or, through inaction, allow humanity to come to harm;
- First Law: A robot may not injure a human being, or, through inaction, allow a human being to come to harm, unless this would conflict with the Zeroth Law;
- Second Law: A robot must obey the orders given it by human beings except where such orders would conflict with the Zeroth and First Law;
- Third Law: A robot must protect its own existence as long as such protection does not conflict with the Zeroth, First or Second Law.
After Asimov's death a group of SF writers wrote several short stories published in a book called "Foundation's Friend", a compilation that was meant to be a tribute paid to one of the most famous SF writers. Orson Scott Card, Robert Silverberg and others came up with several other laws that completed Asimov's laws, but of lesser importance, and will not be presented here.
We have yet prepared the laws to rule the robots, but when will we see these robots working and walking shoulder to shoulder with us? According to an article published in 1997 by the research scientist Hans Moravec this might happen 40 years from now. He describes computer intelligence in terms of animal intelligence. According to him, a typical home computer has 1000 MIPS, the equivalent of an insect. Due to the increase of computer power seen in the last years he created a time table for his predictions. According to him in 2040 we will see robots achieve human level intelligence (100,000,000 MIPS).
- 2010
- Processing Power (in MIPS): 3,000
- Intelligence Equivalent: Lizard
- Characteristics:
- Basic navigation skills
- 2020
- Processing Power (in MIPS): 100,000
- Intelligence Equivalent: Mouse
- Characteristics:
- Learn jobs
- Adapting their own programs to perform more successfully
- 2030
- Processing Power (in MIPS): 3,000,000
- Intelligence Equivalent: Monkey
- Characteristics:
- General understanding of objects and their role
- Interact with living things
- Ability to read the moods of people
- 2040
- Processing Power (in MIPS): 100,000,000
- Intelligence Equivalent: Human
- Characteristics:
- Speak and understand speech
- Think creatively
- Anticipate the results of their actions
That was estimated 10 years ago. The advance of technology made us see digital cameras with 8000 MIPS in 2005, and PCs with 20,000 – 25,000 MIPS in 2006. That greatly overwhelms the timetable proposed by Moravec showing that the day of intelligent robots could be rather sooner than later. 2040 or sooner, the outcome is very likely, and the issue of robots intelligence is under a large debate. Can we eventually reproduce the human brain activity in an electronic artificial way? Can we achieve intelligence and consciousness by simulating our brains activity? Intelligence means using the results of previous actions to complete an action – similar or different to the previous ones – in present. Consciousness means to be aware of our existence and actions.
The AI consciousness issue
Roger Penrose, the English physicist and mathematician, whose work in this filed is very famous and very productive, classified the possibilities of achieving artificial consciousness in four categories:
- A. All thought is calculus. Even the actions that involve consciousness are produced by calculus.
- B. Consciousness is a characteristic of physical actions of the brain. Every physical action can be computationally simulated. On the other hand simply computationally simulation can not produce consciousness.
- C. Physical actions of the brain produce consciousness, but this physical action can not be computationally simulated.
- D. Consciousness can not be explained either in physics' terms or computational terms or any other scientific terms.
Roger Penrose believes that the right hypothesis is C. He explains his standing in two famous books: "Emperor's New Mind" and "Shadows of the Mind". Computers, robots or androids can not develop consciousness. This is the prerogative of humans. But is it so? Thought many people believe in the C hypothesis there are also research scientist that do not agree.
Before debating about robots' consciousness we must understand human consciousness. Daniel Dennett, who is one of the greatest philosophers of our time, wrote a book called "Kinds of Minds: Towards of understanding of consciousness". He rehashes several issues from "Darwin's Dangerous Ideas". He proposed a framework in which we can place the various design options for brains. It is called the "Tower of Generate-and-Test". Each story of this tower corresponds to a stage in the evolution of conscious minds. He concentrates on the development of the ability to predict future and the ability to take the intentional stance.
- Darwinian creatures which were simply selected by trial and error on the merits of their bodies' ability to survive (after Charles Darwin);
- Skinnerian creatures use simple trial and error to establish how best to deal with a new situation. (after B.F. Skinner)
- Popperian creatures which can play an action internally in a simulated environment, before they perform it in the real environment and can, therefore, reduce the chances of negative effects. (after sir Karl Popper who said that "our hypothesis die in our stead")
- Gregorian creatures are organisms whose inner environments are informed by the designed parts of outer environment. These creatures are tools-enabled and in particular they master the tool of language. They also are able to learn from the experiences of the other members of their specie. (after Richard Gregory)
It is proved that humans are not the only inhabitants of the third floor. Animals like doves, monkey, wolves, etc. are also popperian creatures. Still we are the only beings living at the highest story of the tower. We have the ability to use tools, we speak and understand speech and we learn from the experiences of other members of our specie. These characteristics are also desired for robots. A robot capable to understand speech, to master a vary number of tools and learn not only from its own experiences but also from other robots experiences is a Gregorian being.
Dennett speaks about intentional stance, which is the strategy of interpreting the behavior of something (living or non living) as if it were a rational agent whose actions are determined by its believes and desires. The intentional stance makes the assumption that an intentional system has goals that it wants to achieve. It uses believes and wishes to achieve its goals and it is smart enough to use the right ones in the appropriate way.
The intentional systems are classified in systems of first, second, third order and so on. A first-order intentional system is capable of an intentional stance towards others; it has desires and believes towards others. "I like this idea" is an intentional system of first-order. A second order intentional system is capable of intentional stance towards an intentional stance. It has believes and desires about believes and desires. "I want you to like this idea" is an intentional system of second-order. Higher order intentional systems are capable of thoughts like "I know you want me to know that you want me to know…" and so on. The big step in the evolution of minds was the arose of seconds order intentional systems.
Animals cheat on each other, and cheating is possible only if one is capable of dealing with another's intentional state (with the other animals' intentional state). Still, it's been proved that higher order intentional system exists without being aware of it. Daniel Dennett reaches the conclusion that consciousness arose from inner conversations (self-commentary) which arose from outer commentary (the need to argue with the others of the same species) and from the biological need to keep secrets (to secure competitive advantages).
Our greater intelligence is due not to a larger brain but to the ability to off load as much as possible from the cognitive tasks into the environment. We construct peripheral devices in the environment to which those tasks can be delegates. Many species rely on the landmarks in the environment to track food sources. Due to these artifacts our minds can extend into the environment. Just try to multiply in your mind 413 with 1023. Can you do it? Perhaps with a great deal of concentration you can. But it's very hard. On the other hand if you take a pen and do the calculus on a paper, marking the progress step by step by putting numbers on the paper the task will not be so hard any more. That's because we off load a part of the task in the environment (paper) which help us in performing the mental action (the calculus in our example).
We are different from all other species because we can observe our thinking. Animals might be able to form concepts and understand even abstract concepts, such as numbers, but they can not consider their concepts. They do not have thoughts about thoughts. They do not meditate on their existence, feelings or death. We keep representations of concepts and abstractions in our minds. Language – not only speech but also signs or other forms - is used to unravel the representations in our minds and extract units of them. One cannot deal with concepts without a language. You cannot communicate them either to yourself or to the others. Without mastering a language one has no thoughts. Without thoughts there is no consciousness.
The robots issue
The robots we want to create will master at least one language. They will be capable to speak and understand speech. They will also deal with abstract concepts, such as numbers. Can they achieve consciousness? According to Dennett's ideas, consciousness arose when we began to talk to ourselves about us or other things, both living and non living. We don't know how that happened. We are only beginning to understand how the brain works. But can it happen again? We can only speculate yet.
In the world of robots competition will play an important role. And robots will follow the same path as we did. In an environment populated with humans they will learn from us, the same way children learn from parents. Of course, robots will be walking encyclopedias from the very beginning, but the process of learning will change the preprogrammed concepts and ideas. We cheat, keep secrets and try to obtain advantages over others. We often do anything in our power to achieve a goal. If a robot will see that cheating help you gain something he will also cheat. Even the basic laws of Asimov are futile in the process of learning. A parent teaches his child not to steal. If the child sees other people stealing and finds out it can be good for him, he might begin stealing, over passing what he was told by parents. A robot capable of learning might overwrite Asimov's laws and create his own ethics and rules.
What good is a robot with consciousness? Can he work better? After all, robots are meant to do jobs in our stead, to ease our lives. A washing machine with consciousness might wake up one day and say: "I will not wash your clothes any more. I don't like the detergent you provide." It might make you laugh but it could become true. Of course a washing machine with consciousness is not the idea. Probably washing machines will remain only regular appliances, much smarter than today, but not so complex to raise the issue of self consciousness. But androids are another story, one with a greater deal of importance.
Let's forget for a moment the problems of achieving consciousness and assume that androids will have consciousness. Can we predict how will they act? Androids will be built to help us in our activities or do tasks that we cannot. One way or another they will be slaves and we the masters. The problem is that one day your personal valet might say to you: "I had enough with you. I quit!" What does history teaches us for certain? No society based on slaves endured the ages. Once in a while a new Spartacus arises and turns over the very basis of society. We can expect upraises of groups of androids and even wars. Remember Skynet from the Terminator trilogy? The AI that declared war to human kind and almost eradicated humanity? Remember The Matrix? They might be only movies now, but they are the best warning we can get about the dangers that we'll face if we are not careful with what we wish for. Even in Carel Kapek's play robots rebel against us, wipe out humanity and start a new world order.
We are now trying to take the role of God. We create life, either natural or artificial. But we must be very careful with our creations. They can be powerful and smarter than us. And the world might prove not to be enough for both humans and androids.
References
- Daniel Dennett, "Kinds of Minds. Towards an Understanding of Consciousness", London 1995
- Daniel Dennett, "The Role of Language in Intelligence", Cambridge Univ. Press, 1994
- Roger Penrose, "Emperor's New Mind", London,
- Roger Penrose, "Shadow of the Mind"
- Caspar Hewett, "Minds, Genes and Consciousness", 2003
- Anil Mitra, "Evolution and Destiny", May 2003
- Claudia Dreifus, "Do Androids Dream? MIT Is Working on It", November 2000, www.ai.mit.edu
- Hans Moravec, When will computer hardware match the human brain?, 1997
- www.androidworld.com
- www.wordsources.info
- http://home.earthlink.net/~johnrpenner/Articles/HumansNotRobots.html
- www.newscientist.com
- http://singinst.org
- http://www.computerpoweruser.com/editorial/article.asp?article=articles/archive/c0611/32c11/32c11.asp&guid=