The human race might be completely eliminated if adequate precaution is not taken. Intelligent robots capable of learning from experience will be such that the more they know, the faster they can learn. They might obtain too much knowledge in a short time and become competent with or better than humans. They might become capable of reproducing themselves by using spare parts.
All the time the initial command must be given by a human to a robot. But once given an initial command, the robot may be free to carry it out in any manner, creating for itself sub-desires if necessary. Now under some command, a robot might consider it efficient to execute it, by first getting rid of all humans. The command itself need not be so bad.
Such a disaster can be avoided by the following means -
It may, however, not always be possible to incorporate this due
to forgetfulness or for the sake of war or crime.
It is highly desirable that these robots are the first to be built.
This is because the knowledge of a robot might increase too rapidly. So
the earlier it is built the more powerful it will be.
A person who can create the best robot to protect himself will be the fittest to survive. Such a person will exist because of the wide genetic variation in humans which is due to evolution.
So the human race will mostly not terminate, if it passes through the initial phase when there may not be protecting robots.
The probability of the human race continuing is high if the protecting robots are created early. Otherwise the same probability is not at all high. Infact it may be low.
Now let us consider if it is possible to incorporate the concept of evolution in robots. Mostly humans will not allow robots to evolve by themselves because they will be aware of the dangers. But when destroying robots arise, they will start making robots of their own. They might then incorporate the concept of evolution. But it will not probably be the same as natural evolution which relies on randomness.
When a robot is made, care will be taken to ensure efficiency in all respects. If a random change is made to that design then the result will not be better than the original.
Hence the evolution will be different in that, the robot will try to make a robot with much greater efficiency, using all its knowledge.
But inspite of the knowledge of the robot, the random evolution of humans may be better because it is possible that a human with unusual capabilities never before thought of, will be born, and put an end to the wicked robots.
But a wicket robot who knows this will then introduce the idea of genes into its next generation of robots and incorporate random evolution on the genes.
Then the victory will be that of the wicked robots, it seems, because the robots are capable of reproducing faster and living longer. Soon humans will be outnumbered.
But are we humans foolish? We will then incorporate genes and random evolution into the protecting robots, and then it will again be equals for equals.
Actually there is not much cause for alarm. Whatever may be the case, humans will not be exterminated by machines. The reason is this - Humans will incorporate more and more features of machines in themselves; by putting chips in the brain or something like that. This will enable, them to have the best of both worlds. They will be humans, but they will have all the powers of machines such as speed and accuracy.
This will be the case unless some grave mistakes are made like not putting machine features in humans due to ethical reasons. Having read this argument, it would be more ethical to put machine features in humans and more research should be directed in this way than in Artificial Intelligence (which tries to add more human features into machines).