Page 195 - Artificial Intelligence in the Age of Neural Networks and Brain Computing
P. 195

5. IT for Human Survival: An Urgent Unmet Challenge    185




                     Because most of the readers of this chapter have already heard about the
                  Terminator risk from AI (which is quite serious), I will focus instead here on a
                  few examples of other threats.


                  5.1 EXAMPLES OF THE THREAT FROM ARTIFICIAL STUPIDITY
                  Most Americans already understand how impossible life would be if there were no
                  humans available on the telephone, and if one had to navigate traditional dumb voi-
                  cemail systems to do anything at all. From observing such systems, and from certain
                  outsourced help lines systems, we already understand that corporations sometimes
                  deploy such systems because of the beliefs and culture of their management, even
                  when they don’t really improve total system productivity at all. Many of us are
                  already familiar with systems which save the government or a corporation a few dol-
                  lars, by creating pains and hassles which waste a lot more dollars and value outside
                  the narrow books of the group making the change. We know about the risks of police
                  profiling systems which are not intelligent or accurate enough to spot people who are
                  truly likely to commit a crime, and end up causing problems and waste for totally
                  innocent people.
                     But even so, a more graphic example of what AS could do may be helpful.
                     In the year 2000, I led the creation of a joint NSF-NASA workshop to evaluate
                  the possibilities for machine learning and neural nets to make it possible and
                  economic to generate energy in space and beam it down to earth [49,50]. The chair,
                  George Bekey, was not only a leader in robotics research, but also the brother of Ivan
                  Bekey, who for many years was NASA’s top planner for long-term future possibil-
                  ities in space.
                     As part of the workshop, top roboticists were asked: Is it possible, as the Space
                  Sciences Institute once proposed, to design a 50 ton payload which could be dropped
                  on the moon, designed to make use of resources on the moon to expand and grow and
                  export material up to lunar orbit for use in construction of power stations and cities
                  in orbit? The experts generally agreed it could be done, simply by extending tech-
                  nologies already emerging for robots to make robots. There would be no need for
                  these robots to outsmart humans; they would need only to reproduce, swarm, and
                  adapt. But with large numbers, it would be impossible to effectively guarantee no
                  natural Darwinian selection.
                     So try to picture in your mind what this would really look like. An entire planet
                  swarming with energetic, aggressive “metal cockroaches,” no more intelligent than
                  earth cockroaches, but far more capable physically. In essence, robots do not have to
                  be as smart as humans to become a threat to human life.


                  5.2 CYBER AND EMP THREATS TO THE POWER GRID
                  A few years ago, I attended a talk by Texas Republican Congressman Trent Franks,
                  where he described what it was like to become chair of that committee of the House
                  which receives all of the secret information from all sources on the most serious
   190   191   192   193   194   195   196   197   198   199   200