The Mars Landing and Artificial Intelligence

In the NASA video called "Seven Minutes of Terror," which famously went viral over the last month, Tom Rivellini, one of the engineers in charge of the landing, outlines its eye-popping difficulty. As he says, the Mars lander had to go "from 13,000 miles per hour to zero, in perfect sequence, perfect choreography, [with] perfect timing, and the computer has to do it all by itself, with no help. ... If any one thing doesn't work just right, it's game over." The idea of having a computer do it "all by itself," with just 500,000 lines of computer code to allow its artificial brain to work, is at the core of engineering agony. After their years of hard work and emotional and monetary investment (to the tune of $2.5 billion), the humans in charge had to leave the most crucial part of the mission to an artificial proxy. And they were not exactly sure if this proxy would work the way they intended, because there was no way to test it completely. This situation illustrated two pressing issues regarding the development of digital servants: our apparently perennial insecurities about using them and whether we are too hasty to rely on them. The idea of creating artificially intelligent proxies to do what humans cannot -- because the job is too dirty, dangerous or dreary, is a surprisingly old one. It goes all the way back to the ancient Greeks and it reappears in every age in slightly different forms. In his "Politics," Aristotle reminds his audience that the blacksmith-god Hephaestus made robot-like serving stands that could move around the banquet halls of the gods by themselves; and then he ponders the idea of making intelligent machines, such as weaving looms, that could "obey and anticipate" the will of their makers. In the Middle Ages, stories appear about famous philosophers who make artificial servants. One such story is about Pope Sylvester II, who was also a very accomplished mathematician and inventor. Medieval contemporaries claim that Sylvester had made a talking brass head that could predict future events and could also outperform humans at mathematics. In Shakespeare's time we have Robert Greene's play depicting the creation of a similarly precocious metal head. This lineage of artificial servants picks up again in the early 20th century most famously with Karel Capek's play of the 1920's "R.U.R.: Rossum's Universal Robots," in which the term "robot" -- a Czech word meaning "slave" or "worker" -- was first used. Rossum's world is one in which Earth's citizens have come to rely on intelligent robots for everything. Except for the ancient Greeks, the interesting commonality underlying all of these examples of artificial servants is the undercurrent of fear and insecurity about using them. The metal heads of the Middle Ages and Renaissance are depicted as unreliable and dangerous: Sylvester's metal servant gives him bad information that leads to his death, the one in Greene's play implodes after a faulty activation, the robots in "R.U.R." destroy humankind and take over the Earth. And even NASA seemed to wish that they could have had humans at the controls of their landing vehicle. Why? Well, of course, none of us really likes to relinquish control of a touchy situation. But, taken as a whole, there seems to be more to it than that: a cultural narrative of nervousness about our own ingenuity. A fear of our collective feet faltering on an ever-faster technological treadmill and an inability to anticipate needed fail-safes to protect us from our innovations until it is too late. Recent evidence of this type of nervousness is clear in some declarations and actions of the inventors who provide us with intelligent technology. As reported in The New York Times in 2009, a group of computer scientists from around the world met to discuss whether restrictions should be made on development of Artificial Intelligence. Their worry was that human control over AI could soon be compromised, given the acceleration of its capability to operate independently. Their concerns were, in part, driven by the fact that the most rapid advances in AI are being made by the military in the form of automated weapons, such as Predator drones. Even optimists in the technology community do not deny the possibility that our machinery may overtake us; many of them, such as Ray Kurzweil, Rodney Brooks, and Kevin Warwick, simply think that we won't mind being eclipsed by our digital servants because we will have already incorporated so much of them into our lives. So would limits on the development of AI help mitigate the dangers of our ingenious devices? Probably not. Rogue groups and nations would just find detours around any roadblocks that a regulatory body may try to set up. But there are alternative forms of regulation that may work. Scientists and governmental bodies could develop protocols for the way AI is built and tested, guidelines for the kinds of fail-safe controls built into AI, conventions for testing it. And, most importantly, governments could devote more money to research into non-military forms of AI, so that benevolent advances could balance out the more dangerous ones.