Article: Keeping AI from Becoming the Modern Frankenstein's Monster

Photo Source: Flickr

The Paperclip Optimizer continues a long tradition of cautionary tales of the unintended consequences of people giving life to inanimate objects. Throughout history, people have shared stories about animated creations gone wrong, from murderous golems in Jewish folklore to the gothic classic by Mary Wollstonecraft Shelley Frankenstein: or The Modern Prometheus. The tragedy and horror of Frankenstein’s Monster springs from the realization by Dr. Frankenstein that he found his scientific creation completely hideous—a monster. The monster soon discovers that everyone who sees him finds him repulsive, and the monster sadly concludes that he will not find love or community with humans. The monster, enraged by rejection, seeks out Dr. Frankenstein for help with his suffering. The monster demands that the doctor make him a female companion or else he will kill the doctor’s family. Ultimately, the doctor does not make a companion for the monster, and the monster kills the doctor’s family and escapes the doctor’s attempts at revenge.

The Paperclip Optimizer refers to a thought experiment designed to test the ethical boundaries of what could happen if computer scientists developed artificial intelligence that rapidly became far superior to any human intelligence. The Professor of Philosophy at Oxford University, Nick Bostrom, first described the Paperclip Optimizer in a publication titled “Ethical Issues in Advanced Artificial Intelligence,” published in Cognitive, Emotive and Ethical Aspects of Decision Making in Humans and in Artificial Intelligence in 2003. (https://nickbostrom.com/ethics/ai) Bostrom suggests the possible unintended consequences of an artificial superintelligence having a seemingly banal goal of manufacturing as many paperclips as possible with no evil intention. However, the super AI could focus all its attention on the one task of manufacturing as many paperclips as possible, diverting resources away from everything else that does not help in making paperclips. With its superintelligence, the computer would take over the earth and even develop space travel to other planets to build more paperclip factories to satisfy the need to make as many paperclips as possible. Such a singular focus would wipe out humans or anything else that detracts from the master goal of paperclip production.

Bostrom uses the Paperclip Optimizer to argue that the designers of intelligent computer systems must devise ways to infuse these superintelligences with a sense of morals. He posits that a super-intelligent computer could solve problems faster than any human or group of humans, and it could build new technologies at an ever-increasing speed that would protect it from anything trying to stop it. Such a system would be all-powerful, which could threaten our very existence as people on earth.

The unintended consequences of people animating objects have held a place in human imagination and storytelling for thousands of years. The golem, often made of clay and brought to life by a rabbi, has inhabited Jewish folklore for centuries. In 1818, Mary Shelly published Frankenstein, and her story of a monster brought to life through science by Dr. Victor Frankenstein has influenced countless books, movies, and articles that caution the dangers of creating life in clay, body parts, or machines. Nick Bostrom used the Paperclip Optimizer thought experiment to warn what can happen if super artificial intelligence has a seemingly banal goal of optimizing the production of paperclips. He cautions that such a superintelligence could develop technologies to protect itself from harm and eliminate anything that does not help or hinders the production of paperclips destroying life as we know it. Such a cautionary tale suggests that the designers of advanced intelligence must infuse a sense of morals into the super AI that corresponds with human morals, such as the value of life, or else our creation could be our undoing. The next question, one which is easier asked than answered, is what kinds of morals will this machine have? Those who develop increasingly intelligent artificial intelligence will have to wrestle with what that will actually mean and what will produce a good and safe society in the future.

Dr. Smith’s career in scientific and information research spans the areas of bioinformatics, artificial intelligence, toxicology, and chemistry. He has published a number of peer-reviewed scientific papers. He has worked over the past seventeen years developing advanced analytics, machine learning, and knowledge management tools to enable research and support high-level decision making. Tim completed his Ph.D. in Toxicology at Cornell University and a Bachelor of Science in chemistry from the University of Washington.

You can buy his book on Amazon in paperback and in kindle format here.