Artificial Intelligence is a new kind of technological beast

Artificial intelligence (AI) systems will steal your job, and that might be okay. But before that happens, AI developers need to stop developing and figure out how we will live in the world of supercompetent AI.

Written by Jillian Dawley, Biochemistry, LMU '18

An artificial intelligence (AI) system recently developed by Google has created its own “child” and trained it to perform tasks at a higher level than any similar human-built AI system. With figures such as Elon Musk, Bill Gates, and Stephen Hawking openly warning about the dangers of highly competent AI, this achievement makes some of those dangers seem more real. Fears about a robot apocalypse aside, one of the main concerns surrounding AI is the widespread job losses it will cause.

To be sure, the fear about automation taking jobs is hardly new. At the dawn of the first Industrial Revolution in the mid-eighteenth century, workers and contemporary observers were concerned about the way the new machines reduced the labor necessary to do the same amount of work in the textile industry. Prior to the Industrial Revolution, textile production was done by hand and carried out in people’s homes. The process often involved the whole family.

With advent of the Spinning Jenny in 1764, spinning moved out of the home and into mills. At first, the increased output from faster spinning also increased the demand for handloom weaving. This demand for handweavers was eventually checked by the invention of the Power Loom in 1785. By the early 1800s, power looms were fully implemented in textile production and both spinning and weaving were carried out in factories.

Not only did this new mode of production greatly reduce the number of workers necessary to produce the same output, but it also put skilled weavers out of work. Not surprisingly, workers and others observers on their side opposed the transition to factory production on these grounds. On the other hand, employers and supporters of technology viewed the same changes as progress.

Steam-related technological advances brought about the Second Industrial Revolution, which peaked between 1870 and 1914. Massive changes in transportation and communication–for example, the railroad, telegraph, and various electric technologies–brought about the formation of a modern industrial economy. Coupled with the introduction of scientific management to factory labor, this next wave of technological advances revived worries about automation-related job loss.

Fast forwarding to the 1960s when computers began to appear in offices and robots in factories, President John F. Kennedy identified that the main challenge of the decade would be to “maintain full employment at a time when automation...is replacing men.” In all of these periods, however, the reduction of skill requirements and job loss were countered by the creation of jobs in other fields and education that addressed new skills. Some argue that the same shifts and recuperations will occur when, not if, AI takes over the grand majority of existing jobs.

This is a mistake. AI technology is different from advances of the past. Robots will be better at everything than humans are capable of. With highly competent AI, the machines will develop and maintain the machines. As is already possible, AI will improve AI. This will de-skill all human skill.

Even if AI wipes out the need for human labor and humans are left free to live a life of play, we will still be living in a world essentially run by highly competent AI systems. It is doubtful that humans will be “the boss,” so we will have to devise ways to ensure AI serves our goals. One option, from AI-critical Musk himself, is that we will connect the machines to our brains so that they can’t have interests that diverge from ours. Perhaps a more comforting alternative is for developers to program AI systems to operate under set human values.

As simple as option two sounds, there are so many questions to be answered about what those values are and how to account for differences in individual as well as cultural values. Identifying and agreeing upon those values will take a lot of conversation and a lot of time. Figuring out how to ensure that AI adheres to those values will also take some time. Some experts say we have this time because the development of superintelligent AI is not within the foreseeable future.

Even then, the foreseeable future is not necessarily far away. In a speech delivered at the dedication of a new section of railroad in 1847, Senator Daniel Webster remarked that “The progress of the age has almost outstripped human belief; the future is known only to Omniscience.” We are in a similar situation today, except that the technology we are on the verge of creating may be the new Omniscient. Once God AI is here, our previous methods of adapting to technological revolutions will be inapplicable. We need to take action now to slow AI development and instead have widespread, coordinated discussions about how to develop AI the right way–before we develop it.