Artificial intelligence.. new challenges
Daron Ajamoglu * – – -. Every new technology creates challenges and requires decisions that determine who benefits and who loses, and whether the benefits justify the harm caused. The panic over the rise of artificial intelligence tends to focus too much on a point far into the future, when the world will achieve general artificial intelligence. This is the moment when machines, as AI boosters dream, reach the ability to think and perform at the same level as, or above, humans in most activities, including those involving judgment, creativity, and design. Critics of artificial intelligence have focused on the potential danger to human civilization from superintelligence if it were to wreak havoc. The warnings were issued by tech entrepreneurs Bill Gates and Elon Musk, physicist Stephen Hawking and artificial intelligence researcher Stuart Russell. We should really be afraid not of what AI might become, but of what it is now. Almost all of the progress made so far in the field of AI has nothing to do with imagined AI, and instead has focused on some limited tasks. The capabilities of artificial intelligence do not involve anything close to real thinking. However, the effects can be harmful. This limited AI is already displacing workers. Research I did with David Autor, Jonathon Hazel and Pascual Restrepo found that companies that increase their reliance on AI by 1% decrease their employment rate by about 1%. And of course limited AI is used to power the new surveillance technologies that companies use. It is also used in the American justice system to enforce bail decisions and, increasingly, to pass judgments. It also distorts public discourse on social media, and hinders the functioning of modern democracies. The effects of artificial intelligence on the labor market may be the most serious. The American economy has helped create a lot of good jobs – paying decent wages and providing job security, and career-building opportunities – for workers of all kinds of backgrounds and skills. From the end of World War II until the mid-1970s, the United States experienced not only strong employment growth, but also rapid wage growth for both highly educated and mediocre workers. This growth was halted long before the advent of artificial intelligence. From the 1980s onwards, average wages stagnated, and people with less than a college degree began to suffer a sharp decline in their real earnings. During that period, automation and the movement of businesses from one country to another led to declines. But now, artificial intelligence is accelerating this trend, approaching, or sometimes even exceeding, human productivity on some very specific tasks in offices, warehouses, and other places. And many employers, focused on cutting costs, will pounce on any opportunity to eliminate jobs using emerging technologies. Some economists believe that fears of automation and layoffs due to the use of artificial intelligence are overstated. They argue that as work becomes more AI-dependent, the resulting productivity gains will direct employment to other sectors of the economy, sometimes even into the same companies that use AI. And if AI technologies are truly amazing at the tasks they perform today, this argument might be right. Unfortunately, current AI technologies are not only far from general intelligence, they are also not good at doing the things that humans used to do – such as facial recognition, language understanding, and problem solving. This means a double whammy of employment, because AI technologies are displacing labor and not generating any of the increase in labor demand that would have occurred had the technology not delivered more significant productivity gains. Other applications of AI are likely to exacerbate the growing power of companies and capital over labour, adding to these troubling trends. AI enables better monitoring of workers – for example in warehouses, fast food restaurants, and delivery businesses. The applications of AI in government decision-making, particularly the criminal justice system, are no less worrisome. There is also the damage that AI is doing to democratic discourse and politics, not only because of misinformation in social media algorithms, but also because of the growing ability of companies and governments to monitor and manipulate the behavior of millions of people, which is fundamentally at odds with true democracy. Each new technology creates challenges, necessitating critical decisions as to who benefits and who loses, and whether the benefits justify the harm. This is doubly true for artificial intelligence. Not only because of its pervasive effects throughout society, and its impact on areas not touched by other technologies, such as human judgment. Will artificial intelligence be allowed to do the increasing work of displacing and controlling humans, or will it be directed to complement and augment human capabilities? * American economist and professor at the Massachusetts Institute of Technology. Published by special arrangement with the Washington Post and Bloomberg News Service.
Artificial intelligence.. new challenges