Artificial Intelligence Summarized
International trade negotiations, mass migrations, and populist politicians feed fear of unemployment among lesser-skilled workers, despite the fact that technology is a greater threat to jobs. Artificial intelligence (AI) has destroyed both blue and white collar jobs, and is developing ever more powerful computers, designed to rival the decision-making capacity of the human brain. The June 25, 2016 issue of The Economist devoted a 16-page section to a review of the current status and issues regarding artificial intelligence. This blogpost draws upon that article, as well as the several previous Fifty Year Perspective blogposts that have dealt with information technology.
Current State of the Art – AI is evolving rapidly, drawing upon our understanding of how the brain works in order to replicate the function of the brain’s neural networks with software. Layers of virtual neurons process information, and networks many layers deep are capable of higher levels of abstraction, a technique referred to as deep learning. Researchers are using various techniques to extend AI, using databases of text messages or classified images, and building on previously-acquired knowledge. A long-term goal is to build “artificial general intelligence” (AGI) capable of solving a range of tasks rather than a specific problem. AGI is a decade or more away, but short term progress will appear in improved Internet search results, better predictive ability, advanced scientific and medical research, and computers able to converse orally.
Jobs Lost, Jobs Gained – Jobs deemed at risk from automation include security guards, receptionists, cashiers, telemarketers, accountants, and taxi and delivery drivers. However, past predictions that automation will make humans redundant have proven incorrect. In the past technology has created more jobs than it has destroyed. Automatic teller machines (ATMs) are an example. ATMs reduced the number of tellers per bank branch, but the reduced cost of running a branch allowed banks to open more branches, increasing the total number of employees. Similarly, use of software to search legal documents, while reducing personnel costs in the discovery phase of legal cases, increased the demand for discovery and the number of legal clerks in the United States. E-commerce, which reduced the need for brick-and-mortar stores, has increased retail spending and, with it, retail employment. The number of jobs in fields that previously did not exist, such as video game designers, are impossible to project, but will surely increase.
Educating for New Skills – There is no question that automation requires workers to quickly acquire new skills, and that governments and companies must make it easier for workers to do so. On-line courses are being offered by founders of Udacity and Coursera in response to the new skill requirements of AI. There is general consensus that AI will require changes in the way education is delivered, and AI itself has the capacity to tailor courses to individual needs. The U.S. Bureau of Economic Research adds that “character skills” such as sociability and perseverance, skills beyond the scope of machines, must be emphasized in education, as these skills correlate to employees’ ability to adapt to new situations.
Safety Net – Even if new jobs exceed jobs lost, some form of safety net has been recommended to prepare workers for new jobs. The idea of a universal basic income, discussed in previous blogposts, has been examined. But, as the Swiss learned, concern about open borders and free movement of workers may make the basic income for all unworkable, as it would attract free riders from countries that did not implement the idea.
Existential Threat – Concern that super-intelligent machines could turn on their creators has been expressed by such prominent thinkers as physicist Stephen Hawking, entrepreneur Elon Musk, and many others in AI research. Others regard such concerns as hysteria based on stories from science fiction. Some AI researchers differentiate between the intelligence that can be taught to a machine, and consciousness, or inner experience, which is arguably not replicable on a silicon chip. The concept of creativity, the ability to connect two things that are not obviously related, has been argued as beyond the capability of machines. For security against the prospect of machines developing creative capabilities, a big, red on-off switch has been suggested for any potentially errant machine.
Geopolitical Issues – While automation may threaten jobs even for highly-skilled professionals, automation also increases the capacity of those workers. The benefits of that automation may extend to people in developing countries where there are shortages of specialists such as radiologists. Conversely, automation may reduce development opportunities in developing countries by making “off-shore” production profitable again in consumers’ countries.
Ethical and Legal Issues – Expected improvements in facial-recognition systems will heighten existing privacy concerns. Google has developed a system that exceeds human capacity to identify people in photos. AI capacity to digest massive amounts of information improves both crime-fighting and spying on citizens by authoritarian governments. There are also such questions as responsibility in the case of injury in accidents involving self-driving cars. Even more difficult is resolving ethical issues that machines without a conscience cannot resolve on their own. The Economist report offers an example: “should a self-driving car risk injuring its occupants to avoid hitting a child who steps in front of it?”
As AI proceeds to impact more people, jobs, industries, economies, and countries, it is incumbent upon regulators, academics, and corporate executives, among others, to spread AI’s benefits broadly.