Artificial intelligence systems are emerging at a rapid rate. Data centers are sprouting up across the nation, providing the necessary infrastructure to teach AI models to perform a growing number of tasks. The cost of such investments is enormous, both in terms of human capital and energy consumption. AI itself has become a growth industry, creating opportunity for new applications well into the future.
However, as AI systems become more proficient in performing a growing body of new tasks, companies must weigh the cost of using AI versus the cost of employing people to perform such tasks. Over time, implementing AI system costs will inevitably become lower, pushing many companies to reduce their staff as a cost cutting measure.
Yet just because tasks can be done at lower cost does not mean jobs should be eliminated. In fact, the U.S. already has gone through these types of transformation over the past 40 years as a growing number of nondurable manufactured goods are being made overseas, where labor costs are substantially lower than what can be secured domestically.
Clothing, household items and cosmetics can all be manufactured elsewhere at significantly lower cost, given that production processes of such items are labor intensive. Of course, the unintended consequence of this shift is that small cities and towns around the U.S. that housed such manufacturing plants were gutted, as the primary employer moved their operations elsewhere.
In a free market economy, there is nothing inherently wrong with such business decisions. The challenge is retooling affected communities to manage their loss of primary employer.
The parallel to manufacturing facilities moving overseas to save money with replacing jobs with AI systems to perform the same tasks to save money is eerily similar. Much like how companies seek to produce products to get to market as inexpensively as possible, deploying AI systems to deliver tasks as inexpensively as possible assumes AI systems can deliver such tasks at the same quality as people.
This is best seen with chatbots that use large language models, which are increasingly being designed to provide customer service. It has become nearly impossible to call a customer service line without first being vetted by a chatbot. The hope is the chatbot will provide the necessary information to avoid transferring the caller to a human operator, eliminating the need for customers to speak to an actual person to have their issue resolved.
Such AI systems offer some benefits, including reducing customer waiting times and the speed at which information can be provided. Chatbots can also be customized into different languages, opening large markets based on the languages that people speak. The associated reduction in staffing saves companies money, whether their call centers are located domestically or overseas.
Yet a simple one-in, one-out exchange of AI systems and people may not be possible. An ideal use of AI systems is to support jobs that allow people to work more efficiently and accurately, hence reducing the number of people required to deliver the same volume of tasks. Areas where this is occurring include healthcare services, retail customer service,= and even software development.
Common features of tasks where AI systems can provide support is that they all deliver services, not goods. Yet not all services will be impacted in the same way.
AI systems will not be proficient in providing complex custodial services, except to the extent that they can schedule such services and provide adequate inventory for custodial supplies. They will, however, be highly efficient in executing a multitude of accounting services, with well-defined tasks and a plethora of data available to teach AI systems.
The tasks AI systems will be most adept to perform are likely at entry-level positions. The looming problem is many such positions are steppingstones to higher-level positions AI systems are not suited for. Without adequate training and experience at such entry level positions, higher level personnel will be ill-equipped to appreciate the full gamut of tasks that contribute to their business’ success.
AI systems will continue to mature, and how they will weave through the economy and the job market will continue to evolve. No one knows for certain what the steady state will be. What we know is we are nowhere near that place, and change will continue to be the hallmark as AI systems find ways to influence our economy, our livelihoods and perhaps most importantly, our lives.
Sheldon H. Jacobson is a computer science professor at the University of Illinois.