Advances in computer vision, speech, analytics, and mobile robotics promise to affect any jobs related to these skills.
There's much concern about the potential impact of artificial intelligence (AI) on jobs and the economy. And while it's difficult to know for certain which types of jobs will likely be eliminated by AI, one thing is for sure: the technology will have a profound affect on a range of skills.
"It's not easy to predict far into the future, but it is easier to indicate what AI technologies have advanced at an unusually rapid pace over the past five years," said Tom Mitchell, professor of machine learning at Carnegie Mellon University School of Computer Science.
These include computer vision, which has improved in terms of error rates in recent years; speech, which has been enhanced to the point where both Microsoft and IBM have claimed that they've reached human-level speech-to-text performance on the standard "switchboard" data set; analytical skills, evidenced by the ability to defeat top humans at games like Go and poker; and mobile robotics, including the advancements in drones and self-driving cars.
Machine learning is the underlying technology that has effectively caused each of these advances, Mitchell said.
"The shift from human programming to instead training a program from data has resulted in higher-performing AI systems across many tasks," Mitchell said. "For example, even though you can easily recognize your mother in a photograph, nobody has ever succeeded in writing down an algorithm to do that -- we know more than we can articulate. But today it is rather straightforward to train a machine learning algorithm to recognize your mother by providing training examples, photos where you label your mother and others where you indicate your mother is not present."
What types of jobs will be most impacted by AI and machine learning? The potential is there to affect any jobs that require the skills that the technologies have acquired in recent years. "Any task performed online where the inputs and outputs of a human decision are captured, is a task amenable to applying machine learning," Mitchell said.
For example, routine workflows performed online, such as approving vacation requests or travel reimbursements, also generate as a byproduct the training data needed by a machine learning algorithm to learn how to perform that task.
Most jobs involve multiple tasks though, and often machines might be able to assist in only some of those tasks, Mitchell said. "Furthermore, a big unknown is whether computers will replace or instead augment workers," he says. "Will they replace radiologists, or just advise them to make them better performers?"
New business models enabled by AI are already creating new types of businesses and jobs, Mitchell said. "For example, Uber relies heavily on AI algorithms to schedule where their drivers should be placed, to minimize the wait times of passengers while keeping all the drivers busy," he said.
This efficiency, together with the advent of ubiquitous wireless communication, has enabled Uber to create a new kind of business that employs many people. "The gig economy -- part time, just in time, opt-in employment -- is a growing phenomenon enabled by [IT] and increasingly by AI," Mitchell said.