The machines are coming! They’re going to steal our jobs, our livelihoods, and maybe even turn us into goo-covered batteries for their twisted mechanical needs. Or, you know, just take our jobs. Whatever.
But it turns out, even that might not be entirely true, at least to the degree many people fear. In fact, artificial intelligence, the technology many people imagine will be their professional undoing, could actually help them become more efficient at their jobs. That’s according to the results of a study conducted by researchers from Carnegie Mellon University and the Massachusetts Institute of Technology.
“Our conclusion is that what’s likely to happen isn’t the wholesale replacement of jobs by computers,” Mitchell explained. “There will be a gradual evolution and redefinition of jobs so that the tasks that can be automated will be, which will free you, the person, to do other tasks more.”
The study, which was published in the journal Science, examined machine learning, a pillar of AI that allows a computer to learn how to perform specific tasks based on the kind of data it is fed. Think how you’re able to type “dog” into the Google (GOOG, GOOGL) Photos search bar and pull up all of the pictures of dogs you’ve taken.
The study’s authors, Carnegie Mellon’s Tom Mitchell and MIT’s Erik Brynjolfsson, identified 21 criteria to determine if machine learning could do 1,000 different jobs. Criteria included whether there is a large amount of data available for specific tasks, if that data is already online and whether workers have to make well-defined decisions based on well-defined inputs.
Mitchell points to the example of a dermatologist tasked with diagnosing skin cancer based on images of a patient’s skin blemishes to determine whether the marks are cancerous or not.
The large amount of data in that situation would be the vast number of photos of cancerous and noncancerous blemishes available to doctors. That data is also available online, you’ve likely searched Google for them yourself after spending too much time in the sun.
“What happened there is that the machine learning algorithm was given about 130,000 training examples with the ground truths for determining whether or not [a blemish] was cancerous and that program was able to come up with a better strategy than the dermatologists,” Mitchell said.
Essentially, the machine learning algorithm was given examples of blemishes that are cancerous and those that aren’t and the machine was eventually able to make determinations about new blemishes on its own.
Machines can do the tedious work
Machine learning can also be used to eliminate more mundane tasks from workers’ daily lives such as approving expenses or days off, among other functions. As a result, employees could focus on more important aspects of their jobs. The dermatologist, for example, could spend more time with patients discussing issues they may be having, and explaining diagnosis plans and treatment options.
To be sure, machine learning could negatively impact some jobs. Mitchell pointed to mechanical drafters, credit authorizers and brokerage clerks as professionals that could be hurt by machine learning programs. Jobs that are least likely to be impacted by machine learning software include tasks that require physical labor such as massage therapists, animal scientists, archeologists and plasterers and stucco masons.
“The bottom line is I don’t think we are going to see the wholesale elimination of jobs, but we are going to see most jobs impacted by this technology, and it will be the routine tasks,” Mitchell said. “There will be a shift to the tasks that are more uniquely human.”
The study isn’t the only sign that AI could help workers. During a recent interview with CNBC on Monday at the World Government Summit in Dubai, Google X co-founder Sebastian Thrun said that AI tools will help workers perform at superhuman levels.
Like Mitchell and Brynjolfsson, Thrun believes AI will replace repetitive tasks and help transform society by moving workers away from mundane tasks.
“Now, that means that some jobs will go away, very repetitive work, of course,” Thrun told CNBC’s Arjun Kharpal. “But it will be replaced by created work, so we have to move from a repetitive working society into a creative society where we invent new things.”
It’s important to note, though, that some jobs will, unfortunately, be replaced in their entirety by machines and AI more generally. That’s not unique to machine learning and AI, though. New technologies always run the risk of killing jobs. But as those jobs are replaced, new ones open up.
Of course, people who lose their jobs to machines will need some kind of training programs to ensure that they can adapt to the changing market. Without those, the impact on the workforce could be devastating.
For now, machines are far from being able to replace humans in the majority of jobs.
More from Dan:
Email Daniel Howley at firstname.lastname@example.org; follow him on Twitter at @DanielHowley.