Ask Mary "Missy" Cummings if becoming a pilot is a good career choice, and she's likely to say no.
"I have to tell you -- in all honesty and in good faith -- I really cannot recommend that someone choose that as their lifelong career path," Cummings, director of the Humans and Autonomy Lab at Duke University, said at the recent MIT Sloan CIO Symposium.
According to Cummings, people who become commercial pilots don't get to actually do much flying -- thanks to automation. Moreover, the limited amount of time a pilot is in control of the aircraft is likely only a temporary reality. Commercial aircraft might not be programmed for automatic takeoff right now, but "the technology exists today to go gate-to-gate," Cummings said.
As machines get smarter, they're encroaching on job functions in unexpected ways. (The journalism profession is a prime example.) And smart machines are starting to raise eyebrows -- and questions -- about their potential economic impact. SearchCIO caught up with Cummings at the MIT event and asked her to weigh in on the debate.
One criticism of smart machines and artificial intelligence is the potential loss of jobs. How do you respond to a criticism like that?
Mary "Missy" Cummings: I think it's naive to say there will be no loss of jobs. Some people will come out and say, 'We're just shifting jobs. We're creating new jobs.' And that's actually true. For example, in the drone world: We are creating entirely new infrastructures of jobs in terms of real estate, now there's new photography outfits, air quality monitoring so there's new data analytics that go along with the new data that's coming out of these systems.
But in the end, are we replacing a pilot? In some cases, we are. FedEx, UPS, DHL, one day, will be drones. Yes, we will have replaced those people. So there are some people who get replaced in robotic applications. But if you look at the growth of industry around it and the entrepreneurism, the start-ups -- people get creative for what to do to take the place of that job. I'm a fan of that.
And I'm also a fan of automating out humans where they potentially are unsafe. For example, we need to get rid of train engineers. We need to automate rail. We've seen the recent Amtrak crash -- this is an accident that keeps happening over and over again because of human frailties.
These people should lose their jobs because we can replace them and we can make the job safer. Those people can then move to a supervisory center. We still need supervisors of trains, we still need planners, we still need logisticians. So, I think that there are other jobs those same people could do, but we need to take them out of the path where they cause the most damage.
I've heard the argument that rather than automated systems, having the computer and the human work together is a better model. But there's also the worry that as entry-level positions get eaten away by machines, businesses will lose the ability to build expertise.
Cummings: We've been having this debate ever since the invention of calculators. If we let students use calculators, do they really suffer because they didn't have to do long division by hand? And certainly in the aviation world, we worry about pilots who have never actually flown a real aircraft. Do they really understand what's going on?
But we need to disambiguate what is the core knowledge needed to understand that system as opposed to a rote, routine skill that is highly rehearsed -- where you don't need knowledge, per se, you just need a lot of experience to, for example, keep your car in between the two white lines of the road. Does that really take a lot of your cognitive brain power? Once you see how to do it, you can do it, but if you get tired and fatigued, then you don't do it as well as when you're paying attention.
And so, we need to be very careful that when we automate certain jobs, we understand that we are automating the basic skill levels, but we need humans as coaches, guides, filling in the gaps of uncertainty that computers simply cannot do right now.
But we're already hearing about industries jumping into automation without doing much thinking about the consequences. Do you think there's going be a backlash -- that we have to learn from errors before we right the balance?
Cummings: I certainly think the stock market issues in 2008 showed people that -- because they were relying on algorithms to make decisions that turned out to be disastrous. Competitive environments, like the financial industry, are particularly prone to this -- that people might jump very quickly and make a wrong decision because they're hoping that the advantage they get from the speed of automation, for example, puts them ahead of the competitors.
But again, this is why you need to go back to the design process and make sure you're identifying these issues early on and CIOs are asking themselves, what level of supervision do I need? What's my uncertainty? What are my chances of a problem? Do I actually know what's going on in the back end of my code?
I have some students who are working in some financial warehouses right now and I definitely do not want to invest with them because I know how they code. I know the mistakes they make. And so unless you have a really great strong testing and evaluation program -- and most companies in the United States that are looking into advanced automation don't -- then you need to be very careful.