By Jordan Reimschisel & Adam Thierer

[Originally published on Medium on May 2, 2017.]

Americans have schizophrenic opinions about artificial intelligence (AI) technologies. Ask the average American what they think of AI and they will often respond with a combination of fear, loathing, and dread. Yet, the very same AI applications they claim to be so anxious about are already benefiting their lives in profound ways.

Last week, we posted complementary essays about the growing “technopanic” over artificial intelligence and the potential for that panic to undermine many important life-enriching medical innovations or healthcare-related applications. We were inspired to write those essays after reading the results of a recent poll conducted by Morning Consult, which suggested that the public was very uncomfortable with AI technologies. “A large majority of both Republicans and Democrats believe there should be national and international regulations on artificial intelligence,” the poll found, Of the 2,200 American adults surveyed, the poll revealed that “73 percent of Democrats said there should be U.S. regulations on artificial intelligence, as did 74 percent of Republicans and 65 percent of independents.”

We noted that there were reasons to question the significance of those in light of the binary way in which the questions were asked. Nonetheless, there are clearly some serious concerns among the public about AI and robotics. You see that when you read deeper into the poll results for specific questions and find respondents saying that they are “somewhat” to “very uncomfortable” about a wide range of specific AI applications.

Yet, in each case, Americans are already deriving significant benefits from each of the AI applications they claim to be so uncomfortable with.

Flying an Airplane

Consider the poll result that 70% of respondents are somewhat to very uncomfortable with AI flying a plane. That’s strange because it is already the case today that the vast majority of functions on commercial aircraft are automated. According to the Federal Aviation Administration (FAA), automation is already used about 90% of the time in flight. And the number isn’t higher because the FAA has several restrictions on when the autopilot can be engaged. Despite, or perhaps because, the vast majority of aircraft functions being automated, flying is one of the safest forms of travel, enjoyed by over two million passengers every single day.

Driving a Car

It’s also interesting that 65% of poll respondents said they were uncomfortable with AI driving a car considering how humans have been gradually welcoming more and more automation into the driving experience over the years. Cruise control and anti-lock brakes were early examples of this trend that we now take for granted in our vehicles. Today the list of tasks and features in our cars that are being automated or supplemented with AI continues to grow rapidly: rear view “backup” cameras, parking assist functions, automatic emergency braking, lane-departure warnings and lane-keeping assistance, blind spot monitoring, collision avoidance systems, pedestrian detection warnings, voice-activated safety applications, and more. Increasingly, vehicles are essentially rolling computers that do an increasing amount of vehicle operations for us.

Of course, fully autonomous “driverless” cars could be right around the proverbial corner, and rather than fear this technology, Americans ought to embrace it. Driverless cars could offer significant improvements not only in convenience but, more importantly, the potential for significant reductionsin the growing roadway accident and death toll.

Picking a Romantic Partner

Then there’s the poll result revealing that 68% of American don’t like the idea of AI algorithms helping to pick a romantic partner. Yet, that hasn’t stopped millions of Americans from signing up for services like Match.com, eHarmony, OKCupid, or the dozens of other online dating sites or mobile dating apps. “The share of 18- to 24-year-olds who use online dating has roughly tripled from 10% in 2013 to 27%” by early 2016, according to the Pew Research Center. Overall, 15% of U.S. adults reported using online dating sites or mobile dating apps to find a romantic partner, and 59% agree with the statement that “online dating is a good way to meet people.”

Performing Surgery & Making a Medical Diagnosis

Similarly, 69% of people say they are not comfortable with AI helping to perform surgery, and 65% don’t like the sound of AI making a medical diagnosis. That’s surprising considering that, as Jordan pointed out in his earlier essay, several startups and hospitals are already doing just that with promising results. In New York, Memorial Sloan Kettering is already usingIBM’s Watson to review stacks and stacks of clinical data on cancer and provide doctors with advice on treatment options for unique cases. Moorfields Eye Hospital in London partnered with Google’s Deep Mind AI platform to help physicians plan the treatment “map” for radiation therapy patients. The amazing, and perhaps counter intuitive, part of these efforts is that they have actually enhanced the human side of medicine. By allowing AI to assist with data processing and diagnoses, doctors have more time to focus on patient-oriented tasks.

Perhaps those 65% of people are concerned about the accuracy of diagnoses made by AI? Doctors at University of North Carolina School of Medicine are using IBM’s Watson to review cancer data and recommend treatment to the oncologists on staff. In the one thousand cases that the physicians tested, Watson recommended the same course of treatment as experienced oncologists 99% of the time. And earlier this year, in a study published in Nature, scientists at Stanford University demonstrated that AI was capable of identifying potentially cancerous skin lesions just as accurately as a trained dermatologist. AI has demonstrated in numerous situations that it can rival, and even surpass, human ability to correctly diagnose disease.

People should be more concerned about not letting AI help with medical diagnoses. Traditionally, diagnoses have been driven by an examination of a disease’s symptoms. But at that point, it is already too late to prevent sickness and the suffering that comes with it. AI makes it possible to detect traces of a disease before it begins to do irreparable harm to a person’s internal organs and overall well-being. Using AI to catch conditions like Alzheimer’s and cancer years in advance of actually developing symptoms could increase the probability of survival by up to 90%.

Efforts to incorporate AI into the surgical room have also already begun. Researchers from the Sheikh Zayed Institute for Pediatric Surgical Innovation demonstrated that a robotic surgeon controlled by AI could stitch a severed pig’s intestine back together just as well as, or better than, the humans who were given the same task. The study, published in a medical journal last year, notes that “Autonomous robotic surgery … promises enhanced efficacy, safety, and improved access to optimized surgical techniques.” One of the study’s authors, surgeon Peter Kim, said that being able to work with a machine during operations would improve safety and efficiency and be a “tremendous benefit.”

Running a Factory

A narrow majority of those surveyed (53%) also don’t approve of AI running a factory. Yet, the entire history of manufacturing and distribution is one of constant productivity improvements, most of which have been brought on through increased automation of various tasks. “Economists have shown time and again that automation helps overall standards of living rise, literacy rates improve, average life span lengthen and crime rates fall,” notes Kevin Maney in a recent Newsweek feature on AI and robotics. “After waves of automation — the Industrial Revolution, mechanization, computerization — we’re way better off in almost every way.”

With the rise of the Internet and the digital revolution, this process has sped up and now countless factory processes are run by algorithms and robotic processes. In a recent report, The Promise of Artificial Intelligence, Daniel Castro and Joshua New of the Center for Data Innovation identified many of the ways that AI will help improve industrial operations across the economy by monitoring critical systems and preventing breakdowns before they happen; better managing supply chains and improving delivery times; and designing smarter industrial facilities to improve productivity and efficiency. These AI-enabled improvements help both the companies and their customers.

Making Financial Investments

Finally, 60% of those surveyed say they don’t want AI making financial investments. According to a Business Insider article written earlier this year, “machine learning and intelligent systems are already a part of the industry, encompassing everything from virtual customer assistants to complex functions that can find patterns in unstructured data.”

One of the most promising applications of AI is in the detection of fraud. Last year, MasterCard announced that it will employ AI to improve real-time approvals and reduce false declines. Javelin Strategy & Review estimatedthat the value of false-declines is thirteen times greater than the amount lost in actual credit card fraud and is a major source of frustration for customers. Using AI to lower the number of false-declines would be a major benefit for many card users.

AI is also being used to help everyday investors earn more money. Automated investment service Betterment uses AI to reduce “tax drag” and boost users’ earnings an estimated 15% over 30 years. The service constantly calculates the most tax-efficient ratio of stocks and bonds, as well as the most advantageous taxable account (traditional IRA vs. Roth IRA, etc) in which to put those investments. The AI makes adjustments in real-time in order to maximize users’ returns, allowing people to retire with more money in their bank accounts.

We Love AI, Even If We Don’t Know It!

The contrast between attitudes about AI versus real-world use of it suggests that many Americans neither fully understand nor appreciate how AI technologies already improve their lives. That’s not entirely surprisingly because, as Adam explained in his earlier essay:

AI works its magic behind the scenes. AI increasingly powers all the other gadgets and services that we do see around us: smartphones, smart cars, smart health devices, and more. We tend to appreciate those devices and applications more because they are tangible manifestations of technological innovation. But without artificial intelligence, machine-learning, and “big data,” these devices and services would be largely worthless to us.

Think about the many mundane, everyday examples of how automation, AI, and machine-learning currently benefit us by saving us money or making our lives more productive or convenient: bank ATMs, voice-activated apps and devices, language translation tools, sophisticated mapping services, tailored shopping recommendations, quickly finding travel accommodations, and so on. We already take most of these services for granted.

It is also worth noting that younger generations are much more accepting of AI in each of the categories that we have discussed. Survey respondents in the 55–64 years of age category reported being about 20% more uncomfortable with AI performing each of these tasks compared to respondents in the 18–29 years of age category.

It may simply be the case that a period of acclimation will be needed before many Americans accept or appreciate AI based technologies. Of course, that sort of adaptation process is nothing new. It’s similar to what we witnessed for cameras and cars a century ago. At first we feared and loathed them, but very quickly we came to accept and then demand them. In light of that, government ought to take special care not to enact prohibitory policies that prematurely quash AI development.

Generally speaking, despite the initial resistance we often witness toward new types of technological innovation, over time we gradually assimilate new devices and services into our lives and economy. It isn’t always easy or comfortable, and sometimes social norms and various institutions are forced to go through difficult periods of adaptation, but somehow we always “muddle through” and become more resilient and prosperous in the process.

As that process continues, AI technologies will continue to take root in our economy and our lives and improve them in ways both big and small, even when we may not fully realize or appreciate it.

The National Transportation Safety Board recommended yesterday that states ban all non-emergency use of portable electronic devices while driving, except for devices that assist the driver in driving (such as GPS). The recommendation followed the NTSB’s investigation of a tragic accident in Missouri triggered by a driver who was texting.

Personally I don’t see how someone can pay attention to the road while texting. (I’m having a hard enough time paying attention to a conference presentation while I’m typing this!) But the National Transportation Safety Board’s recommendation is a classic example of regulatory overreach based on anecdote.  The NTSB wants to use one tired driver’s indefensible and extreme texting (which led to horrific results) as an excuse to ban all use of portable electronic devices while driving – including hands-free phone conversations.  Before states act on this recommendation, they should carefully examine systematic evidence – not just anecdotes — to determine whether different uses of handheld devices pose different risks. They should also consider whether bans on some uses would expose drivers to risks greater than the risk the ban would prevent.

Texting while driving is generally a bad idea, since it involves taking one’s hands off the wheel and eyes off the road. While not wearing your seatbelt in a car or a helmet on a motorcycle probably only risks your own life, there’s a good argument to be made that distracted drivers put the lives of others at risk. The WSJ reports that 17 states have banned texting while driving outright. But is such regulation really the best way to address the problem?

Technological Empowerment. The WSJ highlights innovative technological solutions that:

1. Block calls and texts while the user is driving; OR
2. Let drivers “speak” their texts using voice-to-text technology.

Those who consider even hands-free cell phone use unsafe will probably insist on the more draconian blocking solution—and want government to mandate it! Such mandates would indeed probably be more effective than relying on the police write tickets to drivers they see texting while driving (especially since such offenses, like calling while driving, usually require some other, more serious offense before an officer can pull over a driver). But do we really need the government telling us when we can use a technology that really might be essential in certain circumstances, or totally safe in others (say, when we’re behind the wheel but stopped at a long light or in a traffic jam)?

The fascinating thing is that these solutions need not be mandated by government: At least some users will actually pay for them! Why? Because, sometimes we’re better off by being able to “bind” our future selves—just as Ulysses asked his crew to tie him to his ship’s mast so he could enjoy the Siren’s enchanting song without giving in to their spell. Similarly, these texting-blocking technologies empower users in three senses:

1. Some users know they shouldn’t text while driving but—like smokers and people who casually pick their noses—just can’t stop, so they want external discipline;
2. Others just want the monthly discount on their car insurance; and
3. Parents want to make sure they can discipline their children, who have a hard time resisting the impulse to pick up the phone.

Obviously, there are limits to how far reasons 1 and 3 can go.  Reason #1 requires at least a desire to change: As the young playboy Saint Augustine prayed, “Lord, make me chaste—but not yet!” Reason #3 requires private sector paternalism in the most literal sense. But reason #2 requires nothing more than the self-interest of users and insurance companies, which can happily coincide with road safety given the right technology. I suspect this solution is much more likely to be effective than simple government mandates, much as insurance company discounts for having smoke detectors, fire extinguishers and sprinkler systems in your home or office are probably more effective at promoting such technologies than the threat of a fine from a government inspector.

Education. Here, as elsewhere, the “less restrictive” alternative to regulation isn’t technological empowerment alone, but empowerment in combination with education. As Adam said in 2007 of anti-cell phone effort: “Banning In-Car Technologies Won’t Work.”

The proper solution here is education, not regulation. During driver training education, teenagers and other new drivers need to be taught the importance of keeping both hands on the wheel and their eyes pointed straight ahead at the road. Operating a vehicle is serious business with serious repercussions if you ignore basic rules of driving safety. But that doesn’t mean you shouldn’t have access to communications / entertainment devices while in your car. We just need to teach new drivers how (and when) to use them properly. For example, set up playlists in your iPod and start them running before you pull out of the driveway. And program the preset buttons on your car stereo or satellite radio device so you can switch stations without looking. If you have to scroll to find something new to listen to, try to do it when you’re at a stop light, not when you’re driving down a busy street with a lot of pedestrians around.

If governments really want to “do something” for road safety, they should build awareness of technological empowerment solutions, especially among parents and kids. They can also fund public service announcements like this one, which is not for the faint of heart:

http://www.youtube.com/v/DGE8LzRaySk&eurl