How Should the Government Respond to AI?

Published on INFORMS OR/MS Today (Joseph Byrum)

Smart companies are developing augmented intelligence to help optimize their operations.

Artificial intelligence will be crucial to future business success … and the government needs to play a critical role in that process.

Artificial intelligence is finding its way into our everyday lives. The U.S. Food and Drug Administration recently gave the green light to an AI algorithm that detects a form of eye disease — a groundbreaking development that will lead to better treatment [1]. Software that makes voice phone calls to restaurants on your behalf to make reservations is also ready to go [2]. But are we as a society ready for the changes this technology will bring?

Often when powerful technologies are not properly understood, there is a desire to pass new laws or take some action to protect the public interest. That isn’t necessarily a good thing. Ronald Reagan famously quipped, “The nine most terrifying words in the English language are ‘I’m from the government, and I’m here to help.’”

You can see what he means in technology regulation, where the feds have a track record of lagging far behind the curve. In the 1990s, myopic Justice Department officials became obsessed with protecting the Netscape browser from the unfair competition of Microsoft’s Internet Explorer. While the government’s antitrust case dragged out in the courts, Sergey Brin and Larry Page realized that it didn’t really matter what software you used to browse the Internet. The destination is everything, and Google became the go-to source of information on the Internet. They didn’t even bother coming out with their own browser until 2008.

So, anything useful the government might do in the AI space must be forward-looking, as when agencies are out to address mission critical problems. The defense industry was the first to develop working and useful A.I systems because it had the capital needed to research the fundamental questions surrounding the technology, and the goals were worthy. With an annual spend of $4.8 billion [3], government AI support is roughly on par with the search for a cancer cure [4] – except the AI figure does not count the presumably substantial efforts being made behind closed doors and off the books for top secret intelligence agency projects.

Without the answers to fundamental AI questions that the Defense Advanced Research Projects Agency (DARPA) uncovered in its pursuits, the private sector finds itself way behind the curve right now. In particular, DARPA’s work demonstrated beyond doubt the value of the augmented intelligence approach to AI. DARPA had the luxury of exploring possibilities without having to justify itself with a return on investment, which is why a government role can pay off in the long run.

DARPA took us from the first wave of AI in the 1970s, where decisions were based on rules and deductive logic, to the second wave in which abductive logic can handle conflicting and overlapping beliefs that more closely match the imperfect environments we find in the real world. In September, the agency announced it was investing $2 billion in developing “AI Next” [5] systems that adapt to new situations and the environment around them, hoping to create a third wave.

Smart companies have taken the cue from successful military applications and are developing augmented intelligence to help optimize their operations. Joint cognitive systems outperform computer- and human-only decision systems, particularly in real-world settings full of messy data, ambiguity and conflict. That’s a win for federal involvement.

The other way in which government can make a positive contribution to the long-run success of AI is to develop a high-quality math- and science-focused curriculum. With automation taking over repetitive tasks and handling mundane calculations, solid judgment bolstered by a firm grasp on the basics will prove more useful than highly specialized training. The workers of tomorrow can draw on STEM (science, technology, engineering and mathematics) fundamentals to make the most of augmented intelligence systems. Humans must have sound critical thinking skills, letting machines handle the details. Having schools that excel in these basics will be key.

A better educated public is also the only reliable insurance policy against bad regulation. The more the public understands a technological advance, the less likely it will see new developments as a threat to their way of life. A scared public invites regulation, and that’s how you get backward-looking antitrust enforcement worried about Internet Explorer. An informed public knows better.

So, if we do a bad job at educating the public about AI, lobbying groups will pressure government into coming up with protectionist measures to regulate AI out of existence. The good news is that government has so far been making the right moves. The White House earlier this year convened a working group on AI and released a plan to continue supporting its development. Congress held a series of hearings exploring what the United States needs to do to stay ahead of the curve.

Should the United States fall behind, Beijing is only too happy to move in. China is pouring billions into a plan to establish itself as the global center for AI innovation by 2030 [6], promising a Manhattan Project-level of effort. China’s leaders rightly recognize just how critical AI will be to business success in the future.

A little competition among governments is a good thing. So, too, is fostering collaborative research across industry, academia and government. These steps are critical in ushering in the third wave of AI, and with the right attitude, yes, that man from the government just might be able to help.

References

1. https://www.fda.gov/NewsEvents/Newsroom/PressAnnouncements/ucm604357.htm
2. https://www.engadget.com/2018/06/27/google-duplex-assistant-public-testing/
3. https://www.nitrd.gov/pubs/2018supplement/FY2018NITRDSupplement.pdf
4. https://www.cancer.gov/about-nci/budget
5. https://www.darpa.mil/news-events/2018-09-07
6. https://www.chinalawblog.com/2018/03/chinas-artificial-intelligence-plan-stage-1.html