On lame-duck AI bill, Congress shouldn’t push too far

Artificial intelligence holds the potential to bring a commercial and economic rebirth for the United States and its allies. Yet the U.S. Congress is getting skittish. Its leaders are reportedly negotiating a lame-duck bill to regulate the AI industry.

As officials push and prod on the new technology, they should exercise caution.

Go too far, and some of the benefits of AI might simply disappear. That would be a very big loss indeed.

Current estimates are that business spending to adopt and use AI will boost global output by as much as $20 trillion through 2030. That's roughly the equivalent of adding another Germany, Japan or India to the world economy over that time.

Much like transformative technologies of the past, such as the steam engine and electrification, AI also stands to offer a real, sustainable cure for many of the world's economic ills by giving a boost to lagging productivity.

In a recent paper for the International Monetary Fund, Nobel Prize-winning economist Michael Spence wrote, "AI is our best chance at relaxing the ... constraints that have contributed to slowing growth, new inflationary pressures, rising costs of capital, fiscal distress and declining fiscal space, and challenges in meeting sustainability goals.” Why? Because “AI has the potential not only to reverse the downward productivity trend, but over time to produce a major sustained surge in productivity."

While there's understandable concerns about AI's impact on jobs like truck driving, AI will also transform many jobs and industries in welcome ways. That could make those workers both more productive and less bored.

AI-driven chatbots are already becoming familiar in customer service. But human agents could also get instant prompts for customer responses and could receive coaching tips afterward.

Research and development costs could be greatly reduced in the crucial pharmaceutical industry by compressing timelines for identifying promising drug developments. Software engineering tasks could be reduced by 20 percent to 45 percent through the use of AI.

The truth is that AI has been in active use in many businesses — even in government — for years, with barely a ripple of concern. The Social Security Administration, for example, uses AI to speed up decisions on disability benefits by searching files to identify medical evidence that satisfies SSA's rules. The Treasury Department has used AI to help prevent and recover fraudulent payments.

Beyond that, AI will have enormous implications for the future of warfare and will likely be key to maintaining — or toppling — the current balance of power.

Ultimately, what this all means is that the U.S. cannot afford to lose out to China in the technological race to master AI and the huge markets and opportunities it will open around the world.

Still, some congressional leaders have been fretting lately that the technology could prove harmful. Their concern stems from recent apprehension expressed by the Department of Justice over algorithmic AI potentially making it easier for business owners to price-gouge.

First, this summer, the Justice Department filed an antitrust lawsuit against AI rental unit pricing software. At the time, antitrust experts expressed fear that the filing would open the door to similar governmental assaults on AI algorithms in other business sectors. These experts may already be saying, “I told you so” because, on Oct. 24, the Justice Department filed a similar amicus brief opposing Las Vegas hotel owners’ use of similar algorithmic software.

Adding further fuel to their concerns over the pressure that Congress may receive to regulate algorithmic AI in the marketplace is the fact that Vice President Kamala Harris endorsed some of the efforts against the technology on the campaign trail.

Although criticizing AI has been fashionable recently, the truth is that both the private and public sectors have used algorithmic AI for the better part of two decades. U.S. decision makers should address this market carefully, in a data-driven manner, and with nuance. They should not act against this technology purely out of fear.

There can be real, significant risks in a government push for greater control over the AI market. At a time when there is a global race to lead in this new sector, the U.S. will not want to lose the economic benefits of this technology.

Even some of the so-called light-touch ways that governments across the globe have been contemplating regulating AI could prove problematic.

Take, for example, the European Union’s right to an explanation for people affected by algorithmic decisions, which allows the citizenry to demand the data behind the algorithmic decisions the algorithms made for them. This might sound reasonable on paper, but the reality is that algorithmic AI is so powerful that sometimes, it makes decisions that transcend human understanding. The AI computing system that beat one of the world’s top chess players after making moves that other renowned players believed were mistakes at the time outlines this point well.

Greater visibility into the inner workings of algorithms has also been associated with higher security risks from hackers. Concerns about competition are another factor arguing against full transparency.

For all these reasons and more, the market — and the paying customers that comprise it — should determine which AI innovations succeed and fail, not government mandates. If a piece of algorithmic AI technology consistently makes poor decisions, individuals in the marketplace will stop buying, subscribing to, or using it.

AI can well be the goose that lays the golden egg for the U.S. economy for decades to come and losing the AI race would be an economic and national security disaster of major magnitude.

Washington should tread carefully.

Oren Etzioni, Ph.D., is Professor Emeritus of Computer Science at the University of Washington, where he helped to pioneer meta-search, online machine reading, and open information extraction. He is the founder of TrueMedia.org, a nonprofit fighting political deepfakes and founding Chief Executive Officer at the Allen Institute for AI, having served as CEO from its inception in 2013 until late 2022.