Exports cap idea raises AI concerns

Experts fear loss of U.S. tech edge

SAN FRANCISCO — A common belief among tech industry insiders is that Silicon Valley has dominated the Internet because much of the worldwide network was designed and built by Americans.

Now a growing number of those insiders are worried that proposed export restrictions could shortcircuit the pre-eminence of U.S. companies in the next big thing to hit their industry: artificial intelligence.

In November, the Commerce Department released a list of technologies, including artificial intelligence, that are under consideration for new export rules because of their importance to national security.

Technology experts worry that blocking the export of artificial intelligence to other countries, or tying it up in red tape, will help artificial intelligence industries flourish in those nations — China, in particular — and compete with U.S. companies.

“The number of cases where exports can be sufficiently controlled are very, very, very small, and the chance of making an error is quite large,” said Jack Clark, head of policy at OpenAI, an artificial intelligence lab in San Fran-

● cisco. “If this goes wrong, it could do real damage to the AI community.”

The export controls are being considered as the United States and China engage in a trade war. President Donald Trump’s administration has been critical of the way China deals with U.S. companies, often requiring the transfer of technology to Chinese partners as the cost of doing business in the country. And federal officials are making an aggressive argument that China has stolen U.S. technology through hacking and industrial espionage.

Tech companies, academics and policymakers are calling on the Commerce Department to take a light hand with artificial intelligence export rules before a Jan. 10 deadline for public comment. Their argument has three main points: Restrictions could harm companies in the United States and help international competitors. They could stifle technology improvements. And they may not make much of a difference.

In August, Congress passed the Export Controls Act of 2018, which added export restrictions to “emerging and foundational technologies.” In mid-November, the Commerce Department, tasked with overseeing the restrictions, published a list of technologies for consideration, including several categories of artificial intelligence like computer vision, speech recognition and natural language understanding.

The restrictions would affect the export of technology to certain countries. Though it does not specify which ones, the Commerce Department proposal points to countries that have faced trade and arms embargoes in the past. Those include China, Russia and Iran.

The Commerce Department declined to comment on the proposed restrictions. After taking public comments, the department will draft a formal plan. It could result in anything from new licensing rules for artificial intelligence exports to outright bans.

Enforcing export controls on artificial intelligence would present an unusual challenge for regulators.

Artificial intelligence is something policymakers call a dual-use technology. It has innocuous commercial applications, like helping someone steer a car or ask a phone a question. It also has important military uses, like helping a weaponized drone find its targets.

Defense planners believe artificial intelligence represents the next notable change in military weapons, akin to the tactical targeting of smart bombs a generation ago or nuclear weapons before that. Artificial intelligence can also aid surveillance systems and even disinformation campaigns through software that can produce fake photos and videos.

But “trying to draw a line between what is military and what is commercial is exceedingly difficult,” said R. David Edelman, a technology policy researcher at the Massachusetts Institute of Technology. “It may be impossible.”

It is difficult to put a “made in America” label on artificial intelligence. Research on the technology is often done collaboratively by scientists and engineers all over the world.

Companies rarely hold on to the details of their artificial intelligence work, as if it were a secret recipe. Instead, they share what they learn, in hopes that other researchers can build on it. One company’s “breakthrough” is often the latest iteration of what many researchers at private companies and universities have been working on.

A lot of the computer code for artificial intelligence is published on sites like Arxiv. org, a repository of academic and corporate research. So, many policy experts believe that if the United States restricts the export of artificial intelligence products and services, it will have little effect on the progress of artificial intelligence in China and other countries.

“The core of these technologies is international and freely available,” said Edelman, who was a special assistant for economic and technology policy under President Barack Obama. “No country — the U.S. or China — has a monopoly on that.”

Federal regulations exempt publicly available information from export control. That means that the government is unlikely to bar companies and universities from publishing fundamental artificial intelligence research. But it could establish controls that restrict foreign access to that information, said Greg Jaeger, a lawyer at firm Stroock & Stroock & Lavan who deals with export controls.

Rules that prevent foreign nationals from working on certain technologies in the United States could also push researchers and companies into other countries.

“It might be easier for people to just do this stuff in Europe,” said Jason Waite, a lawyer with firm Alston & Bird who specializes in international trade.