Later today, President Trump is expected to sign executive order called the “American AI Initiative.” The preview publicity has been underwhelming, at best, and otherwise rather chilling.
The supposed purpose of the document is to set out a game plan for America to compete with at least 18 other countries that have their own national AI strategies. China announced last year that it intended to lead the world in AI, and that it expected its economy to reap $150 billion therefrom within little more than a decade.
If the publicity is accurate, the “American AI Initiative” doesn’t come anywhere close.
First off, it’s an executive order, which is really more of a mission statement than a memo (sorry Jerry Maguire). News reports say it notes five areas of focus, but sets no concrete goals, no explicit criteria for prioritizing work to reach them, and no new money to fund the work.
Then there’s the wishy washy language itself, which suggests that Federal data and other resources will be “freed” for use by researchers, and that “ethical standards” will be developed to guide “reliable, robust, trustworthy, secure, portable, and interoperable AI systems.”
Oh, and agencies will be tasked to “prepare workers for changes to the job market caused by new technology,” and the US will work with other countries in a way that retains our interests (whatever that means).
Perhaps most damning, the AI train left the station long ago, and its engines are companies like Google and Amazon, not the government. It’s as if the government thinks of itself as a customer for the resulting innovations, and the President is issuing a buyer’s guide.
So the executive order might well be too little, too late, and pretty much too irrelevant.
The development of artificial intelligence raises profound, important, and truly scary questions about our future, and they go beyond issues of our collective national security to include the very definitions of individual privacy and free will.
I think our country needs something more than a mission statement…something legislative, lasting, and far-reaching that makes concrete organizational and budgetary commitments to lead development in three areas:
Research — The Advanced Research Projects Agency, or ARPA (then DARPA), was created in the late 1950’s through two acts of Congress to devise and execute R&D on tech topics crucial to the national interest. Working with corporate research labs it pioneered technologies, from infrared sensors (you TV controller uses them) to GPS and the Internet. We need something like ARPA for AI, and wrap into it a Manhattan Project-like collaboration of research minds across government and private enterprise. It should set goals and publish findings.
Ethics — AI has and will continue to blow up our understanding of work and play, and to blithely say that robots and other automated innovations are somehow no different than mechanical looms or conveyor belts and, therefore, will create different if not more jobs for people is either stupid or a purposeful lie. The government should convene a corollary entity to its ARPA AI initiative, focused on studying and talking about the profound changes the new tech will initiate. The entity should also talk openly about how reversible those impacts may or may not be.
Security — While American companies lead AI development currently, their interests are not provincial but global. One giant marketplace made efficient by AI is a neoliberal wet dream, and governments could well be seen as irritants or inhibitors to realizing that new Eden; or, just as likely, the nominally “American” AI might find that it has a lot in common with its “Chinese” counterparts (Dr. Forbin, anyone?). We need our government to get really smart really fast on issues related to security and monitoring, kind of like developing a super-GDPR for AI that asserts its authority (instead of simply trying to limit the authority of other entities). It should resolve to never again hold a public hearing at which tech execs run circles around confused legislators.
I hope I’m wrong about the details of the “American AI Initiative,” because we need a national policy that’s a lot smarter than what I fear we’re going to get.
[This essay was originally published at A Cross of Silicon]