OPINION:
When this column last analyzed the dangers of artificial intelligence, it reported on a company’s exercise in which AI left its exercise area and went beyond human control.
Apparently, there have been other such incidents.
President Trump has announced an “AI force” and wants to appoint an “AI czar,” though the power and the makeup of that force are undetermined. Mr. Trump also announced that he would resist regulating the AI industry.
Now, a few AI corporate chieftains have called for regulating their companies. Anthropic CEO Dario Amodei and OpenAI CEO Sam Altman have urged governments — note the plural — to require the companies to comply with safety regulations and slow the pace of further development.
In response, many legislators in the Senate and House have already proposed their “solutions” to the AI issue. These include the “light touch” legislation of Senate Majority Leader John Thune, South Dakota Republican, and the Secure and Accountable Military AI Act of Sen. Kirsten Gillibrand, New York Democrat. The latter mandates human oversight of AI in high-consequence military decisions, including nuclear command and control, and prohibits AI from substituting for human judgment in such decisions.
Neither approach is practical. The U.S. government, like all others, is eager to regulate another industry and would do so badly and probably incompetently. Among jurisdictions, laws and regulations would vary, and many (or most) would be ignored entirely.
The European Union has already enacted the Artificial Intelligence Act, which supposedly bans some AI systems — e.g., those that try to manipulate human behavior — and requires others to undergo mandatory “conformity” assessments.
The EU has effectively taken its nations out of the AI race.
There is no chance that China will rein in AI. Nor will Iran, North Korea, Russia or even the Houthis. The AI that these nations will develop will have every moral instinct of a feral dog. That brings us to California Gov. Gavin Newsom’s “kill switch” proposal.
A new executive order by Mr. Newsom directs a panel of experts to consider whether to require stricter oversight of companies developing AI, including the possible creation of a “kill switch” to shut down a computer system that goes rogue. However, this requirement will surely bring Mr. Newsom into conflict with his lefty supporters.
Think about the left’s approach to AI. It favors enormous restrictions on the technology regardless of the impact on national security, but what happens when AI becomes self-aware? In the lefties’ view, that could vest any AI system with human rights equivalent to those of real people.
In an Aug. 22 issue, the very liberal British magazine The Economist opined that to vest AI with human rights would be exceedingly dangerous. Suicidal empathy, anyone?
The left will argue that becoming self-aware is the same thing as becoming a person. Never mind that millions of babies are killed by abortion every year. People do not count.
If a computer were to become conscious and function like a human mind, the left will argue, it should be vested with civil rights equal to those of any other person. The left would sink Mr. Newsom’s “kill switch” idea by holding that those computers should be protected by due process in court.
Time is the controlling element in dealing with self-aware computers. Machines will allow zero time to impose human decision-making. Yet that argument, we must assume, will not be relevant in many courts.
Mr. Newsom has stumbled on the key factor in dealing with the machine era: Humans must be in control of the machines’ decisions, whether that means intervening in a nuclear launch or rewriting the Constitution in a manner acceptable only to the machines.
The race to real artificial intelligence cannot be controlled by an international treaty or individual states’ laws. Our adversaries cannot and will not be controlled by either.
We believe we are ahead of China on AI, but we may not be ahead of our other adversaries. If we try to regulate AI by either Mr. Thune’s approach or Ms. Gillibrand’s, we will fall behind our adversaries’ efforts. So what can we do?
Mr. Trump plans to raise the subject of AI in the summit Thursday with Chinese President Xi Jinping. Still, he can rely on the fact that whatever Mr. Xi agrees to will not affect China’s race to AI — or Russia’s, North Korea’s or even the Houthis’.
The president, opposed to regulating the AI industry, may try to impose his own regulatory proposals by executive order. His “AI czar” may propose regulations, but they will come too late to address other nations’ AI efforts.
There is nothing we can do to stop the AI race. We must go ahead with the development of AI to defend our nation’s security, which includes our economy. The stakes could not be higher.
• Jed Babbin is a national security and foreign affairs columnist for The Washington Times and a contributing editor for The American Spectator.
Follow the author
Please read our comment policy before commenting.