The only thing we have to fear is government AI intervention itself

Which is more dangerous - artificial intelligence, or government intervention of artificial intelligence?

It is a question worth asking. After all, politicians are now arguing that artificial intelligence is too important, powerful and potentially dangerous to develop without federal supervision.

You may have seen the proposal by Rep. Ro Khanna of California. He recently called for a new federal agency to regulate AI. He compared his proposal to the Food and Drug Administration, suggesting government experts should evaluate advanced AI systems and determine whether they are safe.

Before giving politicians that authority, we should consider how innovation has actually occurred throughout history.

Did government control produce the automobile—or did competition?

When automobiles began replacing horses and carriages, they brought legitimate risks. Cars were fast, dangerous and disruptive. Roads were not designed for them. Traffic laws barely existed. Accidents occurred. Entire industries built around stables, carriages and blacksmiths faced an uncertain future.

Imagine if the federal government created a car approval committee with the power to decide which vehicles could be designed, manufactured and sold. Imagine if the government had a "kill switch" to stop the regulation of any car that could displace a horse and buggy.

The incentives would have been perverse. Instead of satisfying drivers, carmakers would have spent more time satisfying regulators.

Instead, inventors were generally free to experiment. Companies competed to produce automobiles that were safer, faster, more dependable and more affordable. What began as an expensive curiosity became a source of mobility and freedom for millions of Americans.

True, government eventually established rules for conduct on public roads and held reckless drivers responsible. It did not assume control over the automobile itself.

And that distinction matters. Rules of the road protect people from harmful conduct. Government control determines whether innovation may occur in the first place.

Did government control transform the landline into the smartphone—or did competition?

Today, the smartphone in your pocket contains vastly more computing power than the computers used during the early space missions. It is simultaneously a phone, camera, map, library, newspaper, television, music player, bank, office and gateway to nearly all the world’s information.

Now imagine if Washington had created a Government Smartphone Agency just as Steve Jobs and Apple were developing the iPhone.

Before adding a camera, Apple might have needed permission from photography regulators. Before introducing maps, it might have needed approval from transportation officials. Banking applications could have required another review. Location services, internet access, video calls and the App Store might each have triggered additional hearings, safety studies and compliance requirements.

Regulators might have told Apple that a telephone was for making calls—not taking photographs, providing directions, streaming movies or replacing a computer.

Existing telephone, camera, mapping and media companies would have descended on Washington to warn that Apple’s ideas were dangerous, disruptive and unfair. Each would have had an incentive to use regulation to protect its own market.

Steve Jobs did not need a panel of federal officials to imagine what a phone could become. Apple introduced a new product, competitors responded and consumers decided whether it was valuable.

A government panel probably would have produced a safer, slightly improved landline. After all, millions of jobs depended on phonebooks, answering machines and landlines.

In life, most experiments fail. Some succeed. The successful ideas attract customers, investment and competitors. Competition then produces better services at lower prices.

That is how progress works. It is decentralized, unpredictable and frequently messy.

Government control works in the opposite direction. It concentrates decisions in the hands of a relatively small group of officials. When those officials make a mistake, that mistake can obstruct an entire industry.

Why should artificial intelligence be any different?

There are legitimate concerns about AI. It can be used to commit fraud, impersonate people, attack computer networks and facilitate other crimes. Those actions should be punished.

Fraud committed with artificial intelligence is still fraud. Identity theft committed with artificial intelligence is still identity theft. Threatening someone through an AI-generated message is still a threat.

Government should protect individual rights and hold people responsible when they cause actual harm. But punishing harmful conduct is different from controlling the underlying technology.

Khanna’s comparison to the FDA reveals the danger. The FDA does not merely punish companies when their products harm people. It operates an approval system that determines whether products may enter the market in the first place.

Apply that model to artificial intelligence and federal officials could decide which AI products may be developed, what capabilities they may possess and whether Americans may use them.

Who gets to decide whether something is safe?

Who gets to evaluate the political values the technology represents?

Who is to say a future administration won't pressure AI companies to suppress what it deems to be "misinformation."

A federal AI agency would also favor today’s largest technology companies.

Google, Microsoft, Meta and OpenAI can afford armies of attorneys, lobbyists and compliance officers. A small developer with a better idea cannot.

Once government controls entry into the AI market, established companies will help write the rules and learn how to navigate them. Those rules can then become an expensive barrier against new competitors.

Washington would call that consumer protection. In practice, it could become protection for today’s most powerful corporations.

The people who built the first automobiles could not foresee autonomous vehicles. The creators of the landline could not foresee smartphones. The pioneers of the early internet could not foresee the enormous digital economy that would emerge from it.

Today’s regulators cannot foresee what artificial intelligence will become either. That is precisely why its future should not be placed in their hands.

Let's remember that innovations don't flourish simply because government chooses the correct future.

They flourish because free people are allowed to discover it.