In a free market, there are 3 forms of regulation:
- The market itself. With choice and competition, there is a regulatory effect. Products that cause harm can die out. Milton Friedman said that in a free market the failures are just as important as the successes.
- The rule of law. When someone commits fraud, there should be legal recourse. The job of a legislative body is to determine which harms warrant government intervention, and how laws should be enforced when those harms occur.
- This is what most people consider regulation, as if the first 2 don’t exist. This form of regulation predicts something might be harmful or undesirable, so we will demand government licenses, permits, inspections, and other requirements intended to prevent harm. The downside is the chilling effect it can have on progress.
When we apply this to artificial intelligence, we must start with a definition. “Artificial” means not human. “Intelligence” is harder to define. If a child wins a spelling bee, are they intelligent or just good at retrieving data from memory? As you are reading this there is probably a device on the wall keeping you comfortable. You don’t have someone running into the room every 5 minutes with a thermometer. If a thermostat isn’t Artificial Intelligence, it can only be because we think it is not intelligent enough.
Artificial Intelligence (AI) is not a universal monopolistic entity. There is a fear that AI applications will control us and take our jobs, but depending on the definition we have lived with many types of AI for a very long time without those harms. Even if we restrict ourselves to current day “intelligent” technology, there are thousands of firms offering many more thousands of AI applications. There is a market.
The rule of law as it applies to AI is less mature. It is unclear whether new harms caused by AI have existing legal recourse. If existing law does not apply, legislative bodies have work to do in identifying harms not covered by existing law and prescribing appropriate enforcement mechanisms.
Calls for expansive regulation are often driven by expectations about AI’s future impact. Yet those expectations may themselves be overstated. While AI holds great promise for making life better, there may be an AI bubble that will burst. The book “AI Snake Oil” shows how overinflated expectations bring us back to the reality of what is achievable. In the industries I work in, there are no dark factories that replace all humans with automation. As I tell students every year, if you want to make the world better have a baby.
Any time you hear the term “regulation”, understand there are 3 forms of it. Do not jump to the last one as the only solution without considering the first 2.