With the launch of AI technologies such as ChatGPT, the question at hand is whether we should regulate AI technologies given the potential effects it has, known or unknown. The argument of the moment appears to be that we should consider installing regulatory agencies to govern AI development and its resultant technologies if we are to keep ourselves safe from the potential dangers that may come.
As contrarian as this may sound, regulatory agencies have a bad record of performance for many of the things entrusted to them. Regulatory agencies are generally slow to respond to the crises which are their responsibility to prevent, and when they do respond, they usually provide the wrong “solution” and make the situation even worse. A case in point is the 1930s when the Federal Reserve System allowed the banking system to collapse leading to the Great Depression. The Federal Reserve System had been set up precisely to prevent this kind of problem, but it ended up causing or contributing majorly to its occurrence.
A common feature of regulatory agencies is the endless bureaucratic procedures that come with their responsibility to “solve” a problem within their domain. In painful but unsurprising irony, the very regulatory agencies that are required to approve lifesaving drugs for use refuse to approve such drugs thus leaving the public without the benefit of such drugs. The reason for this behaviour is incentives and constraints: a regulatory agency fears the prospect of being painted as “killing the public” if they happen to approve a drug that inescapably has side-effects. In reality, however, there is no perfect drug and what we have to contend with is a trade-off between no drug at all or a probability of saving lives despite the side-effects. Every choice we make in life comes with such trade-offs, there are no perfect choices.
In the context of technology, a regulatory agency would therefore be inappropriate. The fast-paced nature of the field means that the slow pace of regulators would not be a favourable factor. Also, once a regulatory agency is given a mandate over a specific field, they tend to improperly increase their sphere of regulation so that more and more things, related or unrelated, come under their umbrella. The more things a regulator has under its umbrella, the more money they would be entitled to from the public coffers, and this motivates them to cast their net wider and wider. For example, the Interstate Commerce Commission was created to regulate the “monopoly” of railroads but when the buses and trucks started competing in the industry, the Interstate Commerce Commission was not disbanded, its regulatory powers were extended to the buses and trucks. Regulatory agencies tend to increase in size even if this contradicts the reason for which they were created and may be kept long after their relevance is over. And so with technology, if we put in regulatory agencies, they will behave in the same way because the incentives and constraints are the same: they will interfere with the field, they will increase their regulatory powers into things they should keep out of, they will cause more problems rather than prevent them. At any rate, regulatory agencies are run by flesh and blood human beings like us, why would we expect them to possess any special powers to prevent what no one can?
Whether we are right to fear AI technology or not, we should not rush to install regulatory agencies in this field because history has abundant examples of how regulatory agencies usually make things worse and act contrary to their mandate. In fact, they refuse to act at all for fear of failing at what they should do. The fear of what the new thing will do is not a strange thing, it is a normal part of human nature. We may find, as we usually do, that there was nothing to fear after all. And perhaps the best way to handle AI technology is to allow the market to self-regulate in the same way we let prices freely govern the market. Knowledge is widely dispersed, and each person can get the best deal they can by utilising the little information they have. When installing a regulatory agency, we wrongly expect that it would have more knowledge than all the rest of us and we look to it as our savior. Absence of regulation does not mean chaos – there are many things that work quite well without a regulatory agency’s command and control.
The argument against abandoning regulation is that regulatory agencies are created with good intentions, and yet the road to hell is paved with the very same.