Experts and Leaders Call for AI Regulation Amid Safety Concerns
If New Engines Can Kill Us All, Their Builders May Not Rule Themselves
When those who wield a novel power confess it might end humanity, that power can no longer rest on corporate discretion and political delay, but must be bound by known law and public consent—or be treated as usurpation.
I read with some astonishment that men and women employed to fashion new thinking engines now say they are “genuinely frightened” for humanity’s future. One researcher, having resigned his post, speaks of a coming “point of no return, where humanity is just not in control of the future anymore.” Another, still in that trade, assigns more than a ten in a hundred chance that such systems “could kill all humans.” Others dispute the evidence, calling these estimates “evidence-free,” yet the fearful themselves plainly believe what they say. When builders talk this way about their own handiwork, the question of political power is already on the table, for they confess that their craft may determine who rules, and whether anyone survives to be ruled.
I am told also that leaders of large firms in this line of business now call for “industry-wide deceleration and regulation,” and even for “independent monitoring” of these models, while a former magistrate of great eminence warns in private that this technology may be “dangerous” if not properly managed and urges his fellows to put the matter high on their list. Here then we have three things at once: private insiders raising alarms, rival companies professing to welcome restraint, and public officers dimly perceiving a danger to those they govern. In my Second Treatise of Government, I defined political power (I paraphrase myself) as the right to make laws with penalties only for the public good. Whoever touches a power that might destroy us all already claims such a right in effect, whether he admits it or not.
We ought therefore to begin, not with their promises, but with first principles. By nature, every person is free and equal, subject to no other man’s will, but to the law of nature only. That law, which is reason, teaches that no one ought to harm another in his life, health, liberty, or possessions. It was for the better securing of these that men ever left the state of nature and consented to civil government. Any scheme of rule which puts those very goods to greater hazard than they ran in the state of nature cannot stand as an improvement; it forfeits our rational assent. If there is indeed a fair chance—call it ten in a hundred—that this new power “could kill all humans,” we are not speaking of a small inconvenience, but of the annihilation of lives, liberties, and estates together.
Yet the evidence is disputed. One scientist says the dire estimate is not grounded; others reply that they “earnestly believe” it. I do not pretend to judge their calculations; I was long ago a physician, and I know that when learned men disagree, the first task is to clear our notions. So I mark only what is common ground: that capable observers report “extremely bad, malicious, misaligned behavior” already in current models; that those whose fortunes depend on this business nonetheless call for slowing down and for rules; and that no one seriously denies these systems may, if misdirected, do very great harm. When men who stand to gain most still ask to be bound, reason cries out to know by whom, and by what law.
Here lies the danger that smells most rank of arbitrary power. If the same corporations that design these systems also set the rules, inspect themselves, and decide when to proceed, they are judges in their own case. If rulers, dazzled by promises of advantage or fearing to fall behind other states, allow development to run ahead of clear, consented law, they exercise prerogative in the darkest sense: power against standing rule, on a mere profession of public good, and without a visible way for the people to revoke it. In the Second Treatise I allowed some prerogative only where it manifestly secures the people and is always answerable to them. When a power may render “humanity…not in control of the future anymore,” such discretion, if unchecked, is no longer prerogative but usurpation.
Some will object that sharp restraints might hinder progress, wealth, or national defense, and that confusion over “doomsday” warnings counsels patience. But fear of disorder has ever been the plea of those who would prolong a long train of abuses. The question is not whether this art can yield benefits; gunpowder and the printing press did the same. It is whether the terms on which it is pursued leave each person under known standing laws, with an impartial judge, or instead expose all mankind to experiments run by a few, upon whose private understanding and conscience the fate of the species turns. If the latter, even vast promised gains are but gilding upon chains, bright metal upon bondage.
I do not say that every venture in this trade is already tyranny, nor that every company must be dissolved. The dissolution of a particular form of governance over this technology need not dissolve society, nor even commerce in general. But where corporate or executive power presses forward with systems capable of “malicious, misaligned behavior,” while candid insiders confess a serious risk that they “could kill all humans,” no free people are bound to sit quiet, hoping for the best. They are entitled to demand clear laws made in their name, independent oversight not captured by those it oversees, and a pace of change governed by the common safety, not by the race of rivals. Where that trust is plainly broken and no lawful remedy remains, they may at last appeal, not to the discretion of their masters, but to heaven.
From the works cited
- Second Treatise of Government — Political power defined narrowly as the right to make laws, with penalties, only for the public good — anything beyond that definition must justify itself or stand condemned.
- Second Treatise of Government — The law of nature as analytic baseline: perfect freedom and equality under the law of nature, so that government must be shown to improve on that baseline for everyone who leaves it.
- Second Treatise of Government — Prerogative as revocable emergency discretion — legitimate exactly as long as the public visibly benefits and no longer.