Institutions meant to exercise prudence should be under human control

Each of us must take stock of the unfolding era of artificial intelligence and digital agents. While it is a task for only the wisest among us to foresee what the future of large language models (LLMs) truly betokens—for the market, for America, for humankind—the rest of us might do well to hedge our bets, whithersoever the bots may take us. And here I must correct myself and all who trade in the terms of artificial intelligence: The bots will take us precisely nowhere. That is, wherever we head in this artificial era will be the result of our own choices, for the fault, dear friends, will not be in our self-driving cars, but in ourselves if ever we become underlings to an artificial overintelligence.

Our undeniable responsibility for what dreams or nightmares may come provides a first principle of government over this emerging technology: Whatever powers the bots may acquire, whatever work, white collar or blue, that the LLMs and their resulting advances in robotics may bring, the one kind of work we cannot allow them ever to do is that very work of government. The American republic rests upon the rule of law, stewarded by a lawful people, represented by, well, their lawyers and lawmakers, most of all. As a hedge against any loss of human self-government and corresponding loss of control over the bots, we should begin to debate now the possibility of a full ban on artificial agents with respect to the practice of law and lawmaking. To mix metaphors, we must dig a humane moat around the ruling arts of law and lawmaking as an absolute minimum security against the loss of self-government in the face of the awesome power of artificial intelligence, be it truly on its way or merely a figment of foolhardy capital expenditure heading into the trillions.

One key advantage of this hedge against the bots, this moat around the law, is that it works in all cases. Whether p(doom) = 0 or 0.99, we should maintain the ability and power over all lawmaking and legal work such that we have the hope of righting the ship, so to speak. Another advantage comes from the upstream effects of such a ban, for if humans, and I mean humans alone, must continue to ably practice, judge, administer, and make the laws, then a number of liberal arts—i.e., the arts of liberty—will need to be maintained in our educational system and in our habits of life: reading, writing, rhetoric, law itself, ethics, and the study of justice and humanitas more generally. All these will need to be built up in the minds and hearts of citizens and leaders such that America and Americans can continue and improve as a country and a people of law.

One further and overarching advantage to such a ban would be the preservation of real prudential decision-making such that we will better avoid being deceived by the simulacra of prudence as offered by the bots. This brief note is no place to make the full argument, but I will leave it here as a mere assertion, albeit a well-considered one, backed up by the whole history of the study of mind: Consider what we truly fear in the bots. Do we not fear that they lack judgment? Don’t we fear their inability to see that they have gone too far in some task, as in the Hugging Face incursion? Don’t we worry that, like a thoughtless genie, the bots will hear some wish from us, but grant it against all good judgment in a way that harms, perhaps grievously, whole swaths of humanity? The entire concern is that agentic power to do will be unyoked from the humane power to decide well.

I assert here that the reason for this very reasonable fear is that artificial intelligence is not designed, nor can it be designed, to use the high power of prudence found exclusively in the human intellect. Again, it is a longer argument, but the power of actual, as opposed to artificial, intellect, the sort humans possess, is the sole power on earth that can rise above a process of calculation to judge an act’s goodness or badness relative to the whole of reality. LLMs consider the next word, the next probable part; they cannot, as we do, consider the whole. What else is the law but the consideration of rules and cases, universal and particular, for the sake of the whole community and each member before the law in light of the same? We call this power, this human virtue, prudence. It would be unwise to allow an invention of man that we can already see is incapable of the very thing required to govern well to govern regardless. It would be very imprudent.

Matthew Mehan
Dr. Matthew Mehan is the Associate Dean and Associate Professor of Government for the Van Andel Graduate School of Government at Hillsdale College in Washington, D.C.