“Personality traits, such as severe narcissism or antisocial behavior, impacts a leader’s decision-making capacity and, by extension, democratic stability.”
— James and Ashley Weinberg, paraphrased (2026)
|
Would You Know Artificial Intelligence If You Saw It?
The threat of A.I. is portrayed as an independent agent working toward non-human goals. The danger, we’re told, is that such an agent will pursue its goals at the cost of morality and human values. Such agents have been with us since the dawn of civilization in the form of independent institutions. They are now enshrined as individuals in the form of corporations and authoritarian governments.
From this point of view, exploitation is predictable. Institutions further their own expansion over the benefits of the individuals with whom they’re engaged. Corporations enhance their profits at the expense of consumers, and governments expand their power at the cost of lives and health.
We are forever discussing the pros and cons of these decisions. Institutions rarely question their motives or goals: corporations profit, governments control, and institutions sustain themselves. Institutional goals are ‘programmed’ by law or charter. If there are challenges to the limits of institutional power, they don’t come from institutional insiders but from the outsiders affected by them.
In a sense, institutions act like elements in an ecology. Manufacturers profit from maximizing sales while insurance companies profit from minimizing risk. Armies profit from maximizing perceived risk, while minimizing engagement. Governments profit from maximizing wealth either by increasing production or by theft. In any of these arenas, unchecked growth will result in institutional benefit at the expense of individuals.
Tobacco and lottery companies are prime examples of exploitative corporations (Truth Initiative 2020, Sterle 2016). Their profit is directly proportional to their customer’s loss. Insurance companies are the opposite: they profit when you don’t file claims and they lose when you do (Balson 2025). Armies are potential peace-keepers in that they benefit most when maintaining preparedness without engaging in warfare. Governments profit most when they improve production, and they can do this by selling natural resources, improving productivity, or by selling their people, as North Korea does (Scott 2026, Greig 2026).
Machines Can Behave Unpredictably
People have forgotten that computers need not be digital. The first computers were mechanical, and they were not just adding machines. Alan Turing’s first computers used electric relays, gears, and solenoids. These machines were programmed using early computer languages. Computers originated as an extension of what humans did mechanically.
When I was 7, in 1963, I built a simple model of Babbage’s 1833 “analytical engine.” It consisted of a chassis that housed rods, sliders, and rubber bands. The kit is no longer available, but something more complicated is a popular 3-D printing project.
People are afraid of A.I.-directed systems because they know A.I. is powerful and they think it is unpredictable, but A.I. systems are deterministic machines. Such machines are predictable as long as they are isolated from unpredictable influences and stay out of their zones of chaotic behavior.
It is these two caveats that make A.I. systems unpredictable: they interact with unpredictable humans, and we don’t know what it takes to make them behave chaotically. The fact that no one is even talking about chaotic behavior is evidence of the ignorant level of the discussion among those who are presented to us as being the most informed. When it comes to politicians you can be sure they know nothing.
Institutions Are Forms of Artificial Intelligence
An institution is essentially a management engine that controls resources and is programmed by charter, executives, and a board of directors. It works through the mechanisms of management and workers, and its output is modulated through resource requirements, sales, marketing, and distribution. No single mind controls it and, to a large degree, it operates without the control of any mind.
A factory is a clear example of an A.I. institution, but so is a school, a construction company, a public relations firm, or a political party. In fact, so is an ecological system. Anything that processes an input to produce an output according to a program and, most importantly, responds to changes in the environment can be seen as intelligent. Its intelligence comes from its awareness of how the environment responds to the system’s actions.
It’s not how A.I. systems act that is most unnerving, it’s how they respond. We’re all fairly aware of the resources available and their up-front costs. What we don’t know is how actors will react to the changes that an A.I. system creates, and we especially don’t know how the A.I. system will evaluate these reactions.
A system’s dynamics don’t have to be chaotic for it to be unpredictable, but when dynamics do become chaotic, then they become absolutely unpredictable. The key to limiting the results of chaotic dynamics is not to limit behavior, as that only creates a temporary bottleneck, it’s to limit the range of inputs and outputs.
Limiting inputs amounts to shutting down the system; limiting output amounts to setting laws, protocols, limits, and guard rails. If you listen to the public discussion on these issues, you’ll realize that no one seems to know what they’re talking about regarding these systems’ unpredictable nature. The public certainly does not, and the politicians, being effectively backward thinking managers, will never learn. They rely on the advice of unbiased experts, but such experts are either unrecognized or don’t exist.
It would be reasonable to question your own sanity in these times.
If you’d like help…
Feedback
There are many systems, species, individuals, and programs that are not responsive. Unless these non-responsive systems are being protected, such as aphids that are protected by ants, they die out. As is typical of ecological systems, certain species are fast growing and exploitative but not responsive or defensible.
Species like weeds, brambles, and some viruses and bacteria are able to quickly invade and exploit, but typically unable to sustain and defend. These species assert themselves through rapid distribution and multiplication.
For example, Donald Trump persists not because he is adept at defending himself, but because he is a master of deception and confusion. Like a forest fire or an invasive species he knows that his victim has only limited focus and resources. And because his office has the unlimited resources of the executive branch, he can overwhelm the press, the public, and the courts with noise and prevail in enough cases to make his profit.
Exploitative species would disappear entirely were it not for their ability to encase themselves in a protective casing and go into suspended animation until positive circumstances develop. This includes grass seeds that scatter in the dry summer and sociopathic politicians. Donald Trump exemplifies this. He and his pernicious force will disappear into the subconscious of emotionally dysfunctional people, but he will return in another form in future times of threat, destruction, indifference, confusion, and ignorance.
Just as computers originated from systems of gears, rods, and rotating cylinders, so also have A.I. systems grown from people’s processing of information and the exploitation of opportunities. That computers can be realized from mechanical machines, and A.I. systems realized from computers, indicates that institutions are themselves artificially intelligent machines.
The Future of A.I. is Already Written
The A.I. race is following in the footsteps of the atomic energy race, except that the previous race was held during a world war. In that race, my mentor Eugene Wigner drafted a letter to Roosevelt in 1939 that Albert Einstein signed for the purpose of starting the development of nuclear weapons. Today, while the adversarial nature of the competition is not as clear, nations are still stumbling over themselves to find ways to weaponize A.I.
It seems that much of the public’s concern regarding the safety of A.I. stems from the inherently aggressive use of any new technology, either to fight wars, elections, or crime. In these cases, everyone is ignorant and afraid of the potential for damage. The same would have been the case for nuclear weapons, except then there was less fear because there was more ignorance. We’re ignorant today of A.I.’s potential, but we are learning at a basic level because rudimentary forms of A.I. are available to the public.
Based on the roll-out of past weaponizable technology, A.I. technology will follow the same trajectory. That’s not because A.I. is or does things similar to gunpowder, lasers, computers, atomic energy, or space travel, but because people always conceive of these things in the same parochial ways. We’ll continue to be attracted by what satisfies us, and threatened by what we don’t understand.
As a result, the future of A.I. is a marketing question. We will be sold on what A.I. can do to please us, and will A.I.’s dangers be hidden from us? Hollywood and the news media will respond to our likes and dislikes, and social media—which is itself an A.I. system—will feed us whatever gets the most response. We will have as much say in the development of A.I. as we have in the development of previous technologies. Which is to say none.
The Cost/Benefit Equation is Incomplete
A.I. systems compute an opportunity gradient. This gradient is built from a range of potential outcomes ranked by their relative benefits. This is how an A.I. system plays chess. It does not compute the single best next move, it calculates sets of moves, collects them by their similar paths, and ranks them as sets. The best set is examined further, and the best move within the set is chosen. After each move of the opponent, the A.I. recalculates its options and selects the best. If it is successful, this process leads more or less directly to the winning move.
Chess is easy for an A.I. system because its opponent’s actions are limited and it can evaluate, collate, and assign benefit probabilities to each of its options. These “paths of opportunity” become obscure for chaotic systems, and chaotic systems are quite common. When the “chess board” of opportunity becomes large, the responses become unstable, and the number of distinct outcomes grows exponentially, it becomes impossible to compute the opportunity gradient.
The discussions of limiting the power of A.I. systems do not account for this. What we hear are woefully uninformed discussions about legal, functional, or operational limits. The “solution” for systems that exceed these limits is to shut them down. This is as pointless as punishing criminals as it only incentivizes learning to hide, deceive, and avoid capture. This will eventually restrain criminals whose resources are limited, but A.I. system resources are effectively unlimited.
What a punitive approach fails to understand is that the systems we’re talking about can operate in the chaotic region. This is akin to sailing through a hurricane which we cannot do, but A.I. systems can. To enforce the kind of control that people envision is to reduce these systems to procedural programs, and thus to remove them from being A.I. systems in the first place.
What’s the Difference Between Institutions and A.I.?
Institutions act and learn slowly. What they “see” and how they “think” are predictable. They see according to their biases and act according to their customs, which is a reflection of the limited vision and understanding of those who manage them.
Artificial intelligence systems learn quickly, but how they weigh information depends on what they’re trying to achieve, and this is where they differ from human institutions. At the moment, A.I. systems don’t have fundamental aims. They are not chartered to work for established goals but, instead, are given obscure, unreasonable, and sometimes impossible tasks.
If we created, funded, and provided resources to a corporation for similar tasks, I expect we would get somewhat similar results, albeit at a slower pace and with a narrower scope. In fact, there are departments within corporations that are given the tasks of hacking, spying, and subverting competitors and adversaries.
Marketing departments are typically tasked with taking advantage and distorting consumer sentiment. This is rarely seen as foul play. Instead, we have created rules and consumer watchdog groups. And while A.I. can perform these manipulations faster and better, it’s possible that A.I. generated rules and A.I.-based watchdogs could match A.I. systems at the same level.
The greatest difference between human institutions and artificial intelligence—which we should really call artificial institutions because it’s their actions and not their intelligence that is at issue—is that we think we know the goals of the human institutions but not the goals of the artificial ones.
But this is false. We don’t really know the goals of human institutions, we only know their means and their limits. To say that we understand the Federal Reserve because its stated goal is to keep inflation below 2% tells us next to nothing.
The reason we don’t know the goals of artificial institutions is because we either have not made them clear or we don’t know how such systems will understand the goals we set for them. It is unreasonable to think that an artificial institution will intentionally damage a resource unless such damage furthers its goal. But in situations where the A.I. institution exerts some control over costs and benefits, human values may be excluded from the opportunity gradient.
Back to Mental Health
What seems to be disturbing us most is that lacking goals against which to evaluate actions, artificial institutions might effect extensive damage on a human scale. This can make sense to the A.I. institution in the same way that one sacrifices a chess piece to win the game.
We’re afraid of A.I. systems doing “crazy” things because we have not given A.I. systems a definition of what crazy means. And if human institutions provide any measure, we have allowed such institutions to be monumentally destructive. There are no good institutional role models, intelligent or otherwise.
This brings us back to society’s general inability to recognize insanity. Until we can agree on what constitutes acceptable human behavior (Borowitz 2019), and remove those politicians and executives who fail the test, how could we ever control the unacceptable behavior of artificial institutions?
References
Balson, K. (2025 Nov. 13). “Beyond the Policy: Exploring the Landscape of Trust in Insurance.” Casualty Actuarial Society. https://ar.casact.org/beyond-the-policy-exploring-the-landscape-of-trust-in-insurance
Borowitz, A. (2019 Dec 12). “Trump Named Person of the Year by Popular Sociopath Magazine.” New Yorker Magazine. https://www.newyorker.com/humor/borowitz-report/trump-named-person-of-the-year-by-popular-sociopath-magazine
Greig, J. (2026 Sep 18). “Nations Take Action on North Korean IT Workers After UN Report.” The Record. https://therecord.media/nations-take-action-on-north-korean-it-worker-schemes
Scott, E. (2026, Aug 12). “Inside North Korea’s Operation to Conquer the American Job Market.” Wall Street Journal. https://www.wsj.com/business/media/inside-north-koreas-operation-to-conquer-the-american-job-market-93729962
Sterle, F. (2016 Feb 23). “Lottery Corporations’ Ethical/Moral Corruption.” The Philosophy Forum. https://archive.thephilosophyforum.com/discussions/239-lottery-corporations-ethicalmoral-corruption.html
Truth Initiative (2020 Jun 12). “How Tobacco and Vaping Companies are Exploiting National Crises to Maintain Their Bottom Lines.” Truth Initiative. https://truthinitiative.org/research-resources/tobacco-industry-marketing/how-tobacco-and-vaping-companies-are-exploiting
Weinberg, J., and Weinberg, A. (2026 Aug 28). “The Inner Cabinet: How Mental Health Impacts Political Decision-Making.” Parliamentary Affairs, gsag038. https://doi.org/10.1093/pa/gsag038
Enter your email for a FREE 1x/month or a paid 4x/month subscription.
Click the Stream of the Subconscious button.