The humans have all gone nuts about artificial intelligence

s of today there are 23,899 articles on so-called artificial intelligence so far this year in the scientific literature. I have never seen so much enthusiasm for something that does not exist. Most of them advocate “AI governance.”

Here‘s a sentence from a typical one.

With rigorous governance, artificial intelligence can improve efficacy, safety, and efficiency while advancing mechanistic understanding of hyperbaric oxygen therapy. ^[1] [emphasis added]

Another one declares AI to be a public health problem, citing a laundry list of risks:

. . . algorithmic bias, erroneous clinical diagnoses and recommendations, AI-enabled health disinformation, harms to mental health, AI-driven mass unemployment, lethal autonomous weapons systems, AI-enabled chemical and biological weapons and AI as both a vulnerability and an enabler of cyberattacks on health systems. Each risk is already causing harm, or could plausibly do so soon, to the health of substantial proportions of populations, thereby collectively satisfying the criteria for a public health problem.^[2]

The authors demand “public health-oriented surveillance of AI-related harms and systematic embedding of public health expertise within AI governance structures.” Yet most of the harms are either hypothetical or are based on fake events staged by AI companies.

One rare one that doesn‘t mention “governance” used AI to study how cats get run over by cars.^[3] The authors say they found 111 cats that were killed by cars and 160 that got killed by something else and that AI could help distinguish them. It seems like this would be pretty easy, but for some reason they wanted the AI to use verbal descriptions alone.

But I spoke too soon about regulation:

The implementation of AI-assisted systems in this domain must be guided by rigorous ethical and legal considerations, including the establishment of appropriate rules and regulations to govern their development and real-world application.^[3]

What is happening? Have the humans all gone mad???

Chatbots are merely algorithms

These days you can’t throw a dead halibut six feet without hitting somebody who thinks “AI” is a threat comparable to nuclear weapons. But there‘s a big difference: nuclear weapons exist, AI does not. It is a new algorithm, or technically a collection of them: pattern recognition, speech and image generation, and LLM. They are potentially useful provided you can somehow keep them from confabulating, but unless you’re willing to call a macrophage intelligent, they are not intelligent in any real sense and regulating them would be as pointless as regulating nonlinear curve fitting. Antitrust action, maybe, if the companies are in collusion.

AI algorithms can create grammatically correct summaries of information gathered from a database. If the documents contain mostly true statements, then the summary could be mostly true. It will also be bland because any eloquence, insight, or imagination is stripped out. Regardless of what it might say, it has no concept of truth or falsity. The chatbot expels intelligence from the documents as an old-fashioned wringer dryer excludes water from clothes.

What these chatbots are really doing is testing the theory of wisdom of the crowds: the idea that a consensus of a group is more likely to be accurate than input from an individual. This was the principle behind Wikipedia, but it failed because a group of N editors is N times as likely to contain one fanatic who wants to use the platform for slander and falsehoods.

I spent much of my career designing algorithms and writing software to analyze data. I used them in my research to get results that would have otherwise been impossible to get. My software, mostly pattern recognition algorithms, saved several colleagues who had been falsely accused of scientific malpractice. Conversely, the failure of software to work can mean the loss of an important discovery. Algorithms are vitally important to everyone. AI algorithms might be useful too, but their current design is as far from being dangerous as it is from being intelligent.

Humans and orkels

There are many theories about why people think AI is a threat. Some say AI alarmists are trying to gain political power or government protection from competitors. That might be true, but a bigger problem is that humans crave a source of authoritative truth—an infallible oracle.

Risk from AI is a classic fake problem. Humans invent fake problems because they feel threat­ened by something but wish to conceal their true objective. A fake problem is one that can be easily solved: just ban the offending chemical, introduce “reasonable restrictions” on develop­ment, or discredit somebody. This is much easier than trying to solve real ones because the characteristic of a real problem is that somebody else thinks it’s not a problem and does not want it solved.

‘Governance’ and ‘reasonable restrictions’ are very vague. What exactly is there to regulate about algorithms? Should they be prohibited from becoming intelligent? Should we make it illegal not to contain a “kill switch”? Make it illegal to test another computer’s security or to propose mutations in a DNA sequence? These are impractical and even nonsensical demands and people are getting suspicious about a possible ulterior motive.

The premise of science is that truth is whatever studying the world tells us. The premise of religion is that truth is whatever their deity tells us. The premise of chatbots is that truth is whatever the ‘wisdom’ of the crowds tells us. History predicts that if something is based on a false premise, as chatbots are, the only ones who should be scared are the investors.

sep 24 2026, 4:38 am

No AIs were harmed in writing this article, but we tried.