The other side of scale

The blue departure board at JFK airport during the November 11 ground stop showing mostly canceled flights.

It is a fallacy insufficiently acknowledged that if a bunch of things happen simultaneously they must be connected. Sometimes it’s true: if someone who is incubating measles hangs around a bunch of unvaccinated people who’ve never had it, there’s your common cause. But pit humans’ desire for simple explanations and our extraordinary pattern-matching ability against a week in which Apple, Microsoft 365, the US Air Traffic Control system, United Airlines, the National Security Agency, and JP Morgan Chase all had IT-related outages…and you get pundits arguing that there must be a common cause and a plan to do deliberate harm. The most conspiratorial Americans think the US system is being probed for weaknesses.

On a skeptics’ mailing list, poster Dave Palmer proposes an inversion of Arthur C. Clarke’s first law of science: “Any sufficiently-advanced stupidity can be indistinguishable from malice.”

The idea of a common cause is appealing, I think, because it’s unsettling to face the likelihood that along with the conveniences and functionality scale brings are complex interdependencies with far-reaching consequences. Much worse to think our – and “our” includes most of the world outside the US – infrastructure is so brittle that the ability to function in daily life is uncertain. Nonetheless, until the causes have been forensically established it is more likely that this is a coincidental cluster of unrelated technical issues. Given the complexity, penetration, and scale of modern systems, you would expect that.

The this-must-all-be-connected theory would be more plausible if there were common causes emerging. Microsoft has blamed a network capacity upgrade. The NSA said its secure communications network suffered a “temporary degradation” during “scheduled routine maintenance”. The Federal Aviation Administration is so far pinpointing a faulty router at a remote site (an explanation disputed by the router’s manufacturer, Hewlett Packard Enteprise). United Airlines blames a “technology issue” that borked the web apps pilots use for communications and preflight calculations. Apple’s issues were less consistent across a variety of its services; the company eventually blamed a software update. JP Morgan claimed a software bug. In the middle of all that, news broke that hackers had exploited a vulnerability in Optimizely A/B testing software, which websites use to identify the most effective designs, turning numerous mainstream publishers’ sites into vectors for malware.

On the other hand, as Kate Klonick suggests on X, if there is a common factor it might be upgrades incorporating AI – that is, an unwarranted belief in AI’s capabilities while updating aging systems after firing workers. At Pluralistic, Cory Doctorow more trenchantly blames technical bankruptcy, accelerated by AI into a “new normal” in which “everything is broken, everything is breached”. This seems plausible enough to go on with. Be careful out there.

***

In a different kind of outage, more than 200,000 names are signed to a petition demanding Jensen Huang’s rehiring as Nvidia’s CEO. The company’s board says: not happening. Huang’s reinstatement is being backed by some who may identify too easily with his situation. Among them: David Remnick concludes with the irony that even though Huang helped usher in a new technical age, “the alleged plot that brought him down was familiar in Biblical times”. See what happens when you ignore history?

***

“Would it be better if Wikipedia were written by AI?” The questioner and I were on our way into an event celebrating Wikipedia’s 25 years in operation, hosted at the Royal Society by the Web Science Institute at the University of Southampton. In the few seconds available I couldn’t manage more than, “No”.

Now, a few days later, it’s obvious I should have responded by rewording the question: “Does Wikipedia need to be true?” Wikipedia itself wouldn’t ask that; it aims at verifiability, not truth. Chatbots are often neither. Tell a chatbot that every fact in an entry must have a cited source, and it will obligingly predict some.

In a keynote, Jonathan Heawood considered the many forms of truth. Scientific evidence is important, but it’s only one. The truth sought in a court of law is different from that, as are ethical/moral truths, the “truth” handed down by authorities, and the underlying truth we find in the best fiction. This is one place chatbots fail: much is not attributed, and we can’t evaluate the speaker and their interests. You could say that’s true of Wikipedia, also: you don’t know who wrote the entry you’re reading or what their biases may be. But you *can* examine the talk page and learn from the discussions recorded there how the page has been built and curated..

A chatbot has no idea why a human wants to know anything, where a human writing a Wikipedia page at least has their own genuine interest in the subject. A chatbot tells you birth and death dates because that’s how the biographies it was trained on did things. A Wikipedia editor tells you those details because they share the human need to understand how people fit into our collective history. Time matters intensely to us – but means nothing to a chatbot.

The question reminded me of the early days of the Internet when a fair number of people thought we didn’t need libraries any more because everything would be online where we could just search for it. Yeah…no.

Illustrations: JFK during the November 13, 2026 ground stop (by Sashimi B. via Wikimedia.l

Wendy M. Grossman is an award-winning journalist. Her Web site has an extensive archive of her books, articles, and music, and an archive of earlier columns in this series. She is a contributing editor for the Plutopia News Network podcast. Follow on Mastodon or Bluesky.

Author: Wendy M. Grossman

Covering computers, freedom, and privacy since 1991.

Leave a Reply

Your email address will not be published. Required fields are marked *