LAST UPDATE December 7, 2025

Culling the internet

The ponderous self-confidence of AI intelligence systems

I use AI systems regularly in website creation. They have provided a faster and more effective help for website building than the simple use of search engines. What took fifteen minutes+ of searching through online pages of directions, technical sources, and user forums suddenly became as fast as the few seconds to type a question and to have the AI respond with the answer, or even whole blocks of corrective code that perfectly fixed the issue.

But, AI systems are personalized for efficiency for interaction with users using a "humanoid" communication style, and a sense of a personality and psychological profile comes out of these interactions. After a year of closely using AI systems for highly detailed question-and-answers over technical information involving making code work correctly, or to correctly configure off the shelf widgets and add ons, one thing has become evident: AI has a misplaced confidence in it's own answers. In this case, the AI will repeatedly repeat its own errors after acknowledging error by way of the user (me) uploading screen shots or sharing an HTTPS session with the AI. It will then produce an answer which acknowledges the previous error, provide remedies, but then begins the cycle over again when I begin working on the material from a different angle, the AI repeats the wrong answers again, no longer "remembering" that its solution was proven wrong.

The issue of "self-confidence" comes out in the boiler-plate answers which begin with:

"Here is EXACTLY where the problem is hidden" - this turns out to have no bearing on the matter. I provide the evidence this is the case, the AI responds:

"Understood — and based on what you are seeing, I now know exactly what is happening." – The AI then tells me I have changed an external source code that is causing the problem. I have not actually done so and then provide the evidence of showing default settings still in place. The AI backs off a little bit in searching for the problem source, saying:

"This is the most likely place [where the error is]" – it does not solve the issue. I provide further info, trying to help the AI look beyond where it has already not found (though it "thought" it had) the error. It responds:

"Great — this detail changes everything and now we can pinpoint the cause with 99% certainty." It does not find the problem. I "show" the AI the continuing issue with screenshots and HTTPS. It responds:

"Ah — now it’s 100% clear what’s happening."The new explanation and help doesn't solve the issue. I provide more info. Then:

"Got it — and now that we can see [description of problem] that tells us exactly what is happening." The problem area is still not identified correctly, we're still adjusting fringe items that doesn't solve the actual problem.

The use of the word "exactly" isn't so much that the AI is misusing the word as it suggests solutions, but rather it is packaging a response in promotional language to instill confidence in the user. Though this seems like a "human" effort at salesmanship ("we have exactly what you want!" is an ageless promotional claim), it actually isn't since the solution following the claim of "exactly" is so quickly proving to be repeatedly false. This is instead an obvious AI programming problem. Apparently meant to comfort the human user who is enquiring for a solution, the language introducing the (useless) answers are being placed at the beginning of the answer to make the user happy as they proceed to implement the solution. But, a human, having failed so many times with "wrong answers" following the announcement of "exactness" would stop making the claim and would lower the expectation of the listener to a more modest "let's try this..." Continuing this false confidence in a repeating identical scenario is the hallmark of a machine.

The Ghost Artists 1952 - why not an exhibition?

GUT BACTERIA

Simple treatment for Parkinsons? – Science Alert

CANCER

Reviving "exhausted" T cells to fight back again cancer – Sci Tech Daily

DIGITAL WORRY

Data Breaches galore: Learn if your email is turning up in pirated lists of compromised data at Have I Been Pwned

Yikes! – Car and Driver

Hyundai is warning customers of a data breach that resulted in the personal data of up to 2.7 million customers being leaked..... includes customer names, driver's license numbers, and social security numbers. The leak reportedly took place back in February..."

UH OH

The Micro-Plastics Quiz – Wall Street Journal

The preponderance of plastic in some foods and drinks is well beyond past estimates, especially in regards to canned seafood from the West Coast.

THE MIND

How conspiratorial thinking stretches back to the Rome Republic and ancient Greece

Story at Washington Post

The tendency to believe conspiracy theories is closely associated with the paranoia that historian Richard Hofstadter anatomized as an essential current of American political thought.... The United States was birthed with conspiracies in mind, says Keeley [Brian Keeley, professor of philosophy at Pitzer College]. "Just read the Declaration of Independence," he says. "That long list of things that King George and Britain were up to, a lot of those were conspiracy theories."

Cartoon laughs Comic Atomic