Before I start, I want to emphasize that I am a massive, massive opponent of LLM technology. I absolutely detest it, I think it is likely to collapse and take down the economy, and I think other less likely but possible outcomes from it are even worse. In no way does my upcoming refutation in any way indicate any sort of approval of this shit technology.

We The People, have been able to get 50-60-70 (some argue as high as 100) mpg out of carb driven ICE cars for over 80 years!

But a car company (i.e. the rich folks) bought the patents and shelved them

I first heard this story almost fifty years ago: https://en.wikipedia.org/wiki/Charles_Nelson_Pogue - check out the patents yourself.

It just isn't true, of course. Engineering is hard: there is no magic carburetor or other trick. If there were, China or Japan would build it and in China's case, ignore the law.

Car efficiencies steadily improved over decades - and then dropped like a stone because of massive SUVs. Car companies make somewhat more efficient cars, encouraging people to buy more cars, net increasing the demand for gasoline.

It's what makes the world go around. It's not perfect, but it's better than alternatives.

Capitalism is literally destroying our ecosystem at an exponential rate. It's hard to imagine a worse outcome than decimating our biosphere.

As a programmer you should know all this about LLM's and shouldn't be calling them AI. [More lecturing and hectoring on this idea.]

You: "Everyone else in the world uses this word wrong! Only I use it right!"

AI does not mean artificial general intelligence or superhuman competence. We called silly llittle programs like ELIZA "AI".

And LLMs long ago passed the Turing Test. They perform the imitation of fairly complicated and complex reasoning.

For example, I can give any of the major LLMs instructions on how to build a slightly complicated program, in English, and it will spit out a program that will work, a lot of the time, and a rational explanation of how it works.

This is certainly the impression of intelligent behavior! Until 2003, had you shown me this, I'd have been 100% convinced that that answer would have had to have been written by a human.

So to say, "Oh, there's nothing there," is not logically sound. It is a fairly convincing simulation of intelligence - the average person is wowed almost 100% of the time. Programs like ELIZA are called AI

It is much like the joke about a dog that's playing poker and winning. Someone says to the owner, "Wow, that's one smart dog!" and the owners says, "Are you kidding? He's drawn twice on an inside straight just today!"

They're a close system incapable of thinking or creating novel new ideas or approaches to things. It's literally impossible!

Two completely unsupported claims.

They're an algorithm that has more information to base it's answers on than we as humans are capable of remembering - that's it.

Finally some sort of argument, but unfortunately, not a valid one. Why is it that a complicated enough algorithm and a lot of information couldn't perform some given behavior that you think of as intelligent? Maybe it couldn't - maybe it could.

Helpful with being able to do trivial things? Yes. Can they do some boilerplate for you? Of course. Once a project gets any kind of complexity? They hallucinate and become worthless

Many people claim to be able to dramatically improve their output using LLMs. My experiments have not convinced me that it's so, personally. While it's quite astonishing how quickly one can write a fairly good program/function/module, it has a lot of bad habits, it creates subtle alien bugs, and as you point out, the LLM itself does not learn from experience, though contexts are effecting in providing a local, temporary illusion of learning.

Some people seem to make very fast progress on generating test features using fleets of these LLMs, some more or less to correct the flaws of others.

It's my belief that this will quickly become impossible - the technical debt will overpower them. I am however not certain of this belief.

mostly because they cannot conceptualize, imagine or think.

Three unsupported claims! I tend to believe these claims, but you need to make some sort of argument for them.

Your post comes across as dramatic, for the sake of drama.

Which part, specifically?

I don't see this as either respectful, or advancing your argument.

Relax. Take a breath.

This is condescending, and does not advance your argument.

It sounds like mainstream media has gotten to you, don't let that happen.

This is condescending, and does not advance your argument. (Also the phrase "mainstream media" has often been used with me by people with beliefs that were contrary to the fact, like vaccine and climate deniers, free energy, and, well, magic carburetors.)

Up until mid-December, I had a functioning career, and then I was summoned into a video conference, asked to describe my AI competences, and when they were limited, shown the door. Now every it seems single job matching my fairly wide array of skills wants AI, though I keep looking.

I believe I have been involuntarily retired. So why shouldn't I be pissed off? What are your feelings about having an income, yourself?

AI has over ten times as much money invested in it as went into the dot com boom.

If it collapses, on top of the energy crisis and the tariff crisis, it will tank the world economy for a long time, fscking a lot of people.

If it is moderately successful, it's going to destroy a lot of jobs, putting huge downward pressure on all other salaries, and also, an evaporation of value from all the massively overvalued AI companies.

And if it's totally successful, the majority of us will never work again, and have to depend on the charity of people like Bezos, Altman and Zuckerberg to eat. I'm waiting for my check already!