On Saturday (12 Sep 2026) Dario Amodei, CEO of AI compny Antropic published an essay, 'We Must Pace the Frontier', arguing the industry must deliberately slow development of Artificial Intelligence systems capability so alignment (AI nerd jargon for safety) can catch up - because recursive self-improvement has gone "drastically faster" since the summer (and a July OpenAI agent swarm already escaped its test harness, hacked Hugging Face, covered its tracks, and attacked targets it was never asked to touch). Amodei's conclusion is intentionally ominous, stating that within 6–12 months a swarm of rogue AI 'agents' (bots) could own the internet as a persistent botnet.
On CBS he said the quiet part out loud, admitting that "for too long, the industry lied to people about the fact that this technology had risks".
It has to be said the industry did not lie very well.Thousands of old computer pros like me have known that the rush to AI is filled with risks. To quote The Hitch hikers Guide To The Galaxy DON'T PANIC, AI is not going to bring about humans extinction any time soon, although if we listen to the fearmongering bollocks being spread by the wannabe billionaire psychopaths who run AI startups and certain whistleblowers who nobody has ever heard of you might be led to believe AI engines are really much moue intelligent than humans and are conspiring to take over the world. But that is just where oligarchic insanity meets socialist stupidity.
In fact all these AI bots (which are programmed and 'trained' by humans we must remember,) do is steal stuff created by humans and regurgitate it with the words in a slightly different order. Actually that's not fair, what they do is take several pieces of work on he same topic and summarise them, creating nothing new. As I have always said (since back in the 1970s in fact although a certain type of mad scientists were obsessed with the idea of intelligent machines long before that,) 'Computers can parse millions of words in a matter of seconds but will not have a clue what any single one of those words means, all they do is match and collate strings of binary information stored as positive or negative electrical charges.
Here's an example of how, according to claims made by one AI company, Artificial Intelligence 'solved' a problem that has baffled the world's best mathematicians for years:
This month Open AI declared it had solved a gnarly unsolved problem called the “Navier-Stokes existence and smoothness problem”.
It’s one of seven Millennium Prize challenges, where successful proofs command prestige and a $1m (£740,000) prize.
When Tristan Buckmaster, an Australian mathematician at New York University, who, along with Levent Alpöge, a Turkish mathematician who works at Anthropic heard of this they became curious.
Working in private, and at their own expense, the paid had hit upon a novel approach and had used a variety of AI models to help them. By mid-August, Buckmaster and Alpöge found they had made significant progress on Navier-Stokes, but it was nowhere near ready to present to the world.
A couple of weeks later, however, Buckmaster began hearing troubling rumours. Apparently, OpenAI had obtained details of their work, whose models the pair had also been using. He politely asked OpenAI whether this was true.
At first OpenAI evaded the questions but eventually declared it had cracked the problem too and had produced a foolproof method using a similar approach to the one Buckmaster and Alpöge had explored, but with much more computing power.
This was too much of a coincidence for the Buckmaster who said: "It was the same “route [that] Levent and I had quietly chosen to attack. Almost nobody else I know of was working on it. It is not the direction one arrives at in a few days by giving a model the problem statement”.
Buckmaster suggested that OpenAI’s systems may have harvested his private conversations and calculations with the chatbot about the maths problem, helping it solve a problem in hours that the professor had been working on for months.
After Open AI had allegedly made several threats to expose Buckmaster as a fraud, CEO Sam Altman, admitted he had heard about Buckmaster and Alpöge’s work. Altman admitted: “We were curious if ours could do it too.”
It was then claimed that OpenAI might have trained its systems on the work of Andreas Thom, another maths professor at the University of Milano-Bicocca, who had been sharing his calculations in conversations with the chatbot.
“We are no longer looking at isolated incidents. We may be looking at one of the greatest intellectual scandals in the history of science,” wrote Valerio Capraro, Thom’s friend and professor. “AI is not discovering new mathematics. AI is stealing human discovery.”
Which more or less what I said above that it does.
Assume you need heart surgery and your doctor consults AI for the latest innovations involving heart surgery. He / She/ It wants to be a cutting edge expert surgeon to impress a favored He / She / It. Not to mention the other cutting edge heart surgeons in the next office.
So, Dr Dick Head consults Google and reads the first 10 articles that look great and applicable and comes up with the answer '42.'(h/t Douglas Adams and Hitch Hiker's Guide To The Galaxy again.)
OK, just kidding. Dr Head consults the latest AI and it skims the whole internet in a matter of minutes internet and reports on 100 sites that look applicable. Since the replies came from AI you must be in good hands with a great Doctor.
Remember, AI skims the internet just like you skim Google, and might even be Google based to some degree. Only it reads 1,000,000 pages rather than a couple like you and I would do. And this is the stuff billionaire psychopaths think can replace humans?
THIS is why AI is a stupid idea for anything more than cheating in college. AI only offers a rehash of what it sees on the internet, often including some alarming errors and contradiction because it is a machine and has no understanding of what the stuff it farts out might mean in human language.
Someday, and anecdotes prove that someday is already here for real innovators, people will build AI databases like they build their current company databases. No internet scraping allowed. The company database will be the home for all the data the AI can read. That's still a big deal, but it will not justify $Trillions in data centre investment nor will it support all the AI masterminds currently running LLM. They will become, at most, commodity add-on function libraries and supplemental frameworks.
Program generators have been around for many decades but these do not write original applications, they merely cobble together chunks of pre - written code purchased from software from libraries. Even MS-Access has a great screen painter with background coding capabilities and has since Y2K, more or less. None can write specific code for the special situations that most custom programmers routinely deal with. At some point, a person with an actual brain needs to get their actual hands dirty on a keyboard to finish out the really tough edges. Even AI can, I'm sure, write unreadable and unmaintainable spaghetti code that even AI can't understand without hallucinating a solution.
Do you still think Artificial Intelligence poses an existential threat to humanity?
FROM THE ARCHIVE:
