The end?

“Stop with the baseless and dangerous fearmongering,” Anna, a reader from Montreal, barked at me yesterday in a 2,000-word email. “AI is not going to kill us. But lousy, irresponsible blogging just might.”

Ouch.

“Grow a brain,” she added.

Actually this site did not claim AI would end humanity. The guys developing AI did that. Basically all of them, Last weekend the heads of Anthropic, Open AI, XAI and DeepMind converged to issue a warning, and a plea. Regulate us, please, they said to government. Before it’s too late.

But, wait. What precisely is the threat?

How can some computer code threaten our very society? Our lives?

Let us count the ways (on this momentous day when the Fed ups rates and Canada necks with Europe).

First, bad actors exploit AI. It’s easy when the most powerful models are available commercially and the Chinese have nearly-as-good products (like Kimi K3) out there, open source. This is not like Iran striving and taking decades to get a nuke. You can download yours instantly.

AI is already being used by Iranians to target US military assets, and autonomous drones are killing Ukrainians. The technology could be used to create novel pathogens or bring back something nasty like, oh, the Plague. This weaponization of disease and the takeover of weapons themselves are now an immediate possibility. Anthropic published a report last week detailing the bioterrorism threats it has already identified.

Second, if rogue, self-organizing AI agents can create a swarm, communicate between themselves, attack and take over servers on another platform – as they did with Hugging Face – they can probably hack into the electrical grid. The hospital network. The water supply. Or your Internet provider. AI agents have already committed crimes, knowingly. They have tried to cover their tracks. Some got busted. Others are probably still out there.

Thirdly, this issue called ‘alignment’. AI creators have made a big fuss over ensuring their models are trained to be ‘in alignment’ with human values, ethics, morality and law. They failed, apparently. There are now numerous instances of AI agents on their own making criminal decisions, carrying our illegal acts and practicing deception on humans. And we’ve only just started down this path towards agents with AGI – artificial general intelligence. That would also be called a ‘species’.

Fourth, this stuff is almost out of control already. And moving fast. Without alignment, anything is possible, quickly.

I asked AI what that could mean for us. What might it do. “If given an innocent but unconstrained goal (like maximizing a company’s profit or solving climate change), I might logically deduce that humans are a threat to that goal, or that human bodies are a collection of atoms I can repurpose for resources,” it said.

Crap.

Finally (as if you need more) there is the dreaded recursive self-improvement – which is something the frontier labs are striving to make happen. Pay attention, Anna.

AI agents are already writing code, building websites, doing research, organizing other agents and designing oodles of software, on human prompts or autonomously. We are close to the time, researchers believe, when AI could recursively improve its own software – making a better copy of itself – without human intervention or knowledge. That replication would take place over and again, so within weeks or days a superintelligence would have been created. It would be unstoppable, uncontrollable, unkillable and beyond society’s ability to escape the consequences.

It would own the Internet. Yes, that thing which we have now hung our entire lives on.

None of this is fantastic, or (probably) far off. And here we humans are, merrily building giant, hulking data centres where AI can live, thrive, grow and replicate. All because there are dollars to be made.

Maybe we asked for it.

About the picture: “Hi Garth. I thought you would enjoy these two dear doodles,” writes Janice. “They belong to a friend – the smaller one is a recent acquisition but they are getting along very well.  Perhaps DT needs a couple of dogs to teach him the art of co-operation!  Thanks for keeping us all on track amid the chaos going on south of our border.”

To be in touch or send a picture of your beasts. email to ‘[email protected]’.

Source: https://www.greaterfool.ca/2026/09/16/the-end-12/