The extinction risk isn't robots

This week, AI researchers from Anthropic came out with a warning. Artificial intelligence could wipe out humanity, said one of them who recently left the company. Jacob Coxon has worked at both OpenAI and Claude. He thinks both companies are being reckless in the chase for superintelligence. Evan Hubinger is another of the researchers at Anthropic. He reposted Coxon's X thread with a grim prediction:

We really do think AI could wipe out humanity! Personally, I think there's more than a ten percent chance it happens within ten years. – Evan Hubinger on X

After the story rolled out across just about every international news outlet, Anthropic sent a short statement to The Guardian:

We have always been transparent that AI will bring both enormous benefits and unprecedented risks. To address these risks, we continue to build models with some of the strongest safeguards in the industry. – Anthropic to The Guardian

We can of course speculate whether the safety measures Anthropic follows are enough when the outcome is the end of humanity. Any other technology would have been shut down if there was a chance of it ending in mass extinction. So surely "the strongest safety measures in the industry" would be to turn off the AI tap?

At the same time as the death prophecy spreads, there are others eyeing Anthropic sceptically. And they tend to fall into two camps:

One side claims Anthropic is deliberately spreading catastrophe scenarios to strengthen the perception that artificial intelligence really is a technology of godlike capabilities. That kind of narrative could be very convenient with a potential stock market listing as early as October. It's a logic I don't quite understand; that investors would happily spend all their money on technology that might take our lives.

The other side claims Anthropic is using this narrative to get greater control over regulatory change. By maximising the crisis around AI's impact, the hypothesis is that regulations and guidance will become so restrictive, so extensive and so expensive that only companies like Anthropic and OpenAI can afford to comply with them.

Here’s Why Anthropic Is Pushing States to Regulate AI Faster

The company endorsed landmark AI transparency laws in California and New York last year, but its head of US state and local policy says they may already be outdated.

WIREDMaxwell Zeff

David Sacks is President Donald Trump's tech adviser. He's one of those who thinks Anthropic is trying to catch new AI players in regulatory flypaper so it can have the market to itself.

But all of this feels like a completely different world. Here at home in Norway, we've got our own AI problems.

What on earth are the real estate agents up to?

This is a development I've been squinting at for over a year now. And all the while I've been counting myself lucky not to be house hunting. Because while AI researchers race towards The End of Humanity, estate agents have worked out that a fisheye lens and an ultra-wide angle aren't enough when it comes to prettying up a flat for sale. Now they've started AI-generating gardens that look nothing like the real ones, interiors that don't exist and streets that aren't real.

This week brought the freshest example of estate agents in AI psychosis. The housing market in Oslo isn't the hottest right now. After many good years where the homes sold themselves, agents have had to do something they aren't used to: they have to market the properties. Anthropic and OpenAI warn about the power in their AI models. And I can absolutely see how humanity is hanging by a thread – at least if the estate agency Nordvik's St. Hanshaugen office gets its way.

In the picture, giant teenagers have taken over Stovner and Tante Ulrikkes Vei in eastern Oslo. But the agency didn't stop there. The actual street has been joined by another AI-generated street!

– This is something the photographer has done. The image has been edited using artificial intelligence, says estate agent and partner Haleem Wahid at Nordvik St. Hanshaugen to Norwegian Business news site E24.

The photographer's fault or not. When an agent's fee comes to well over 10 000 punds, the least we can expect is that the agent goes through the pictures of the property they're being paid to sell. The agency surely doesn't come out of this blunder in a good way.

We can have a laugh about it, but I also think there's a serious undertone here. When an estate agency AI-generates fictional streets and bus stops into a picture, they're manipulating the basis on which people make one of their biggest decisions; where to live. The image wasn't labelled as AI-generated either. The photographer maintains that the image is purely edited – not generated.

In April, the Norwegian business daily Dagens Næringsliv wrote about agents that AI-generated images adapted to different seasons. The grass turns green, and the plants bloom on the balcony. But for me, questions arise, for example: does the AI-generated image reflect the correct sun and light conditions?

None of this is done with bad intentions. But AI generation is a shortcut our brains love, and the more time you can save, the lower the critical thinking gets. Still, that isn't the only thing being broken down: creative thinking, memory and attention span all get worse from excessive AI use, a BBC article shows.

‘Think outside the bots’: How to stop AI from turning your brain to mush

GPS ruined our sense of direction. Search engines weaken our memory. AI, scientists warn, could do the same to everything from creativity to critical thinking.

BBCThomas Germain

The article draws parallels to the Google effect. Research shows that we remember less of the information we search for online when we google, because it takes so little effort. Our brains quite simply get weaker from farming out our tasks.

We can laugh at agents who make an area look like Jonathan Swift's Lilliput. We can tear our hair out when powerful, family-owned businesses has no qualms about AI-generating poor designs for their products:

Maybe the Anthropic researchers are right: humanity is threatened by artificial intelligence. But it won't be armed robots, or AI-generated viruses, that kill us. It could very well be our own cognitive decline. Because we live in a world where we no longer read the property prospectus – AI can do that for us. A world where we can't be bothered to click the link Google gives us – we can just read the AI Overview.

So we end up the same way, all of us: unimaginative and incapable of concentrating. That means we've lost two of the things that got humanity to where we are today.

Thus far, and no further.