Remember how scary Covid was? A lot of the fear came from uncertainty. 2026 feels the same — except now the uncertainty comes with better graphics and worse petrol prices.
The first reason for the uncertainty is geo-political, aka Trump. At this rate we’ll need to rob the petrol station before we can afford to buy petrol.
The second reason is AI.
My news might not be your news
This is a sort-of disclaimer.
The news you hear/see/read depends on the source of that news. I read BBC news highlights, for a (hopefully) more balanced view. But I also get news on my social media feed. And social media algorithms target your weaknesses — including your anxiety.
So if I’m seeing more AI-apocalypse stories than you, it’s not because I’m paranoid. (Well, maybe not.) It’s because I liked an interview with Tristan Harris. Although I’m also shown hairdresser videos, which I really can’t explain. It’s not like I’m worried about hair. The algorithm clearly disagrees.
Claude’s opinion of humans
A recent story combines geo-politics and AI. American journalist Shane Harris asked chatbot Claude how it “feels” about the US military using it to select targets. You can watch an extract on Facebook.
This led to lots of discussions about Claude’s response and ethical issues. But I was interested in Claude’s unintentional insult to humans:
“When a system like Maven generates hundred of targeting recommendations and humans spend roughly the equivalent of a glance approving each one, the human is not really making a decision in any meaningful sense. They are ratifing an algorithmic output under time pressure with incomplete information and the institutional pressure to move fast. …
That’s not human judgment. That’s automation bias with a human signature attached.”
Automation bias is when we trust automated decision-making systems more than we trust our own knowledge or other evidence. It isn’t just about AI. It also happens with autopilot systems, GPS navigation, financial forecasting systems and more. This bias can lead to errors. That’s why we have procedures to check automated suggestions before we act on them.
Closer to home
Most of us aren’t approving life-or-death decisions. We’re approving invoices, code reviews, reports.
But scale this down a bit, and it might feel exactly like what is happening in your world right now.
Are we under time pressure: Yes
Do we have incomplete information: Always
Is AI generating output faster than we can check it: Yes
Are we under pressure to move fast: Yes
AI works fast, humans work tired. So is this good human oversight? Probably not.
Oversight or undersight
We know AI output needs to be checked. You’ve read the stories about freaky output. If you are using AI responsibly, you’ve checked the output and had to fix it more than once.
There’s a problem with checking someone (or something) else’s work. It’s slow. It’s boring. It’s thankless. There’s a reason no-one volunteers for manual testing.
Oversight isn’t fun, but it is a human job. It’s ours to do. And ours to do better. We can question AI. We can challenge it. And we should.
Beware the hype
Big Tech — especially those with serious money in the game — have hyped AI relentlessly. Elon Musk even claims we will have a future of amazing abundance.
Before you believe these claims, think about this. It is a quote from “Feet of Clay”, a Discworld novel by Terry Pratchett, who is one of my favourite authors.
‘Funny thing, that,’ said Nobby. ‘You never get bad fortunes in cookies, ever noticed that? They never say stuff like: “Oh dear, things’re going to be really bad.” I mean, they’re never misfortune cookies.’
Vimes lit a cigar and shook the match to put it out. ‘That, Corporal, is because of one of the fundamental driving forces of the universe.’
‘What? Like, people who read fortune cookies are the lucky ones?’ said Nobby.
‘No. Because people who sell fortune cookies want to go on selling them.’
What do you think? Are you suffering from oversight bias, or advocating for more human checks? Share your stories.