Oh, look! There's a bird outside!

– Five-year-old Kai, whenever his teacher tried to get him to read the word 'cat'.

I didn't blame Kai. He was what you call 'reading-resistant'. He was five and had better things to think about than whatever I was spelling out with the big foam squares on the kindergarten floor. Just because his cousin Soulamath was already reading up a storm? Let her. He'd take his time. But he needed to get the ESL teacher off his case. He'd use anything available as a distraction.

I was used to this. After all, I had half-a-dozen Latin-resistant baseball players in my morning class over at the college¹. Their idea of a distraction ran to asking me to suggest more of those cool books. I'd endeared myself to them by recommending a scifi novel called Legions of Hell. Some of the characters were ancient Romans who deserved to be in hell – fighting, among others, the Vietcong. My students liked the book because they learned a lot of shocking Roman swearwords. This was a ninja move on my part: it distracted them from their distraction. I could have given them Catullus, but scifi is more fun than Latin poetry.

I am reminded of this by the behaviour of Google lately.

Google, this morning: would you like AI to help you write your emails?

Use "help me write" in Gmail and Docs to quickly draft and refine content, or use Gemini Advanced to research and brainstorm using our most capable AI models.

Let's see: do I want your grubby little paws in my writing style? Apage, Satanas. Go peddle your wares elsewhere. Those wares have been weighed in the balance and found wanting. Coming on top of this week's AI news, the timing of this 'offer' could not have been less appealing.

No, a person can only stay in the air for about one second after a single leap. However, some say that if you run off a cliff, you can stay in the air as long as you keep running and don't look down.

These and many other astounding statements were made by Google's AI Overview. It told people to eat rocks, preferably geodes. It suggested using superglue in pizza sauce to prevent the cheese from sliding off the pizza. It listed the benefits of nuclear war.

Twitter users were quick to diagnose the two main causes of Google's AI malaise:

1. AI's inability to recognise sarcasm and satire.
2. Reddit users.

Which leads to one immediate piece of advice for Google programmers – or anybody, really, with a baby AI: don't train Junior on the contents of the internet. Especially not snark-filled fora. Monkey see, monkey do. It's not safe out there. We haven't infant-proofed the place.

I do have some good news for that software engineer who was convinced Google's AI was sentient. Unless and until that thing learns to detect sarcasm as well as a Pittsburgh third-grader, we humans have nothing to worry about.

Except letting them into our emails.

What does that have to do with distractions? Well, it is my personal opinion that online service producers like Google are trying to get us to use AI to distract us from the fact that not only are they failing to make their products more useful to us – like they promised to do, back when we were all young and charming and in love with the internet – they are actually making them less useful. Google searches used to be productive. Nowadays they present more and more of a challenge to the determined researcher, who has to wade through page after page after page of advertising and 'sponsored' garbage to get to any information at all. In case you're too young to remember, it worked better in 1999, when h2g2 started.

I think Google and others have decided that end users are a drug on the market. They wish we'd go away. Or at least, that we'd take what we were offered without complaining. How dare we not want to know what they want to tell us right this minute? What do we think we are, anyway, important or something?

Hence the increased use of AI-this and AI-that. AI will tell you what to want. You will buy it. See? We're all happy.

AI will tell you what you're hungry for. Where the recipe is. Where to buy the ingredients. Look: you've barely finished typing in your credit card information and lo! There's the doorbell AI, letting you know that your package of ready-prepared foodstuffs has arrived on the porch. Should we tip the drone that delivered it? Better not, it might get mugged by the neighbour's drone.

You take the box inside. Your AI-operated stove asks if you want help. Your AI appliances offer cheery advice. You slip the stuff in, it is cooked, you eat.

It would be rude to criticise the cooking. Besides, you're afraid the tech will sulk. Sulking, it can do – but it still can't detect sarcasm. . .

So you loudly proclaim, 'Yum. Yum. Yum!' It is happy.

I, personally, would like for the AI enthusiasts to slow down the alleged 'progress' here. At least until AI learns what a joke is and stops telling us to eat rocks.

The other part of the tale of Kai and Soulamath: Kai kept refusing to spell out C-A-T. Soulamath kept reading.

One day, there was an unscheduled fire drill at the elementary school. It turned out the fire alarm had come from my kindergarten class. A tearful Soulamath confessed that, in her enthusiasm for reading, she'd made out the words 'pull down' on the fire alarm – and happily pulled down.

The assistant principal said this was my fault. I said it was the fault of premature education; also putting fire alarms where five-year-olds could reach them.

I wonder if that school has an AI security bot now?

Dmitri Gheorgheni

17.06.24 Front Page

Back Issue Page