Credit: BrianAJackson, istockphoto.com

1997 was a pivotal year for me. It was meant to be a year of travelling the world with my partner, but due to family issues we only made it as far as Murmansk in the Russian Arctic. But in our two months there, I had my road to Damascus moment in Monchegorsk, where I decided I was going to do this for the rest of my life.

When I came back to the UK in the early summer, determined to get stuck in to my new career plans, things had changed dramatically. We had a shiny new Government promising to modernise everything and, when you switched on the TV, there were adverts for websites. It seems so quaint now that someone would be spending so much money promoting a search engine in the middle of Coronation Street, but that black Lycos dog was bounding about everywhere. It was a new world. The dotcom bubble had begun.

Investors starting shovelling money into any business with .com in its name, with grateful young entrepreneurs spending this mound of cash on shiny modern offices strewn with beanbags, fancy coffee machines and table football, er, tables. This was the future.

Until they ran out of money.

Pop went the bubble.

When I look at the current hype around AI, I feel like it’s deja vu all over again. Huge amounts of cash being invested, but how are investors going to make their returns? The vast majority of AI we use is free, and often unasked for. So who is going to pay for all those data centres, their grid connection (or ‘behind the meter’ gas turbines) and their colossal energy consumption? The ‘burn rate’ of these investments is eyewatering – last year OpenAI was spending $1.60 for every $1 of income – and some paying users have found their costs running out of control.

But there are other reasons to see a sudden correction to the AI surge:

1. Copyright: there are already moves to try and protect the intellectual property that is being fed into the AI sausage machine. Whoever invents an algorithm to prove that a particular piece of AI generated work has leant heavily on uncredited pieces of copyrighted material will be quids in.
2. Risk: we have already had cases in the UK of AIs inserting invented suicidal ideations into social workers’ notes from interviews with vulnerable children. One of my last speeches as a Councillor was to warn of over-reliance on algorithms, as they simply don’t have the gumption to understand the difference between a normal conversation and one littered with red flags. “It was the algorithm, not me, Your Honour!” isn’t going to cut much ice if some poor kid ends up on the wrong end of a protection decision.
3. Liabilities: as well as those serious risks, there have been at least two cases of Big Four consultancies’ work being found to contain invented (‘hallucinated’) references. If you’ve paid through the nose for a piece of work that turns out to be an ephemeral algorithmic fairy tale, surely you’re going to call your lawyer? Hint: employ a small consultancy instead – we won’t try and scam you like this.
4. Boredom. I always get frustrated with any algorithmic music platform as wherever you start, you always seem to end up at the same song – on YouTube that inevitable destination is Dani California by the Red Hot Chilli Peppers. By contrast, I’m writing this listening to BBC 6Music and Nick Grimshaw, a human being, has just played a track by HAAi who I’ve never heard of before (yey!). How come all those cartoonish AI images on Twitter/X all look as if they were drawn by the same person? When will we get tired of AI slop – yesterday?
5. Public resistance: there are already campaigns to resist data centres around the world and Erin Brockovic is on the case. A moratorium has been placed on data centres in Dublin and I suspect more will follow.

Don’t get me wrong, AI will be a huge part of our cultural and economic future. But we’re at a vertiginously high peak on the Gartner hype cycle and, as per dotcom, there’s a deep trough gaping in front of us. After the inevitable crash, we will all put our sensible pants back on and start treating it as another tool with benefits and costs. Part of that correction will be to find much cheaper ways of running data centres either through more efficient technology, better designed algorithms or both. Another part will be AI companies charging users the true cost of AI, plus a profit margin, which will curtail use. So of all my worries about rising to the climate challenge, AI isn’t top of my list.

I am, however, really worried about the socioeconomic impact of the inevitable crash.