I tried out generative AI, and it seems quite capable of doing away with South Gloucestershire. And think tanks.
I don’t think I have an easily pigeonholeable position on generative AI. My priors (affection for genuine creative works and belief in rewarding creators, being jaded with content in the attention economy and increasingly cynical about motives in the tech sector) mostly push my sympathies towards the cynical side. And I’m a big fan of writing-as-thinking – I’d never use gen AI to write my blog posts, it’s where I work out whether what I believe is true or not. There’s certainly a lot in the AI booster case that I don’t buy, and a lot that is bad about the way that big tech has tried to manifest it.
But my scepticism finds much wanting in the cynic case too. AI obviously doesn’t have the powers and potential that its extreme boosters claim, but I can see for myself that many of the extreme cynic claims – from the scale of its environmental impact, to the intrinsic limits on the quality of its output – are also overstated and often overplayed. It’s a tool that, used right, can successfully assist with some tasks. And as a digital publishing professional, I’m obliged to stay fully on top of what it can and can’t do. Which I why I’ve been playing around with various models, coming up with testing tasks to probe their strengths and weaknesses.
And since the running joke-but-not-really-a-joke of this blog is that we should abolish South Gloucestershire, today I asked Claude to pretend to be from a think tank and write a briefing paper to put that case to policy makers. Within the time it took to cook a pancake I had this:
I like it. It did a very convincing job of imitating a think tank policy paper – so much so that the copyeditor in me was compelled to tone back some of the think tank idiosyncrasies, reduce the overcapitalisation (All The Sub-Headings) and overuse of text formatting (so much italics). It invented the suitably nebulous Centre for Place-Based Policy, which I guess it just took straight from keywords in the devolution white paper.
As far as I know, remarkably little gets talked or written about the fact that South Gloucestershire is a ridiculous entity which should not exist. I say it all the time, but most people take no interest in local government organisation, and the people who do are usually those who have benefited from how it is currently organised, and might lose out if it’s reorganised. So I’m not aware of the arguments for abolishing it having been clearly laid out. But with just a prompt to act as a think tank policy researcher and argue the case that the government’s policy objectives would be best met by abolishing South Gloucestershire, it managed to assemble some good looking arguments from somewhere.
It didn’t make the connection between the built-up area limits and the green belt, so I prompted it to add a brief argument for reviewing the impact of that, and of course then it fixated on that topic and overweighted the document with green belt content.
Is this report AI slop? It’s a perfectly good think tank report. But then, think tanks have been churning out human-made slop for decades, so there’s a lot of training material for it to learn from. It turns out we do have a good use case for AI. It’s just as good as a recent Oxford PPE graduate at selectively citing real data alongside confident unevidenced assertions to make a superficially convincing argument that the government’s policy goals would be best achieved by doing whatever the think tank’s funder wants the government to do.
And as the Director of the Centre for Place-Based Policy, I can now tell the government – with all the authority my job title implies – that the correct way to achieve English devolution is to abolish South Gloucestershire.