BlAIting Goat

Can you tell AI from the real thing?

There is, among newsletter writers, an unspoken covenant that every sentence has been extracted by hand from the author’s weary brain like artisanal honey from a difficult hive. Readers want to believe the paragraphs arrived after hours of pacing, three coffees, and a meaningful stare out of the window. The possibility that a chatbot quietly assembled the whole thing between breakfast and the school run is considered vaguely indecent.

Imagine, then, if the Bleating Goat had been using AI all along.

Not for everything, obviously. That would be cheating. Just the introductions. And the endings. And those clever transitions where one anecdote about a broken kettle somehow becomes an observation about late capitalism. The important bits, really.

One can picture the ritual. Open laptop. Type, “Write 800 words that sound effortlessly conversational but also suggest I’ve read at least three serious books this year.” Wait seven seconds. Complain publicly about how exhausting writing is.

The great irony is that nobody would notice. Readers already assume every newsletter is secretly produced through a blend of memory, selective exaggeration and whatever happened to trend on social media that morning. AI merely industrialises the process, replacing inspiration with probability while preserving the comforting illusion that someone has suffered artistically on your behalf.

Of course, any accusation would be denied. “I use AI only as a brainstorming partner,” the statement would insist, with exactly the sort of polished cadence that makes you suspect AI wrote the denial as well.

Perhaps this is the inevitable endpoint of internet authenticity. We perform sincerity with increasing sophistication until eventually the software becomes better at sounding like us than we are. The machine learns our hesitations, our favourite rhetorical flourishes, even the carefully deployed self-deprecation that convinces readers we remain gloriously human.

And somewhere, behind the curtain, the goat continues bleating—only now with suspiciously perfect grammar.

The techlash is finally coming. The term was coined in 2013 but the predicted hostility against big tech never really arrived. People enjoyed the benefits too much.

But now, there’s definitely a change in the weather, in more ways than one. 70% of Americans now believe big tech companies hold too much power, and you have to remember that 38% of Americans still approve of Donald Trump, so over a third of them are morons. 52% of UK citizens believe AI has more risks than benefits and 58% are anxious about it.

We are all angry about all these new data centres, gulping down electricity and water to answer each of the 30,000 messages per second that we fire at ChatGPT and its equivalents. Take a look at this Reel on Facebook.

They make our electricity more expensive and because they are critical infrastructure they are ahead of us in the queue for water when it’s scarce.

But AI writes so smoothly, and when asked to write a 300 word essay on the Bleating Goat secretly turning to AI to write its pieces, it produced an essay in three seconds, about 250 times faster than the original caprine analogue model. It makes up the first section of this post, up until the word “grammar”. Did you spot it was AI? If you put those 300 words into Pangram, the AI detection tool, it says “we believe that the entire text is AI”. They are right.

But I don’t think AI and its creators realise the hostility that is building up against it. Many writers have zero tolerance of AI use for generation of any text, though most will use for research then verify the sources independently. But using AI is not “vaguely indecent”, as ChatGPT called it just now - it’s fraud. Big tech like to think of large language learning models as the “human hive-mind” - we should use it as the stock to which we add our own small splash of individual genius. Switching metaphors, like Isaac Newton we are all “standing on the shoulders of giants”, reaching just that little bit higher thanks to all those who have gone before.

Writers are entitled to feel less charitable. AI has stolen their work and is busy trying to put them out of business, forever. It can already write bog-standard news articles and pot-boiler fiction. It is improving rapidly. By combining genres, perhaps in a way that would not occur to humans, it has the potential to produce genuinely innovative fiction and music, albeit based on what we humans have done already.

Most readers probably won’t care. Half of the UK population never read by choice anyway. Writers are expending considerable effort for very limited reward, like beekeepers trying to extract artisanal honey from a difficult hive, as someone once said.

I don’t mind, but I write mainly as a hobby - try living off the pittance I get from the few paid subscriptions among all you free subscribers. For full time professional writers, with a handful of very lucky exceptions, times are hard.

AI is still lagging behind the professionals, but writers feel like the breakaway in the Tour de France, casting anxious backward glances at the approaching peloton. And that peloton are all riding electric bikes.

It’s essential that we writers aren’t tempted to sneak a motor inside our own bikes. Even if one day the machines can seemingly do it better, we will always want the real thing, because the machines will always be faking the human experience. Only we humans truly know what it is to be human, and only human writers can transmit that genuine experience to their readers. AI is just faking it. We need to make sure that readers don’t start to suspect that all writers are using AI, so we don’t lose their trust in the genuine article.

Even cheating a little bit can be fatal - for example, Keir Starmer got some free glasses and Farage got given millions by dodgy businessmen, but still most of the public will make the false equivalence and say “they are all the same”. Writers need to hold the line.

By the way I asked Pangram to check this whole piece and it said “we believe that this text is a mix of AI and human-written content. 62% is goat-written”. It was actually 66% goat, so 4% of my writing has AI-tendencies. I’ll keep an eye on myself.

Comments welcome.