I’m pretty sure that the following line - ‘It’s funny that words don’t have worse consequences than they do’ - was written by a famous copywriter, whose book I know I have someplace, and whose name now escapes me. 

He was right. Sometimes those consequences can be terrible (“oven ready deals” anyone?), sometimes amusing (“we’ll be ready by Tuesday. At the latest.” Yeah. Ok) and sometimes unintentionally, disarmingly honest. Honest enough to make me guffaw at the unintended irony.

Take this little extract, from a press report a few days back, where a government spokesperson is reported to have said: 

“We’re considering a range of factors to address inactivity, and further details on this will be set out in due course.”

Compressed into one short unintentionally illuminating sentence; the government’s answer to dealing with inactivity is a bit more of the usual inactivity, and double-sneakingly illuminating as the ‘inactivity’ here is clearly a euphemism aimed at the ‘lazy workshy b*stards’ demographic.

If you’re failing to sway opinion with honestly, best dissemble seems to be the lesson.

Now, of course, the quote could have been truncated by the subs looking for a little light relief. Fair play if so. Made me smile. If not, you have to wonder whether pratfall or a rare blooming of honestly, or perhaps a bit of both, give a bit of time to celebrate the art of doublespeak is still alive and thriving (and probably for a good deal longer than when Orwell coined it in 1949 in 1984.) 

So intentional doublespeak uniquely a human trait? I’m wondering as the art of the euphemism, that lovely thing at doublespeak’s heart, feels like something only actual people can do. Any yet, as Stephen Wolfram sets out in his amazing dissection (‘What Is ChatGPT Doing … and Why Does It Work?’) of how these Large Language Models work, they’re

“adding a “token”, which could be just a part of a word, which is why it can sometimes “make up new words”.”

Which, I guess, means it could potentially learn to not only coin the next new post-core-core trend hanger (perhaps unintentionally. Which, frankly, is exactly how I happen upon interesting collisions) but also intentional doublespeak.

Which leaves the question; what if the accidental parses generated by our prompts were actually the Model’s attempt at irony? Or doublespeak? As The Doublespeak Awards (created by our NUT - that’s the National Union for Teachers in the U of K) in 1974) observed in 2017, that for some - well, Kellyanne Conway anyway, there are always “alternative facts.” 

Maybe the Model already knows.

Postscript:
Interestingly, first time I happened on a commentator (Mark Scott, POLITICO’s chief technology correspondent) describing ChatSHT (apologies, my naming convention) as bringing “possibilities of.. technologies into people’s imagination in the same way the iPhone did in 2007.” Which is a very interesting take. Wish I thought that, it’s sounds like the kind of thing will say about the 2022’s AI to whatever we’re all obsessed with 16 years hence.