When I heard the news that the NZ Herald was using Artificial Intelligence (AI) to write their editorials I happened to be staying overnight with an elderly couple who are subscribers who get the paper edition delivered every morning. The epitome of civilisation rendered a piss-take of civilisation. I had to ask them what they thought about this grotesque revelation.
My own initial thought was that it was one of the signs of the media apocalypse, and there have been so many. It was like finding out the village’s landmark tree had not been dying naturally but had been poisoned deliberately by its owner. Grievance. Of all the myriad self-inflicted wounds the media in New Zealand had given themselves over the last few years this one struck deep into the brain stem. It struck at its very purpose of being. Destroying the core personality function, like repeated ECT. Wrecking what little remains with so little thought.
So irrelevant is the Herald, so devalued its currency, that no-one even seemed to care terribly much – it would be just a hastily scribbled post-it note stuck into a book on the inevitable decline of the newspaper industry somewhere near the end chapter. It wasn’t even a scandal per se, the publisher NZME just brushed it off with a nonchalant statement that, yeah, so what, we’ve been doing it for ages, forgettaboutit. Stuff labels their AI assisted material on the article itself and so the debate was narrowed to how fully should it be disclosed rather than what number horseman it was.
The first thing I noticed when picking up the paper was how thin it was. So awfully thin – free community paper thin. And of course it’s a tabloid which psychologically we still associate with a lesser product, now irrefutable. Then the price, the ludicrous, laughable price: $4.70. Not an error – that’s four dollars and seventy cents. Absurd. The paper, a pint of milk, a loaf of bread, a fastpost stamp, one stage on the bus – they were approximately aligned for many, many years; but the outlier now is surely the cost of a paper at far in excess of the other items in that basket of goods. Totally over-valued even if the quality and quantity was the same as it was 20 years ago (which it certainly is not). The couple assured me their subscription gave a hefty discount on the cover price, but even so – Jesus.
I flicked through it. Didn’t take long, did it. The two stand-out differences inside from ye olde worlde times of the late 20th century of my youth (45c and yellow honesty boxes the length of Queen Street) and today were the absence of classifieds (once referred to as “rivers of gold” they are now an anaemic dead sea of about one and half pages) and the concentration of ads directed at the boomer gerontocracy (tourist cruises, retirement villages etc.) to the almost total exclusion of any ads to any other demographic. The world news section was completely supplanted by our access to international channels on the internet and TV – all of the stories were out of date already. One of the couple (her) had already been talking to the other (him) about the AI news before I came into the kitchen so I knew I would get an opinion. I could hear her down the corridor sounding exasperated about it.
“Well – what’s the point!?” I think is pretty much what they said. It was sort of creepy – was another response. How will we ever really know what anything is? What else is there that’s AI – all of it? They felt cheated, and could not understand why the editorial leader of all things would be generated by a computer when that is the absolute last thing that should ever be left to a programme. If the voice of the paper is Chat GPT then why would anyone listen to it – it cannot be a genuinely held opinion in any real sense. It is only fit for a computer to read isn’t it – not a human. We want humans to write for humans. The abdication was worse than 1936. What do they pay the editorial staff at the paper for if not writing editorials? I had to agree. The editorial was premium content behind their paywall I reminded them. They were perplexed. But nothing to cancel their subscription over I might add. It might come in handy for supporting a plate of crackers in bed I suppose.
The news story I heard on the radio (we must have been listening to RNZ) was that it took an AI programme to catch out the AI writing – the filters had detected tell-tale signs of AI composition. It’s not that difficult – I asserted confidently to them – the tells are obvious if you know what to look for, at least with Chat GPT which had installed itself onto my laptop without permission at the end of last year. It produces a rather formulaic style whereas humans tend to vary what they do even within a tight register.
We read the editorial in that day’s paper together. Yes, there were signs. Maybe not all of it, but parts of it definitely. It was banal stuff about All Black selections or something, but note the tone and the form – didactic. Reads like a lecture to a 12 year old foreigner. Classic AI. Look at all the extraneous bits, essentially filler, telling us things every New Zealander would already know and not need to be told in a series of clichés. I read it to them. It was like PR guff and devoid of actual opinion. And note the style: short sentences. Painfully short. Hemingway short. Unreadably short. And note the humour, the sarcasm, the quirky turns of phrase – there were none. An eerie absence of humanity reminiscent of Ayn Rand.
I found yesterday’s Herald on the armchair and went through that too. Even worse, even more obvious. The Herald has shredded so much credibility it’s on the rims and heading for the cliff.
You don’t need a computer to tell you what is a computer, you need a human to tell what is not human – and bub, this ain’t human. A human can sense The Terminator, we can. It can morph into whatever it wants to deliver a message or just occupy space, but it is detectable if you know what to look for. Incredibly it turns out that AI is like the annoying android character “Data” from Star Trek The Next Generation who we mocked at the time for speaking without the ability or inclination to truncate words and use slang as a human would. How unrealistic we scoffed: it’s the simplest thing ever you useless bag of bolts! Like too many fingers on a hand and spelling mistakes in the AI images it transpires that Chat GPT is similarly averse to imitating the most average aspects of everyday informal speech and common colloquialisms every bit as much as Data was averse to it thirty years ago. They were bang on – at least with the first iteration.
I subscribe to Dr Bryce Edwards’ NZ Democracy Project emails – the free version that teases and cuts off Trotter’s articles right at the point you really do consider paying – and have noticed a prodigious output ever since around about the time AI came along and I’m wondering, huh, is there a connection? Seriously, is there? I’ve read enough to think he may be being assisted at least a nudge by good old Prof. Chats. A lot of his articles include esoteric information I doubt AI could ever hope to scrape, so that is kosher, but other times it does appear to be formulaic to a fault. Chat GPT will usually do a series of quotes and then a balancing scenario of contested positions near the end and go for an anodyne positive conclusion – I’ve seen this in some of Dr Edwards’ articles. Norm be it of the formula it is. I’m not sure either whether starting sentences with “therefore” and “hence” is just his eccentricity or whether that’s an AI thing; so too when “and” is used to start a sentence and it occurs repeatedly.
Is it just mechanical writing by a lecturer run off his skates trying to make deadline, or is know-it-all professor Chats doing that voodoo that he/she/they do so well? I’m not a hundred on this at all. And as long as he’s read the Chatfeed before he hits send, then – for a news summary or review that can take an angle without partisan prejudice – fine, that is probably what AI is most competent at. If I asked Chat GPT to summarise the last 24 hours of news on a subject and it prompted me for a style, to be fair, I would probably think something challenging like Goebbels in English (and read it to me like Orson Wells), but in all probability, I would just type in “like Bryce Edwards.”
The ultimate human version that comes closest to that sort of AI neutrality programming and relentless over-production would be Tim Watkin. If Chat GPT is The Terminator then Tim Watkin is Robo-Writer. Short version: AI has completely 100% made Tim Watkin redundant.
Long version: Apart from the digital exercise of fingers jabbing furiously at a keyboard writing quantities of bland copy is pointless in this era because he pretty much precisely does what AI does, but… Chat GPT can do it within a few seconds and at no cost. And without any smugness either, which is a real blessing. Since he never has an opinion and can only recite news summaries like a lunatic town crier chained up at a library internet kiosk since 2006 (when did he launch NZ Pundit?) and since AI his ritualised summation is shed of the last possible justifications for its production his neurotic Sisyphean task is over – permanently obsolete – practiced, perhaps as it always was, as an archaic artform (with PhilU – who used to post 50+ news stories to his blog every day – we are going old school). One man’s journey to bring order to the world by cataloguing it. Sure, we all put those number and letter stickers onto the cassette tapes didn’t we, but think next level, it’s in spectrum. Since it’s fanciful he could sue Chat GPT for an algorithm derived from his soggy robotocisms or could blag himself heritage funding from the Ministry as a subsidy for being irrelevant (in the way the Spinoff has harvested the bounty of Wellingtonian mediocrity), how long can the format survive as a craft, as a viable pursuit albeit a relic? All hope of salvage is desperately unrealistic for those specialised in the generic. It is a serious, existential matter for anyone on the lowest rungs of the storytelling ladder. There is nothing that can be done to compensate the loss of purpose to one’s life, to wonder no more whether your gift, your contribution to society is irreplaceable, and not just irreplaceable but to know it’s preferrable – nothing, just nothing could mitigate that I wouldn’t have thought. To continue would be to wilfully buy a Ham radio set and sit in the shed with tea in a thermos calling Derek at 4HPNQ Tamworth at 0230 NZST. It would be the height of reality TV-grade, Farrier/Braunius subject matter, family-rated mental illness to continue doing the natural work of AI but at your very best still slightly worse than AI and – even at this early stage – about 50,000 times slower than AI. Most of them are probably incapable of climbing – rat-like let’s be honest – up the media personality ladder and out of harms way. It is an irresolvable problem when AI exhibits more recognisable personality traits when it isn’t even instructed to than the human author does when they are trying. Will these doomed lemmings or the Herald be the first over that looming cliff? Are they half way down already? And it’s a dreadful thing to have to say, of course, but I think it’s right to give a little tribute here because going by the views no one else cares: RIP Tim Watkin’s career 1993-2024.
[And works a fucking treat by the way Bryce I see why you use it]
