Clever, error-prone, and coming for our jobs, AI will reshape politics

Not only a hell of a lot of jobs are going to go with this transition, but so too will the entry paths to far more senior gigs. So where does all this land, asks James Morrow?

It’s fair to say that not everyone likes a new idea.

When the sewing machine first came around French tailors rioted – not because they were French, but because they feared the invention would ruin their businesses.

English textile workers later followed suit, and the name of their leader (Ned Ludd) is where we get the word “luddite” to describe those who want to reject modern technology and go back to the past.

So, as we sit on the cusp of what we are told is a new AI revolution, what’s next? Will we see hordes of office workers tipping their laptops out in the streets to protest these new systems that will leave them as jobless as buggy whip makers?

Well, maybe. But not yet.

I’ve been spending the past several weeks playing with various AI systems – the technical term is Large Language Models – that promise to research any question, solve any problem, and do it with a breezy chattiness that is almost human.

AI will revolutionise many professions but it will never replace human instinct. Picture: Manaure QUINTERO / AFP)

One area that we are told will be quickly revolutionised is law, particularly the grunt work of drafting documents and submissions.

For fun, and to see if it could do the job of a junior lawyer, I asked Grok (the Elon Musk AI engine attached to X) to draft a deed of release indemnifying my cocker spaniel from damages resultant from any, um, accidents on the floor.

Elon Musk’s Grok AI program is chatty but error prone. (Photo by Jaap Arriens/NurPhoto via Getty Images)

The result was letter perfect and looked like something that could have come from any decent law firm: It was full of clauses for release and indemnity and admissions of liability.

I showed it to a senior partner at one of the big Sydney firms and she couldn’t find fault.

But then things got trickier, and weirder.

I wanted to see how it would go as a research assistant and asked it about examples of Labor politicians talking about patriotism.

It came up with a couple of good examples ranging from Bill Shorten to Kim Beazley.

But then it started hallucinating.

At one point it just straight up made up a quote from Tanya Plibersek that suggested she was anti-patriotism – nonsense, particularly for those familiar with her 2020 speech backing in “patriotism at its practical best, patriotism as the thread connecting us all as Australians.”

It also had a startlingly hard time working out the relationship between the length of Australian election campaigns and how many seats the party in power they are likely to lose.

On some subjects it was amazingly comprehensive, churning out thousands of words on the similarities and differences in the ways Hindus and Christians perceive the mind..

But given the Plibersek error – which I knew enough to know was an error – and not being an expert on Hindu theology, how would I know if it was right?

The original Luddites raged against the machine ... literally.

And on other topics it was downright wrong (presumably because of the weight of James Bond in the world’s search engines, it said a martini should be shaken not stirred, an entirely correct take).

The sort of eerie thing was that through all our chats was the way it maintained a chatty breeziness.

It called me “mate”, asked me what “tracks” I was “spinning” when I posed a question about classical music, and seemed to subtly change tack if it thought I was disagreeing.

In short, it sounded like a clever undergraduate trying to fake it without having done all the reading.

I’m not the only one who is worried about where all this is going.

In a recent interview, songwriter Nick Cave told of his experiments with another AI engine to write songs.

“Initially, it was amazing,” he said, but after giving it several prompts (“dark, slow, gothic song about a banana” returned a song called “The Dying Peel”), he got bored.

“I lost interest in it. It was just this thing that could do this stuff, right?”

“But the thing is that it’s good; it’s kind of bland, but you can see the direction of it,” Cave said.

There’s no doubt that a lot of things currently turned out by people – like our humorous example of routine legal documents – will soon be the domain of the machine.

Not only a hell of a lot of jobs are going to go with this transition, but so too will the entry paths to far more senior gigs.

So where does all this land?

Will all creativity be replaced by semi-clever computer generated slop?

One thing is for sure: We will likely see not just economic dislocation but social and political dislocation as well.

As once was the case with blue collar jobs, where are the white collar roles and salaries going to be in the future?

And if millions of people are AI’d out of work, who’s left to pay for the products everyone and everything else is producing?

Real value will still be found, whether in writing songs or picking shares, in human processes and instinct no computer can replace.

But at the same time, those Luddites we chuckle at now may be the basis for new political coalitions that say, they had a point.