News

When Barry Humphries died last week, a third ventricle burst in the hearts of many of his fellow Aussies. It was hard enough with Robert Hughes and Clive James gone. We pray Germaine Greer keeps pumping forever, so to speak.

Some may wonder what the fuss is about. But a sense of loss for a person or thing one has scant to do with is a universal phenomenon. I was told recently that Zurich has been in a sombre mood since Credit Suisse passed away.

Coincidentally, Humphries and the Swiss bank have occupied much of my time over the years. And odd as it seems, they are both responsible for the most rewarding lesson of my career — one I’ve been mulling over with regard to artificial intelligence and what its believers are promising us.

Let me explain. Humphries was famous for playing outrageous comic personas who would have been cancelled at birth today. Barry McKenzie and Les Patterson were bigoted misogynists. Dame Edna was a housewife — appropriation!

How did he get away with it? The characters were hilarious and often had an eyebrow raised, which helps. But crucially, in my view, audiences knew Humphries was an intellectual.

Even if you hadn’t seen his 50,000-volume library, erudition seeped through, no matter how vulgar the joke. Humphries had a confidence and range when performing that comes only from a deep knowledge of the world.

Indeed, when I first read his scarcely believable autobiography, More Please, it was hard going. My sizeable dictionary didn’t contain half the words Humphries dropped into every page. I had to buy a bigger one, and even then I struggled.

Why did someone so adept at connecting with audiences in simple English use such obscure words in his book? Years later, a consulting project at Credit Suisse opened my eyes to how this particular confidence trick works.

It was my debut assignment. A full European strategy review for the board. No sleep for a fortnight. Thousands of numbers crunched. Hundreds of slides (our advice was ignored so don’t blame us!)

The final presentation was only six pages long. A weighty appendix was not even referenced and probably ended up in the bin. My boss assured me the latter had served its purpose, though. It landed on the table with a thud.

He explained that you only get away with showing the top of a pyramid if the base is solid. Just as Humphries could crack penis jokes because he had read the history of the male form, our three recommendations relied on Credit Suisse knowing that we had done the groundwork.

What is more, just as humans have an incredible nose for hypocrisy, they can tell if you don’t have the foundations in place. There’s no fudging it. I learnt this again years later, when I was writing and editing the Financial Times’ Lex column.

The Lex notes that failed were the ones where we didn’t crunch the numbers. When the two extra calls weren’t made. I can’t explain why, but readers somehow sensed when we rushed to a conclusion, even if it was correct.

If we’d sweated, however, it was amazing how little analysis we had to reveal. Half the note could be anecdotal and readers didn’t mind. Somehow, the effort shone through regardless — just as it did in every gynaecologist pun that Humphries made.

I took this revelation to my next job in charge of research content at a large bank. We began to collect data on how many of our publications were being opened and the number of pages read. The obvious conclusion was that we should write much less.

This would have been the wrong way to go. I knew clients needed the 100-page report, even if they didn’t read past the summary. Even if we had the most accurate forecasts.

That such elbow grease is still admired when dealing with ideas may seem particularly incongruent. But it’s true. And just as we wouldn’t admire Humphries’ comedy as much if not for his scholarship, making our lives easier by ridding humanity of graft is a false goal.

Yet that’s exactly what AI advocates are saying the new technology can do. Large language models will shoulder the back-breaking work so we are free to ponder the pointy end of our pyramids (as Sir Les might have said).

But you cannot flaunt an appendix if ChatGPT has written it. In the future, how will we prove we have laboured when we won’t have done? AI can only triumph if it incorporates the secret of Dame Edna’s success.

stuart.kirk@ft.com

Articles You May Like

Biden vs Trump: a week of contrasting presidential campaigns
The Paris Olympics will be political. That’s a good thing
Hobbies are not supposed to be productive endeavours
Ether ETFs appear set to launch on Tuesday, six months after massive debut for bitcoin funds
Loyola University of New Orleans gets negative outlook