ChatGPT – have you heard the one about Jewish artificial intelligence?

I wanted to tell you an AI (artificial intelligence)-generated Jewish joke about AI. Written in the style of Winston Churchill, or even better, Spinoza or Freud, because, well, they were Jewish. ChatGPT was going to make up this joke for me, because this is exactly the sort of parlour trick it loves.

But here’s the funny punchline. ChatGPT servers have become so overloaded with eager-beavers like me who are eager to find out its limits that it denied me access, complaining about user overload.

Ha! ChatGPT is impressive (and perhaps even funny) only when it feels like it. Some AI!

But seriously folks, what is this thing?

AI has been around for a long time, but it really got serious with the birth of computers about 70 years ago when some researchers found themselves beguiled by the prospect of making a computer mimic human brains. Some of the best minds on the planet tried to figure out a way to recreate intelligence. Alan Turing, the famed British mathematician, was one, devising the “Turing test” in which it’s stated that if a human doesn’t know whether they are communicating with a computer or a human, then AI has been achieved.

Well, no. We passed that point many years ago. AI needs to do more than trick a human into thinking it’s human. It must innovate. Learn. Invent. Create. Surprise. So, work has been determinedly underway for decades, including by me, who once wrote a piece of software to compose music and managed to get an academic paper published about it. It never made the hit parade, but still, I knew that this day would come.

Which day? Well, to put a pin in it, it was 29 November 2022 when the AI application called ChatGPT was released by a company called OpenAI on a largely unsuspecting public. And nothing will ever be the same again – at least for AI. Look up ChatGPT on Google. It’s as though the second coming has, er, come. Or the end of the world has arrived, depending on who’s hyperventilating.

ChatGPT is, indeed, surprising. You simply type a question into a dialogue box. Really. Any question. In plain, even error-riddled English. Long question, short question, impossible question, trick question. Ask it to write an exam (and give it the exam). It passed both the United States Medical Board License and the US Bar Exam, and a number of graduate-level Masters in Business Administration courses. It didn’t get an A, it got Bs and B-minuses. But hey, it’s about eight weeks old. Could you pass the Bar exam when you were eight weeks old?

Oh, and one of the professors asked GPT to grade an essay it had written. Apparently, it graded it as well as the professor would have. That’s pretty meta.

Not to get too deep into the weeds here, but here’s how it works. ChatGPT was fed billions of text modules (including the entire works of Wikipedia) from another conversational and text-based research effort called GPT-3. It then set to work looking for relationships, connections, entanglements between all of these text-based blobs, using fancy maths and stats. Some of these relationships have never before been spotted by humans because ChatGPT calculates faster than we can, can remember more stuff, and will soon have more connections than we have synapses in our brains.

And, most impressively, ChatGPT can write in perfectly structured, conversational English. Better than most people can. Not just sentences, but even longer texts, like essays and articles. It can write near perfect computer code in any language. And if you have quibbles with its answer, you can ask it to try again.

How much of a sea-change is this programme? Jerome Kahn from Fortune says it sharply:

“A few times in a generation, a product comes along that catapults a technology from the fluorescent gloom of engineering department basements, the fetid teenage bedrooms of nerds, and the lonely man caves of hobbyists, into something that your great-aunt Edna knows how to use.”

You can see the problem here. Many people are going to use this stuff to make their lives better, easier, more productive. A call-in listener for a recent radio show I was on said, “English isn’t my native tongue. But at work I have to write emails all the time. ChatGPT writes the email in perfect professional English.” A simple use case, incredibly powerful.

But, of course, there’s the other side. Education, for one thing, with its essays and assignments is going to break very quickly as students find this AI crutch. There are already American universities and schools which are moving to oral-only exams. The art of software development will get battered. In programming, practice makes perfect. Why spend three weeks coding when ChatGPT does it in five seconds. And obviously, journalists and writers are going to lose their jobs. That’s already happening as some major publications move to AI-produced content and don’t hide it

Yes, ChatGPT sometimes gives dumb answers – there are social media sites dedicated to its bloopers. But as I say, this baby is just out of the womb. The next version reportedly upgrades the system from 175 billion parameters to one trillion parameters.

Not to mention the furious competition to ChatGPT now sprouting up everywhere, like at Google.

This is going to be the year of AI. Perhaps the decade. Perhaps forever.

Sorry about not being able to give you a ChatGPT joke. The system is still overloaded. Ha!

• Steven Boykey Sidley is a professor of practice at the Johannesburg Business School, University of Johannesburg. He has written seven books, and is a columnist at “Daily Maverick”.