Programmers, beware: ChatGPT has ruined your magic trick | John Naughton

BEnedict Evans, Technical Analyst whose newsletter is required reading for those who follow the industry, made an interesting point this week. He had, he said, spoken to mainstream journalists who “always felt like ChatGPT was a trivial parlor thing and the whole thing was about as interesting as a new iPhone app.” On the other hand, he continued, “most people in tech are walking around slowly, holding the top of their head with both hands to keep it from blowing away. Inside of that, I think we can see a range of attitudes.

We certainly can – on a spectrum ranging from the idea that this “generative AI” is going to be the greatest boon since the invention of the wheel, to fears that it portends an existential risk to humanity, and many opinions in between. Seeking respite from the fire hose of conflicting comments, I suddenly remembered an interview that Steve Jobs – the closest thing to a visionary the tech industry has ever had – gave in 1990, and I dug it up on Youtube.

He recounts a memory he had of reading an article in American scientist when he was 12 years old. It was a report on how someone had measured the efficiency of locomotion for a number of species on planet Earth – “how many kilocalories did they spend to get from point A to point B. And the condor won – made it to the top of the list, topped everything else; and humans made it to about a third of the list, which wasn’t such a great achievement for the “crown of creation” .

“But someone there had the imagination to test the effectiveness of a human on a bicycle. A human on a bicycle blew the condor, all the way to the top of the list. very big impression – that we humans are tool builders and that we can create tools that amplify these inherent abilities that we have to spectacular magnitudes.

“And so for me,” he concluded, “a computer has always been a bicycle of the mind – something that takes us far beyond our inherent abilities. only in the early stages of this tool – in the very early stages – and we’ve only come a very short way, and it’s still in its formation, but we’ve already seen huge changes, [but] it’s nothing compared to what’s to come in the next 100 years.

Well, that was 1990 and here we are, three decades later, with a powerful, mighty bike. How powerful it is becomes clear when you inspect how the technology (not just ChatGPT) tackles particular tasks that humans find challenging.

Writing computer programs, for example.

Last week, Steve Yegge, a renowned software engineer who – like all uber-geeks – uses the ultra-programmable text editor Emacs, conducted an instructive experiment. He typing following prompt in ChatGPT: “Write an Emacs Lisp interactive function that appears in a new buffer, prints the first paragraph of A tale of two cities, and changes all words with ‘i’ to red. Just print the code without explanation.

ChatGPT did its thing and spit out the code. Yegge copied and pasted it into his Emacs session and posted a screenshot of the result. “In one fell swoop,” he wrote, “ChatGPT produced completely working code from a sloppy English description! With voice input hard-wired, I could have written this program by telling my computer to do it. And not only does it work fine, but the code he wrote is actually pretty decent Emacs Lisp code. complicated, Of course. But it’s good code.

Consider for a moment how important this is, as tech investors like Paul Kedrosky are already doing. He compared tools like ChatGPT to “a missile aimed, even unwittingly, directly at software production itself.” Sure, chat AIs can be great at producing undergraduate essays or creating marketing materials and blog posts (like we need more of them), but these technologies are great enough to work black magic for quickly produce, debug, and ramp up software production. and almost free of charge.

Since, ultimately, our networked world runs on software, suddenly having tools that can write it – and that could be accessible to everyone, not just geeks – marks a significant moment. Programmers have always seemed like magicians: they can make an inanimate object do something useful. I once wrote that they must sometimes feel like Napoleon – who could order legions, all at once, to do whatever he wanted. After all, computers – like troops – follow orders. But to become masters of their virtual universe, programmers had to possess arcane knowledge and learn specialized languages ​​to converse with their electronic servants. For most people, that was a pretty high threshold to cross. ChatGPT and its ilk just lowered it.

what i read

Write about
A masterful essay on reflection on writing by Helen Lewis on her Substack blog.

Meeting of the minds
A sharp analysis of the meeting between Xi Jinping and Putin by Nathan Gardels in November magazine.

The Monster Reveals is a Clever Try on the appeal of conspiracy theory in the hedgehog review by Phil Christian.

Leave a Comment