The staggering rise of ChatGPT

OpenAI’s chatbot reached a million users in an extremely short period of time. What does it mean for how we communicate?

An astounding chart appeared in the Times this weekend. It plotted the different lengths of time it took several tech companies to reach 100 million users of the product they offer. Netflix, for example, as a famously slow-burning offering that pivoted from disc-rental to streaming, took 42 months, or three-and-a-half years. Spotify took five months. Facebook took ten. Instagram reached the landmark after a cool 2.5 months.

ChatGPT, the online AI chatbot developed by OpenAI, took 0.16 months to reach a million users.

In other words, within five days of launching their free preview in December, OpenAI announced that they had received a million sign-ups (at least half a dozen came from NJF Capital’s office; the Times’s headline “How we all got hooked on ChatGPT” felt deeply apt by the end of the day). Outlets have scrambled as quickly to understand the software and as they have to come up with diverse takes on its implications, from the New York Times (“How ChatGPT Kicked Off an A.I. Arms Race”) to Wired (“ChatGPT Has Investors Drooling—but Can It Bring Home the Bacon?”) and, of course, everyone’s favourite clever contrarians at the Guardian (“ChatGPT isn’t a great leap forward, it’s an expensive deal with the devil”). The software has been hailed as the revolutionary tech development that no-one saw coming; an invention that will forever make redundant everything from coursework to novels to work-out plans.

It might not be a great topic for a blog, considering ChatGPT is also supposed to be coming for the livelihoods of copywriters, but it’s impossible to ignore. It is, as Wired said, “ingenious, garrulous, and occasionally unhinged”, as all the best writing tends to be.

And yet. For all its world-changing potential, it sometimes feels as though ChatGPT’s offering is still that: potential. The AI can write, certainly, but it doesn’t have a voice. Imagine the most boring person you know, or the smartest 14-year-old, and ChatGPT tends to compose the sort of thing they might come out with.

How long before ChatGPT does reach a sort of “writerly singularity” and comes up with such distinct prose that it feels like it’s sprung from the Comment section of the Times itself? How long before ChatGPT knows to drop in knowing references to early-Noughties pop acts if you ask it to write in the voice of a 23-year-old Vice contributor? Whatever your guess, it’s probably worth halving it. After all, Microsoft doesn’t bet $10bn on a bit of software unless it really thinks it has found the Google killer.

Perhaps the most likely situation is that ChatGPT will eventually lead us to a middle ground, where rather than replacing human writing and expression, the easiest, simplest written tasks will be done by AI, and those with a more distinct style and voice will continue to flourish. If you teach high schoolers, for example, you should probably be suspicious of ChatGPT’s impact on essay-writing, but if you’re the founder of Substack, are you going to feel threatened that the thousands of bloggers and essayists who share their thoughts on your platform are about to be replaced by generative writing? It seems unlikely. Compare AI to photography, whose development many at the time thought heralded the end of painting. It didn’t – it just made painters find more imaginative ways to create.

The writer Freddie DeBoer – himself a Substack user – offered advice for young writers on his own blog a few months ago. “If you want to make it as a writer, for God’s sakes, be weird,” he wrote. “In broad strokes: if you want to make it as a writer you will have to differentiate yourself, in text, from the vast rising oceans of texts that surround the digital world.” Ultimately, it seems like good advice. As long as you don't write like a robot, you'll probably be fine.