Antonym: The Pinterest beats ChatGPT Edition
Antonym: The Pinterest beats ChatGPT Edition
A year after launch, the $13 Trillion AI revolution is still less popular than virtual mood-boards.
Dear Reader
So much happened this week that I could have written a newsletter every day. Perhaps I should have.
OpenAI’s CEO spoke at its developer conference on Monday and said that the service had 100 million weekly active users. Most commentary was uncritical and simply gawped at the BIG NUMBER and the isn’t-it-all-amazingness of it.
Two things felt off: first, weekly user numbers is an odd stat to pick. Most online services report daily active users (DAUs) or monthly active users (MAUs).
Look at social networks in the graph below and you can see MAUs range from Facebook’s 3 Billion to Pinterest’s 465 million. That’s right: Pinterest probably has 3X more users than ChatGPT, the most powerful, openly available new technology in the world, the leader in a field that PWC predicts will contribute $13 trillion to the global economy at the end of this decade.
The second thing about the 100 million number is that this includes all the users of the free, more limited GPT 3.5 service. OpenAI doesn’t say how many people pay for the much more powerful GPT 4 service, but estimates for freemium services (where some users pay for extra performance of features) usually expect 2 - 7% of total users to pay, maybe 10% at most.
That suggests less than 10 million people using GPT 4 every week, less than two or three million every day. (Yes, I’m playing fast and loose with the sums, so please do your own beermat calculations.)
We’re not comparing apples with apples here. But the surprisingly low usage of this wonder-tech is a humbling reminder that however powerful your box of AI tricks, you may be performing them in a bubble.
So what?
Why does this matter? It’s not that everyone should be using GPT 4 but to make informed decisions about the technology we all need to know or have advice from people who have tested and have a sense of the power of the latest models. Organisations are making calls about whether and where to invest time and money in AI – but may not be getting as much perspective as they need. There is a lot of confusion about whether and what the potential of AI is and at the same time it is developing very quickly.
In the last month alone, users of the premium version of ChatGPT have been able to start building bots, have the system analyse pictures, create photorealistic images good enough for commercial use and have conversations with the app. I have tried all of these features and been shaken by how good they are and how immediately useful. We put them to work straightaway. Some of my colleagues have been able to write useful software, even though they know very little about coding.
Next year billions of dollars will be going into developing the next generation of Large Language Models (LLMs). Many improvements will be based on the experience and data created by the real world use of the current models.
Rather than focusing only on the technology, it’s better to think of human minds and AIs as two systems that get better by interacting with each other. The tech will get faster and the experience of the best users will get deeper, their skills more powerful.
That’s why I’m convinced the best policy is to invest in human + machine capabilities now, and not wait for tech vendors or consultants to rock up in the near future with packaged solutions.
Here’s a bit of our webinar where I talk a bit more about the scale of the AI models’ growth:
Are freelancers the AI canaries?
The optimistic view of Gen AI in white collar or knowledge work is that it will remove drudgery and that the extra time workers get back can we be put back into more interesting work. But with big tech slashing jobs and a grim economic outlook, it seems to be making things worse before it gets better.
An excellent piece by John Burn-Murdoch in the FT this week looked at the emerging evidence on AI’s effects on productivity and employment.
One study showed that in the five months after ChatGPT launched freelancer jobs and pay both declined.
These are short-term effects, made worse by a deepening advertising recession in the marketing sector (which employees a lot of freelancers), but it doesn’t bode well for workers in the short term.
The demand for freelance marketers is dwindling in part due to in-house or agency teams using AI tools to cover some aspects of the work. However, this short-term effect is ironically fueled by a lack of understanding on how to effectively use AI tools.
It is possible to construct an promotional plan and content swiftly using tools like ChatGPT, yet it could result in subpar work if not operated by knowledgeable personnel. The key to effective marketing has proven to be strong creative ideas. But without human experts who can distinguish good from bad creative content, low-quality work might be accepted for marketing purposes.
It's not all bad news, according to Burn-Murdoch:
Taken together, the studies tell us three things. First, regulation will be key. Online freelancing is about as unregulated a labour market as you will find. Without protections, even knowledge workers are in trouble.
Second, the more multi-faceted the role, the less risk of complete automation. The gig-worker model of performing one task for multiple clients — copywriting or logo design, for example — is especially exposed.
And third, getting the most out of these tools, while avoiding their pitfalls, requires treating them as an extension of ourselves, checking their outputs as we would our own. They are not separate, infallible assistants to whom we can defer or hand over responsibility.In millennial-speak, generative AI is the white-collar worker’s frenemy. It’s wise to be wary, but this could become a flourishing relationship.
Men at tech conferences need to shut up
I appreciate the irony of writing a long letter about things I think and then talking about how men should shut up. Rachel Coldicott’s Twitter thread on male privilege and bad behaviour at tech conferences and her Substack post on the same topic are very useful reading, especially if you’re a man who doesn’t think they are sexist and gets a bit prickly at the suggestion.
When we ran the Dots series of conferences we worked hard on getting gender balanced line-up of speakers and banned questions. The latter was motivated by the fact that questions are usually tedious for everyone more than anything else, if I’m honest. Comment-not-question and three-part-questions being the main vibe killers in my experience.
This week I’ve been…
Watching…
It’s hard to tell if it is the hangover from the American writers’ strike or slowing investment in streaming content, but there have been fewer things to get excited about in TV recently. Last week’s recommendations were both movies, while this week we’re going backwards to a pandemic cult hit.
The Midnight Diner (Netflix) is odd, story heavy, people-are-good emotional broth. Half hour episodes make it a light and agreeable entertainment supper.
The other post-lockdown hit I’ve been revisiting is Mike (Disney+) which is incredibly good.
Generation V (Prime) is probably only going to make sense if you watched the series it has spun-off from The Boys. It’s gory in tone and goring in its satire of contemporary American culture. Generation V takes place on a campus for superheroes, most of whom will be destined for a future in theme parks and content from the Vought Corporation, the company that has created “the supes” with its dangerous and unpredictable chemical treatments. Imagine if the Marvel Universe was real and run by Trump Inc. and you’re on the right tracks. Dark, weird and fun.
Reading
A tidal surge of news and work has knocked me off balance with my reading. A Hacker’s Mind by Bruce Schneier is my priority, while in terms of fiction I’ve retreated to re-reading the dark pleasures of The Heart Goes Last by Margaret Atwood. It’s a story about a community where half the population live as prisoners one month and the other half the next, in order to share resources in the town’s simple but comfortable suburban homes.
That’s all for this week, folks. Thank you for reading and if you liked it, please share it!
Antony
For more of this sort of thing: I’m speaking at the Chartered Institute of Marketing (CIM) event on AI on November 23rd in London: details here. This week Brilliant Noise published an analysis on the OpenAI announcement of a GPT Marketplace in our BN Edition newsletter.