Substack should have a category for AI.
Asking for a new category is not enough to deserve it, so the rest of this post is my attempt at convincing you (and hopefully Substack) that AI is a compelling case. To be fair, I will also do my best to refute the main arguments against my position.
When I came to Substack in mid-2022, AI was blooming and, to the surprise of many, had been for well over a decade. I realized that only a tiny fraction of the top tech newsletters had anything to do with AI. By the end of 2022, that indifference was no more: ChatGPT happened and, with it, AI found a permanent place in the mainstream.
At first, I thought the influencer-type daily roundups promoted on Twitter would be people’s preferred go-to choice. I was wrong. My newsletter grew more than ever in the post-ChatGPT period. I saw an unprecedented influx of writers coming to Substack, willing to put effort into creating insightful, high-quality, in-depth writing exclusively about AI. Readers, of course, followed.
The presence of AI-first writers on Substack is substantial today. 22 out of the top 100 paid tech newsletters in the tech category are exclusively or mostly dedicated to AI (some cover various related topics).
That’s almost 1 in 4.
Even if the ratio is impressive, extrapolating it to the entire category would likely yield a conservative estimate. The interest in AI has grown exponentially in the last year. Most AI-first writers are new to Substack and have had little time to climb the ranks. As you dig into the leaderboard I expect the ratio to be even larger. Not to mention free AI newsletters and writers focused on AI that belong to other categories, like business.
One counterargument is that, just as AI has attracted so much attention over the last few years, it could lose it equally fast.
Interest in AI is correlated with interest in ChatGPT (see graph below) so, as ChatGPT’s novelty inevitably fades, it’s reasonable to expect a comparable reduction in AI-related searches. Does it make sense to create a new category today that no one will care about in three years? Perhaps not.
I have two responses to this. First, ChatGPT and AI are intertwined concepts today, but they won’t be before long. Second, interest in AI won’t diminish for other reasons as it’s shielded by both 70 years of history and the most successful decade ever. Let me explain.
I expect interest-over-time curves in Google searches for “ChatGPT” and “AI” to stay approximately parallel for a while because, to most people, AI and ChatGPT are synonymous. However, ChatGPT will become a relic of the past when much better tools enter the scene (it’s happening as I write these lines) and as a result, lose traction without necessarily taking AI down with it.
The terms are associated but not causally linked. AI existed before ChatGPT and will survive it. AI, as an engineering practice and as a scientific discipline, will retain as much (or more) status and relevance as it has today thanks to ChatGPT’s successors.
Although ChatGPT’s decline won’t affect AI, it could fall out of grace for other reasons, like overhype and unfulfilled promises. So here’s my second response.
What people need to know about AI—especially those whose interest was awakened with ChatGPT’s release—is that it’s an old research field. It was officially founded in 1956 and even earlier works had a huge impact in current directions (e.g. McCulloch and Pitts neuron model in 1943). That’s 70-80 years of history. For reference, ChatGPT is 1.5 years old. The transformer is 6.5 years old. And the deep learning revolution is 12 years old.
It’s important to distinguish between trendy innovations that are actually trends and others that are instead late bloomers bound to be paradigm-shifting technologies once mature. AI may look like a trend but it’s evergreen (I’m not equally sure about generative AI, though). AI is everywhere already, even if it doesn’t feel like it.
Contemporary technologies that have gathered a comparable amount of followers, like cryptocurrency, are harder to defend as atemporal. I could be wrong because the future always has the last word, but crypto sure looks like a fad. It started to be a thing around 2009-2010 and although it peaked a few years ago, never lived up to expectations. In contrast, AI has already survived two winters.
Still, despite the evidence, crypto does have a category of its own but AI doesn’t.
And it’s not the only case where Substack has promoted a sub-category. A couple of non-tech examples: Politics is now subdivided into three—US Politics, World Politics, and Health Politics—and Literature, Fiction, and Comics are separate.
Doesn’t AI qualify?
An important nuance to the picture of extreme interest that I am painting here is that it goes both ways: AI is as hated as it is loved.
I can’t argue with that. Substack is home to writers trying to make a living through their passion (one of few such online homes that are truly worthwhile). Many of them of the creative type, focused on fiction, literature, and other categories similarly threatened by generative AI tools. They’re entitled to hate AI and I understand if they’d opposed my proposal. (If you think hate is too strong a word, then take annoyance, tiredness, or indifference—the counterargument is still valid.)
Why would Substack risk enraging writers by giving such prominence to AI?
My take is that AI being its own category doesn’t provide it with enhanced visibility (you have to look for it anyway) but instead, it provides writers (and readers) who can’t stand it or don’t care about it a simple means to avoid it altogether. It also doesn’t make it more threatening than it is. If anything, high-quality information readily accessible is the best resource to those who choose defense and opposition—which is a perfectly licit stance. (When I refer to AI newsletters, I don’t include AI-generated content, which I have zero intention to defend or protect.)
AI being its own thing is good for those who care about it, for those who want to never cross paths with it, and for those at the intersection.
Another counterargument (this is mind-reading of Substack’s executive) is that adding too many categories could dilute the value of the main ones and mess up the user experience of the platform (which I assume is what Substack cares about the most). Why give it to AI and not, say, European politics, macro and microeconomics, or have literature further split into novels, theatre, poetry, essays, etc?
I’m biased, but the answer for me is straightforward: AI passes the bar of importance in a way very few other sub-categories do. In the amount of writers and readers as well as in relevance to the world and people individually. If crypto passes the bar, AI does as well (I’m in favor of using the same measuring stick for every candidate; perhaps others pass, too, but I’m here to defend my case!)
Admittedly, this change would benefit most of all writers who, like me, focus on AI. But I see this as a win-win situation for everyone, now that AI remains a relevant and controversial topic in equal measure.
Ultimately, I think Substack as a platform would improve, as would the user experience of readers, both those who are interested in AI and, perhaps even more, those who are not.
Signed,
, The Algorithmic Bridge, Mostly Harmless Ideas , The Muse, Artificial Ignorance, Traction Design, Why Try AI?, Buy the Rumor; Sell the News, THE STARTUP from AFRICA, Teaching computers how to talk, Pandora’s Bot , Lotus Insight, The Daily Alchemist, AI Supremacy, Focus (English), Exploring Ideas Across All Horizons, The AI Observer, Educating AI, AI Changes Everything, GAI Insights, Paul’s Substack, The FuturAI, , AI Health Uncut, When Life Gives You AI, The Bayesiant
You can add your name by commenting on this article with your name and Substack newsletter or by replying to this note:
