How an artificial intelligence expert successfully launched a new AI publication on Medium
Entrepreneur Thomas Smith talks about starting The Generator, writing on Medium, being a Boost Beta curator, and more.
Thomas Smith is a longtime Medium contributor and the founding editor of the new Medium generative artificial intelligence publication, The Generator. A graduate of Johns Hopkins University with a degree in cognitive science, he’s also the CEO of Gado Images, an AI-driven company whose purpose is to “digitize, capture, and share the world’s visual history.” Like many Medium contributors, Smith’s interests extend beyond his primary area of expertise. He’s also a photojournalist, tech product reviewer, and food writer. Recently, Smith began participating in Medium’s Boost Beta program, the aim of which is to help great stories on Medium reach even more readers across the network.
We talked to Smith about writing on Medium, starting a publication on Medium, and being a Boost Beta curator. His answers provide a roadmap for anyone interested in writing or starting a publication on Medium and a glimpse at some of the world-changing issues generative AI is raising.
His remarks have been lightly edited for length and clarity.
Jon Gluck: Before we begin, how do I know you’re not a robot?
TS: Good question — the output from things like ChatGPT is already so powerful. It hasn’t done video yet, but in a couple months you may not be able to tell. I’m working on a story right now where I cloned myself with AI and made a video version of myself. The technology isn’t dynamic yet, so this is really me, but if we have this conversation in another few months, it might be different.
JG: Like so much about AI, I find that equal parts thrilling and terrifying. In any case, now that we know you’re real, how did you get started on Medium?
TS: I got started in late 2019. I had worked in the AI and photography spaces at my company Gado Images for about 10 years. I had developed an expertise in those areas, and I wanted to share that in a way that could help people. Medium felt like a good place to do that because I wasn’t interested in building my own website or blog, and it felt like this was a platform where there would be a receptive audience for those topics. I went through the list of topics you see when you create your account, and I was like, Photography, artificial intelligence, machine learning … I can write about that. I now have more than 600 articles on Medium.
JG: How has your audience grown?
TS: It started out slowly at first. I remember seeing the first earnings report after a month of writing, and it was $7. I was actually excited by that because it proved that there was potential! It ramped up more quickly once I figured out the fit between me and the platform, which was, in most cases, the tech space and looking at how tech impacts business. It helped that I had knowledge built up from a decade of being an entrepreneur. Submitting stories to publications was another thing that helped me get more traffic. I’m also a professional photojournalist, and during the pandemic I started shooting photos and writing about what was happening. Obviously, a lot of people were trying to process that, so that was a big jump for me, serving that need.
A good month for me now is somewhere between 30,000 and 120,000 views on the platform. Those numbers can be a lot bigger if something goes viral. I wrote one story during the pandemic, about supercomputers’ analysis of COVID, that got about 11 million views. That kind of upside potential is one of the things I love about the platform.
JG: I assume your earnings are more than $7 these days.
TS: I’m typically in the four figures per month, and sometimes more, like the month where everything took off.
JG: Is money a significant part of your motivation?
TS: To me, content creation is a business, and I approach it as an entrepreneur. One of the easiest ways to tell if a post is creating value is if it’s earning money. On Medium, I like that when you bring value to the community, you’re going to participate in some of that value. On a lot of other platforms, creators are expected to churn out great stuff, but the platform sees all the upside. They dangle the idea of reach, that the algorithm will promote you, but you’re not being compensated when that happens.
JG: How do you feel about Medium’s reach?
TS: Medium has a large reach, but it also has the potential to spread your writing and thoughts far beyond the Medium community. The community is very engaged, but you also see external platforms tying into Medium and people reading Medium and quoting from Medium stories. You see excellent SEO impact; Medium ranks highly with search engines because it’s known for quality writing. You can capture people’s emails through a newsletter on Medium. And you can share links to your own blog or other sites. My Medium stories will send a lot of traffic to my YouTube channel and vice versa. It becomes this symbiotic thing where all the content ends up being elevated by the connection between this platform and the others I create content on.
JG: You recently launched a new publication on Medium called The Generator about generative AI. How did that get started? What motivated you to do that?
TS: My background is in cognitive science. I was studying this when it was, like, you do a neural network and it’s on a whiteboard and you draw the nodes and calculate the weights with a formula and paper with a calculator. And I’ve been writing about the subject for some time. Back in December of 2020, I wrote a story for Medium with the headline “Experimenting with GPT-3 Felt Like Witnessing a Technological Revolution.” Now it’s the same basic technology, but it has scaled up into this world-changing thing. It’s exciting to have an interest in a sort of obscure topic that’s taking off.
I want The Generator to be a home for people who have things to say about generative AI who maybe don’t have the platform to say it.
I’d already been writing about generative AI on the platform and it seemed like something that was reaching a tipping point with ChatGPT. With more public awareness about generative AI, it felt like we needed something to talk about that, and to talk about it from different perspectives. You often see great technical articles about generative AI, and that’s important to understand — how do I integrate with this API? How do I use this particular tool? But it’s also important to address it from a cultural perspective, an ethical, legal and philosophical perspective, an artistic perspective, from the perspective of people whose jobs might be disrupted. There are so many different ways that this should be approached.
JG: You write a number of the stories yourself, right?
TS: I do, but I like to bring in a lot of other voices, too. I can write about AI from my perspective — I have a degree in the field, ten years of experience, I was an Open AI beta tester from the earliest stages — so my perspective hopefully can deliver value. But also I want The Generator to be a home for people who have things to say about generative AI who maybe don’t have the platform to say it.
JG: You launched on March 1, which was almost impossibly good timing.
TS: With everything from GPT-4 to the launch of Google Bard to Midjourney 5 coming out, and the incredible explosion in the use of these technologies, this last month has turned out to be the most exciting, exhausting, and thrilling month in AI I’ve witnessed.
JG: How’s The Generator doing so far?
TS: In the first month, we published 44 stories, a little more than one per day, which I’m really proud of. I’m even more proud of the fact that 19 different writers have already contributed to the publication. It’s an awesome set of people, coming at AI from literally every angle and every level of experience, knowledge, and fluency in writing.
In terms of readership, the site has gotten about 100,000 views and more than 100 days’ worth of reading time in the first month. It’s humbling.
JG: Of the 19 writers you talked about that have contributed so far, can you break down where those folks have come from?
TS: The vast majority have come in through people reading a story where either I or other writers have mentioned The Generator. I’m also actively doing outreach to people who I know have great perspectives because I’ve worked with them or I’ve heard them speak. I’m trying to encourage them to come over to do an interview with me or to create a Medium account and post a story. That’s the next phase: continuing to bring on more platform writers, and also to find people who are on the forefront of these technologies off the platform and bring them in so that they can share a perspective that might not be on Medium yet. You’ll get pieces that wouldn’t necessarily be published in the New York Times, and that’s the idea—to give people a voice, even if their thinking is at a very early stage. You have the potential to see things that other people haven’t yet thought of before anyone else is seeing them.
JG: What are the most popular stories you’ve published so far on The Generator? What do you think made them so popular?
TS: A writer named Paul DelSignore wrote a great roundup about AI and the law. I think it was popular because people have heard about this technology now, but they don’t necessarily know what its impact will be. It’s gotten about 11,000 views so far. A story about ChatGPT and something called the dialogical self, by Samuel Sullivan, which delves into how to unlock greater self-awareness using generative AI, has also done extremely well. I think people resonate with the idea of, Hey, I can talk to GPT in this different way that might serve me, not just help me write better blog content. I’m fortunate to get early access to things like GPT-4 and Google Bard, and the pieces I’ve written on what I think of those, like this one, have been popular. And there’s a really awesome piece that I hope will get some traction that we just published where Sam Sullivan used Midjourney, an image-generation platform, and genetic test data on their dog to see if the platform could create an image that looked like their dog just from their genetic test results. He took the genetic test and the breakdown of what percent of different kinds of dog his dog was and formatted that in a way that Midjourney can understand and said, “Now make me a picture of my dog based on this description.” And the end results are sometimes disturbing, sometimes really weirdly accurate. That’s the kind of piece that makes me happy to see. It’s not just another technical essay. It’s something really novel and interesting.
JG: Let’s talk about the nuts and bolts of starting a publication on Medium. How do you find stories on the platform, reach out to writers, and so on?
TS: The first thing I did, besides create the basic page, was write out my brand strategy — the types of stories I hoped to cover, my plan for recruiting writers, what I wanted the tone to be. Then I took some of the stories I had written previously about generative AI and put them into the publication, so when you landed there on day one it wouldn’t be empty. (Those aren’t included in the 44 stories we published in the first month.) On March 1, when we launched, I wrote a post introducing the publication — what its voice is going to be, what the goal is — and invited people to participate.
It helped to be able to say I was in the Boost Beta program. I’m excited about the idea that community curators with specialized knowledge like myself can nominate stories we think should be promoted on Medium. It gives me the opportunity to find stories that maybe aren’t from people who have a big audience, people who have amazing ideas, but might not necessarily get the reach they deserve.
Because I believe in leading by example, I started publishing about one story of my own a day. I immediately got interest from people who said, “This is awesome, I’d love to write for you,” but then it was crickets when it came to actual stories coming in. After a certain point, that sort of tipped and I started to get two or three new story submissions a day. I think people were excited, but just needed some time to write their first pieces. Since then it’s been an outpouring. People see their stories can be successful and they can find an audience on the publication and on the platform, then more people seem to be submitting. We’ve found some amazing stories that way.
It gives me the opportunity to find stories that maybe aren’t from people who have a big audience, people who have amazing ideas, but might not necessarily get the reach they deserve.
At the same time, I’m searching the platform for stories that are suited to my publication, and I use SEO tools like Ahrefs to help me find generative AI stories on Medium that are getting a lot of interest. Then I’ll find the people who wrote those pieces and invite them into The Generator or boost their stories. Maybe a programmer who has worked on AI for five years in a deep technical way has a piece of advice they want to share, but their post isn’t getting the attention it deserves because it’s their first piece on Medium. If I can find that story and nominate it for Boost, then that’s value for the writer, for my publication, and for the platform. I also read as much as I can about generative AI off the platform to find potential writers.
JG: Let’s talk a little more about the Boost Beta program. As a curator, how do you decide if a story is worth nominating?
TS: Medium has shared a list of general guidelines. Then each curator applies their own expertise and judgment. Generally speaking, I like stories that are surprising or interesting or somehow original. Something the writer has firsthand experience with can be good. An historical example or pop culture reference that grounds the story in some kind of broader context can help. Those are the things that make me think, Okay, this is something I want to nominate for a boost.
JG: On the spectrum of wholly credulous on the one hand and cynically skeptical on the other, where would you say The Generator falls in its coverage of AI?
TS: As a publication, we don’t have an agenda. I would be equally happy publishing a super detailed, enthusiastic technical piece by a programmer who uses the stuff every day and a piece from a teacher or a writer who thinks it’s going to ruin their classroom instruction or take away their job. We want to share everything along that spectrum. Personally, I’m excited about it. I understand the social implications, and they need to be talked about. But I don’t share the existential dread a lot of people have about the march of AI or whether this is going to degrade the quality of human writing.
JG: The tech world has seen a number of wildly hyped Next Big Things recently — the metaverse, NFTs, crypto — that haven’t panned out as promised. Why is AI different?
TS: The biggest difference to me is that those technologies didn’t have an immediate way to tie into people’s business processes. It’s all in the future: the metaverse will exist with all of these applications for it and everybody will use Bitcoin instead of cash. Generative AI, on the other hand, can immediately tie into your work. If I’m a writer and I’m writing a blog post and I’m not good at writing titles, I can paste my article into ChatGPT and say, “Write me five titles for this.” Then I can use my own judgment and say, “This one doesn’t make sense, this one’s too clickbaity, and this one’s just right,” and a story I wrote that might have gotten buried will get in front of people. And that’s just one example in one industry.
JG: Did you use an AI to pick the name for your publication? Or the logo?
TS: I came up with the name. The logo and cover art were generated with DALL-E. The logo sort of captures the idea of building. It’s essentially a house, but constructed with the nodes of a neural network. It’s the idea of creating a space to discuss generative AI. The cover art is this organic looking, almost woven, design showing what could be nodes in a neural network, too. I also used Midjourney to produce the image that accompanies this story. I’m a photographer, but I’m not a good illustrator, so for me, letting the AI do that work was super helpful. I’m writing a story about it.
JG: What’s been the best part of running the publication so far?
TS: Interacting with such a broad set of writers has been awesome. I’ve worked with writers for other publications, where I’m hiring them and saying, “Alright, go write a 2,000 word essay about the best time to drive in San Francisco.” And it’s like, the writer will do it, but they usually do it for the paycheck and the passion isn’t always there. But the people who are writing for The Generator are extremely passionate about the topic — excited, fearful, angry, all the emotions are coming out in their writing. And that makes for some great interactions. And to see somebody bring something novel where I look at it and go, “This is really cool, I’m going to nominate this as a boosted post,” and then to see it get, in most cases, the boosted stories have gotten somewhere between 3,000 and 30,000 views — that’s really cool, too.
JG: How many pieces do you accept versus reject?
TS: We haven’t really had to turn down a lot of pieces. Usually I’m able to work with people to get their stories into a form where they can be published. In some cases, I’ve been able to recommend changes that I think will make a piece more likely to be boosted and that’s been fairly successful, too.
JG: What’s the most interesting feedback you’ve gotten from writers so far?
TS: One of the biggest things I’ve heard from writers is that they’re excited to have an opportunity to share their views on this stuff. In a lot of cases, there’s an expectation that generative AI belongs to the techies and that this is a subject you don’t really have any business talking about unless you’re a devotee of Elon Musk. But this is a world-changing technology that’s going to impact everybody. So everybody deserves to have a voice on it. Some writers say they never thought they would write a story about generative AI, but they have an idea that relates to the things they normally write about. A relationship writer, say, on where ChatGPT is good at giving relationship advice and where it’s not. Or an illustrator on what they think about Midjourney. That kind of thing has been great.
JG: What about feedback from readers?
TS: I’ve gotten a lot of enthusiasm for those first looks and the first perspectives. I think people love to see the technology before anyone has access to it, or a very small number of people have access to it. They like to see somebody sort of delicately but enthusiastically pick it apart and say, “Here’s where it’s good, here’s where it’s not.” And they like to get in on that and share their responses. Like, “Wow, I can’t believe it failed at that.” Or, “Whoa, I can’t believe it’s going to be able to do blank.”
JG: What are your goals for the Generator? Where do you want it to be a year from now, five years? What’s your strategy for achieving those goals?
TS: I’d like to see it expand dramatically. I want to get a lot more writers and perspectives into the publication. People who are scared, people who are angry, all of that. I’d also like to get broader reach off of the platform. I’d love to do more pieces that are evergreen so they turn up in search. To borrow a term from AI, I’d like to see The Generator become more multimodal. I’d love to tie in more videos, which I’ve been trying to do for my own stories. I’d love to tie in audio. Maybe add a podcast. People see my video demos on YouTube, and they come and read the detailed piece on Medium. They read the piece on Medium, and they find my YouTube channel. As I mentioned, it flows both ways.
JG: How should people submit stories to you? What are the nuts and bolts?
TS: Email me at tom@gadoimages.com. You can also follow me on Twitter at TomSmith585. I share a lot about the publication there, stats and stories and things. You can DM me. You can tweet at me. You don’t need to send a draft, just tell me what you think you’d like to write about. A couple sentences is fine. If you have a pitch you want to send, fantastic, send it. If it feels like you’ll be a good fit, I’ll add you as a writer. You can submit drafts.
JG: How quickly should people expect a reply?
TS: I try to review things within hours, certainly within 24 hours. I can’t guarantee that I’m always going to be able to do that, but that’s my goal. If you want feedback or editing on the piece, let me know. Some writers don’t necessarily care if a piece gets nominated for Boost; they just want to write it in the way that they want to write it. And if it gets an audience of 200 people, they’ll be happy with that. Other writers are like, “I want to get 10,000 views. I just don’t know how, so help me.” And that’s really helpful for me to know.
JG: Any other advice for writers?
TS: Once your story is published, that’s just the beginning. You want to be promoting it, you want to be sharing it with people, you want to be sending it out to your friends and family. You want to be sharing it on your Facebook. People feel squeamish about that sometimes, but it’s really important. For one, the people you’re connected to probably want to see your posts. And when people read and share your stories, it signals to the algorithm that people care about those stories and they should be amplified even further.
JG: Okay, last question: How do you know I’m not a robot?
TS: From the expertise I’ve developed over the years, I’d like to think that I can tell from the Zoom video. But again, stay tuned. In six months, will anyone be able to know?
Want to write on Medium? Get started.
Follow The Generator and learn more about Medium’s Boost Beta program here.
Also in this series:
- How Kaki Okumura’s Medium essays led to a book on health and human psychology
- How Devon Price redefined ‘lazy’ and turned his Medium essay into a book
- How Google data scientist Cassie Kozyrkov found success (and a fulfilling creative outlet) on Medium
- How U.S. Army veteran Benjamin Sledge found a way to share his experiences with the world
Who should we interview next? Let us know in the responses.
