The Exchanges

Q I think. So, so computer science, very different then, than it, than what it is now. I read that you wanted to study, and I was blown away by this, large language models in the early seventies. I didn't even know, well, you think of LLMs today and AI, right? You think of generative AI. I didn't even know that that was something you could study in the early seventies.

A They didn't use the term large language models. I was influenced by the people looking at neural networks, which are the basis of large language models. The deal is I got real interested in that. I got interested in the idea of natural language processing, thinking that was a core to understanding consciousness. Because I had been very influenced by a lot of science fiction, but the idea is that I was studying this in 73. I remember back then taking an AI course, writing a theorem prover in a language called Lisp. I don't recall what a theorem prover is anymore. And all I remember about Lisp is that it stands for lots of irritating, stupid parentheses, which is what you see when you look at Lisp code.

AI assessment note: “They didn't use the term large language models. I was influenced by the people”

Q Okay. You graduate, um, from Case Western, and like a good computer science graduate, a responsible young man in the mid-seventies, you go and work for one of the biggest companies in the world, IBM. Uh, and, and you, I think initially you're in Boca Raton for some time, you're in Detroit for, for part of the time. What were you doing for IBM?

A Well, uh, in IBM Boca Raton, which was not very fancy at that point, I went to work on a forgotten minicomputer called the Series One. I did some good work there, did some good software development, but the, uh, you know, it was never taken really seriously by the whole, uh, industry. The only thing I did which I thought was, was really good was I was a core member of a team building a series one as a card that would fit inside of a PC. My big mistake was not getting more and more involved with, uh, the PC back then, but, uh, I got a good education. I made the right mistakes, and, uh, it's worked for me.

AI assessment note: “I went to work on a forgotten minicomputer called the Series One.”

Q I want to know what's going on, can I post a job here? Can I, I've got an apartment or a room for, for rent. So it just kind of naturally becomes this thing. You're not, you're not necessarily architecting that. It's just, you've got a list, but People are reading it, and then other people are like, oh, can you add this thing to your email list this week?

A That's it. I had caught the vibe of the time, as the kids might say, and in the way I started Craigslist, I created its own vibe, where user-generated content became a genuine community thing, and my problem was that it was becoming a little too successful too fast. I got lucky. A guy named Eric Theis offered me the use of his Listserv. He had a real server, which was, uh, an astounding 1.5 megabits. And, uh, the dilemma I had was when you're using a Listserver, you had to give the thing a name. As a nerd, I'm literal. I wanted to call it San Francisco events, but there were people around me Who are much smarter than me. Like a guy named Anthony bat. He told me people were calling my mailing list. Craigslist. He told me that I had created a brand. He was right. I had created something called Craigslist.

AI assessment note: “That's it. I had caught the vibe of the time”

Q the collapse of the newspaper industry. So of course, What does the newspaper media industry love to cover more than anything else? Itself. Right? So, a lot of attention is being focused on Craigslist and how it's Responsible, allegedly, for the collapse of local newspapers. Did you feel like you were being personally targeted by critics of Craigslist for the alleged, or the, what would become the demise of newspapers?

A I, in the past, felt that Craigslist did have a substantial effect on newspaper survival, but until No one had done any of the numbers. I'd spoke to industry analysts like Ken doctor and Rasmus Nielsen, who said it was ridiculous to think Craigslist had that big of an effect. And they had intuited that the cause of the newspapers problem goes back to the fifties and sixties. TV news interfered with newspaper revenues because they've replaced display ads and so on. But I really didn't ever see the numbers until Thomas Bechtel published, uh, some serious research in 2018 showing that newspapers started suffering in the sixties. He ran all the numbers, found that there's just continuous straight line decline with newspapers until around 2008, at which point Facebook and Google started getting a lot of that revenue. And that, uh, trend has continued. He, uh, tells me that if I feel Craigslist had that big of an effect, I'm deluding myself. What really hurt newspapers was TV news followed by Facebook and Google. Uh, but, and this is something that I find very frustrating. Whenever an article is written about this, the reporters never look up the research. They just start with the conclusion and continue with their, ah, with the conclusion that they've already made.

AI assessment note: “Whenever an article is written about this, the reporters never look up the research.”

Q model, which is still more or less the same today, is you would charge for job ads, some real estate listings. And a few other things, but most things would be free, right? And, and, and, and again, in 1999, you're still manually put, like, people are, are emailing into the email address, and somebody you, or now you're paying people, are manually taking that ad and putting it up?

A Well. At some point in the evolution, I don't recall when, we automated a lot more in terms of posting, and we relied on the community to let us know when there was a problem, but we still needed a substantial team. For example, we had a technology team, we have a customer service team, we had a billing, Very, very briefly, I think in 2000, we tried a little marketing, put a couple of job ad ads up in HR magazines that lasted maybe a month. Um, so we ran, uh, very lean, and that was through, uh, 99. But towards the end of 99, I had hired a guy named Jim Buckmaster to take over technology. No one needed me to do technology anymore. But I was CEO and I was a customer service rep.

AI assessment note: “we automated a lot more in terms of posting, and we relied on the community”

Q Yeah, for sure. Um, okay, so, so you have a competent, strong partner now running the business, and you've got cash flow coming in, and this is really, I mean, still, 90% or more of the ads were free, but you were making enough money on the ads that weren't free. People would pay, what, 25 to 50 bucks for a job posting, for example?

A Uh, that was it. Um, the philosophy Which was first articulated by the people who use the site. They were saying, Hey, charge people who normally pay a lot more money for less effective ads. And that meant charging the people who put up a job, not the people who put up a resume. We would charge, uh, apartment owners, homeowners for their ads when they wanted to rent stuff out. We would not charge the people who were looking. And that's the philosophy, uh, to today. The deal is that Craigslist has done, uh, better than anyone expected. Our business model is, uh, in a way it's doing well by doing good. We've, uh, avoided becoming gentrified like so many sites. Jessica Linga in her book, The People's Internet, said that we're one of the few remaining un-gentrified sites, and I love that because I look at, uh, neighborhoods which were only partially gentrified or not at all, some of which are really successful, some of which are really about community, And I like that. I like that in neighborhoods. I like it in websites.

AI assessment note: “Uh, that was it. Um, the philosophy Which was first articulated by the people”