Pop culture obsessives writing for the pop culture obsessed.
We may earn a commission from links on this page

The oral history of Radiohead’s GooglyMinotaur, a chatbot ahead of its time

We may earn a commission from links on this page.

When Radiohead released Amnesiac on June 5, 2001, the album arrived with a mysterious online companion: GooglyMinotaur. The Where The Wild Things Are-like character was a goofier take on Stanley Donwood’s figure on Amnesiac’s cover, and he was friendly. AOL Instant Messenger users could add the GooglyMinotaur to their buddy list and chat with him like they would a friend or family member. He was music’s first chatbot, an “interactive agent” that used natural language to have a conversation.

As it turns out, the origin stories of GooglyMinotaur (affectionately called “Googly” by those who worked on it) and its popular peer, SmarterChild, are inextricably linked. Both bots had a deadpan, snarky sense of humor but tended to be genuinely helpful if users had questions. That was no accident: The team at ActiveBuddy, the start-up behind both GooglyMinotaur and SmarterChild, was deliberate about how the bots responded to and interacted with fans and meticulous about refining answers to queries.


That human touch perhaps explains why both bots developed rabid cult followings. According to a 2003 New York Times article, 250,000 people were talking to SmarterChild every day when it was most popular. Fifteen years after Amnesiac’s release, ActiveBuddy’s work with bots and communication looks incredibly prescient. Apple’s Siri is the sporadically helpful robot buddy iPhone users love to hate, while Amazon Echo has its own built-in assistant-like voice persona called Alexa. And in April, Facebook launched bot development and implementation capabilities for Messenger—leading to things such as And Chill, which trades in Netflix-based movie recommendations.


To mark the 15th anniversary of GooglyMinotaur, original ActiveBuddy employees and then-members of Capitol Records’ new media department shared their memories of the innovative bot—which today is merely one of Radiohead’s many forward-thinking, career-defining promotional gestures.

The origins of ActiveBuddy and SmarterChild

“I could have sworn I saw a light coming on”

Tim Kay and his brother-in-law, Robert Hoffer, were start-up vets who had worked together at a directory services company called Query Labs. (In the ’90s, their company went head-to-head with Zip2, owned by future PayPal co-founder Elon Musk.) In January 2000, Kay and Hoffer hit on an idea for computer-based, instant-messaging-like technology that would answer questions posed by users.

Tim Kay (ActiveBuddy co-founder/chief technology officer): There were precursors to Twitter at the time, on SMS [short message service], and instant messaging was new, too. We hit upon this idea in that very first call. It was like, “Well, you’ve got instant messaging. Let’s just hook up computers to it, and it can answer your questions.” I spent a few days doing an early prototype, which went really well. It was pretty straightforward. It did stock quotes and some other things like that.

Robert Hoffer (ActiveBuddy co-founder): The idea of bots was not me. That was a thing that we’d had. Artificially intelligent robots, or artificially intelligent computer personas, was ancient in the computer business.


Kay: I think the first investors [Robert] talked to gave us $4 million. The idea was in January. We had $4 million in the bank in February. That’s how crazy the dot-com bubble was: We got $4 million in investment in two weeks’ or three weeks’ time.

The nascent company, ActiveBuddy, was bicoastal, with offices in Sunnyvale, California, and Manhattan. Kay developed a proprietary programming language called BuddyScript, which Hoffer’s team then used to code the nuts and bolts of ActiveBuddy’s interactive agents, or bots. These BuddyScript-based bots could process natural language and respond in kind, which facilitated two-way conversation.


Hoffer: We were way first to the market with the idea of having bots in the instant messenger windows. [But] how are you going to even do it: natural language-processing bots in 2001? Can it be technically achieved? We came up with a plan for doing that. We had a whole technical team in California that we hired to do that.

Kay: It used to be when you tried to do natural language, you had to have a Ph.D. and had to have tape on your glasses and wear a white lab coat. BuddyScript was approachable. People who had the ability to do Excel-type programming could also create natural language.


The goal with academics for doing natural language was, “If I just got the right 100,000 lines of code, then it would be smart and it would come alive”—you know, artificial intelligence and stuff. The way we designed it, we weren’t expecting that to happen, but we expected that if you put an hour of work in, it would get an hour better. That’s really what made SmarterChild. People put lots of time in, and it got better with every hour put in. We had a lot of resources behind it.

In Sunnyvale, we did the technology, which included the BuddyScript language. In New York, they then implemented the applications in BuddyScript. It was just like a perfect tension, which you see when things go really well. We explained how [BuddyScript] works, and [the New York team would] write something. We anticipated they’d write five or 10 lines of code to answer stock quotes, and they’d write 100 lines of code. We thought, “Ah, geez! It runs too slowly!” Then our guys would work really hard and make it run faster. And then they’d write 1,000 lines of code, and it went back and forth. It got very sophisticated very quickly.


The story behind SmarterChild is interesting. We put up a prototype [and] called it ActiveBuddy. We put it in front of the AOL guys, for example, as well as some others. They played with it for a couple of days, and then we said, “Let’s shut it down,” because we don’t want them to have too much access to it until they make a commitment. So we needed a new name to put it up, which we wouldn’t tell them about. And that’s why I picked SmarterChild, just off the cuff.

But it was still a really boring product—what we now call the “old white guys app.” You know, we put our heads together and said, “What should this be?” and said, “Oh, it should do stock quotes and news and weather, and it’s great! This will be perfect for our 13-year-old-girls audience!” [Laughs.] There was no resonance.

Still, ActiveBuddy’s staff was anything but staid. There were musicians-turned-programmers such as Chris Bray, who was into glam rock and now runs Apple.com, and talented illustrators and graphic designers. In fact, as the company’s technology started coalescing, these other departments became more important: They helped these eventual forward-facing bots—including SmarterChild and GooglyMinotaur—develop their individual personalities.


Jennifer Villany (ActiveBuddy project manager/developer): We really wanted to think about how to automate things in a way that the bots were self-learning. A lot of our work in 2000, 2001, was really manual. And we loved that, but like any technologist or team, you always want to be evolving and making your technology better.

For us, it was always very interesting—like, where could they just have memory and start to build their own personalities? We would create the system or the functionality, so that it could be self-learning. Or we could, with some variables, allow for it to be programmed male, female, different attributes. Is it surly? Is it friendly? That sort of thing.


Hoffer: Christer Manning and Ed Hepburn were our graphic geniuses. SmarterChild, as a bot, came out of [their] minds, [and they] are both brilliant illustrators. Ed, particularly, is brilliant at coming up with these goofy, cute little characters. And Christer was able to give them personality. We even had banner ads where SmarterChild, as a bot, would come rolling out. He’d see Ask Jeeves, which was our enemy at the time, [who] would say some stupid stuff. [SmarterChild’s] head would open up, and a ray gun would come out, and he would shoot. [Laughs.] Those were all Christer.

Villany: [There] was massive learning about instant messaging culture. We wanted to be appropriate and contemporary with how we responded. Then we would dive into, “How are people chatting with each other? How would we either replicate that—or what would we do that might be a little bit different, but yet funny, for a bot to respond to?”


Pat Guiney (ActiveBuddy copywriter/editor): I first came to ActiveBuddy from another internet start-up of that era that was failing. This was 2000. I was hired as the writer/editor for ActiveBuddy. Basically, the idea was that I would help give the interactive agents, or whatever it is we called them, some personality in their responses. Not just delivering factual information, but doing it with a little bit of humor and making it more compelling, readable, and fun.

I remember on my very first day, I was given a long list of the most obscene profanity you can think of, and my job was to try to think of responses to it. In other words, if someone typed some incredibly offensive thing to one of our chat buddies, how should we respond? That was a fun introduction to my job.


Hoffer: [Pat’s] a very talented, sardonic writer, that was essentially the underlying foundational personality of all of our bots. When SmarterChild hammered you into an apology, that was all Pat Guiney.

Villany: Many of us joked, because when you talk to Pat and you know Pat, and then you talk to SmarterChild—it’s Pat. [Laughs.]


Capitol Records and ActiveBuddy find one another

“You and whose army?”

Rumors of ActiveBuddy’s cool new technology reached the Capitol Records’ new media department, which was full of innovators. In fact, it had recently launched an online stream of Radiohead’s Kid A—a completely novel and visionary promo move in the year 2000.


Hoffer: The first real commercial bot—in that, the first bot that we actually got paid to do—was the Radiohead bot. In fact, GooglyMinotaur was instrumental in ensuring the success of SmarterChild.

Robin Sloan Bechtel (Capitol Records, head of new media): We had been looking for and discovering new technologies in the Capitol Records new media department since the early ’90s, at the web’s infancy. We would find the right technology and then come up with the right marketing application for it, with the right band. Or we would have an idea and find the right technology, or build it ourselves.


At this time, we were thinking of potential pioneering ways to market Kid A on the web. We wanted to come up with concepts that no one had done before. The music industry had never even streamed a full album on the web yet, which we planned to do.

Meanwhile, we had heard some whispers about SmarterChild and its parent company ActiveBuddy, that their technology was “mind blowing,” but not much else other than that. What they were doing was very mysterious, so much that we could not even find them or get ahold of them.


Hoffer: The Greenwoods, Jonny and Colin, saw SmarterChild and talked to their manager and reached out to us through their label and wanted to actually put a Radiohead version for the release of Amnesiac. They wanted to have a bot just like SmarterChild. (Neither Radiohead nor its management is doing press, according to the band’s publicist.)

Sloan Bechtel: Our new media department was constantly tasked with challenges and hurdles trying to innovate in the early years of the web. We had a saying: “Can’t lives on won’t street.” So we tasked our brilliant assistant, 18-year-old Mio Shibata, to once and for all find ActiveBuddy. She took it upon herself as a personal challenge and became a super detective and got them to call us back. Their question was, “How did you find us?”


Then we had to convince them to work with us.

Hoffer: The board of directors was completely constipated on this. It was a five-day fight where we split the board: “Do we do this, or do we not do this?” Three members of the board wanted to do it: me, [venture capitalist] Andy Weissman, and [ActiveBuddy co-founder] Peter Levitan, who was our CEO and one of our core investors. Two other guys [including Tim Kay and future CEO Steve Klein] didn’t know who Radiohead even was. That’s how clueless they were about popular culture and what we thought we were doing as a business.


It was one of those amazing dichotomies that you find between the suits and the creatives. They were interested in selling the technology of the platform, which we thought was just a means to an end.

Kay: I was skeptical that there would be traction, because it’s the same as SmarterChild to a large degree. And I was wrong about that. But I don’t recall having any business objections. I just may not be remembering.


[The New York office] could pull together these media presentations, these great bots that could talk about a specific topic fairly quickly. I was totally impressed. They were able to reuse a lot of the assets. It’s easy, obviously, to just change the hangman game to be pertinent to the particular topic. As a technical guy, I thought, “The audience is never going to buy that.” Right? That’s just obviously what it is, a rehash of the existing. But they ate it up! They just loved it! Nobody minded that it was the same thing over again to a degree, plus implementing a lot of pertinent questions about the particular topic. Our technology made that pretty easy.

Hoffer: Ultimately, somebody had to make to make a decision, and it was Peter. Peter said, “We used to say at [ad agency] Saatchi & Saatchi, when you get a gift from God, you take it.” That was the decision.


ActiveBuddy inked a deal to create the Radiohead bot in April. Called GooglyMinotaur, it was set to launch in tandem with the early June 2001 release of Amnesiac. Shaping the look, feel, and tone of the bot so it aligned with the band’s skeptical, enigmatic, anti-marketing, and anti-commercial stance was next.

Guiney: You could think of a product like GooglyMinotaur as consisting of two things. One of the things was legitimate answers to questions about Radiohead. And also switching out messages—if you just said, “Hello,” it would try to talk to you about Radiohead’s new album. Of course, people did a little bit of that. The other part of it was open-ended: “How do we deal with people asking the craziest questions you can imagine?” And that was what I really took charge of at my job with GooglyMinotaur.


Sloan Bechtel: The band came up with the name GooglyMinotaur for the bot. It’s a character in their artwork for Amnesiac. The name really helped give him a personality, since they already had artwork of him as a character.

Villany: For me, the dev part [of my job] was a no-brainer. The stretch was the project management part. I was responsible for managing the timelines and the deliverables and the client expectations around it but also working with the engineering team to build.


Capitol shared with us the overall arc, creative, voice concept. Then we had two creative directors, Christer Manning and Ed Hepburn, who evolved that voice for the bot personality, if you will.

Kay: The icons for example, you saw the little AIM icons? That was Ed Hepburn, and he was just brilliant at doing those things.


Guiney: Aesthetically, I remember everything was lowercase, and there were these interesting graphical kind of things we did with the text that were unusual and interesting. That really helped give it a unique look.

Jennifer Bird (Capitol Records, director of new media): We were working with the fan sites starting with the Kid A release, and Googly was launched with Amnesiac. So we already had a great working relationship with them, and they were very supportive of everything we put out digitally. Three I remember speaking with on a regular basis were Beryl [Tomay] from Follow Me Around, Jonathan [Percy] from Green Plastic, and Adriaan [Pels] from Ateaseweb, who I still email with to this day. But we weren’t picky. We sent info out to every fan site we could find, and the response was fantastic. (Tomay, Percy, and Pels did not respond to interview requests.)


Michael Beam (Radiohead mailing list administrator, 1996-2002; ran the Underworld.net server, which hosted several large Radiohead fan sites): [Radiohead wasn’t] big on self-promotion. They left that up to the label. They let their fans on the internet do it organically—and their online fans were very savvy.

Villany: [I had to] learn Perl and MySQL, because we had to create the databases to put the content in. For example, all the bloggers that we worked with on GooglyMinotaur would send us their blogs, discography, their FAQs around the band, any of their fan content. We would have to ingest it and then put it into a SQL database so that we could modify the content in the right way, so that Googly responded with that content.


Pat would write the copy that surrounded those responses, or introduced those responses, but I was creating the database and the mechanisms to respond back with.

Guiney: We had a lot of fun. Of course, we had a lot of leeway just to use our imaginations and do what we wanted. I don’t remember getting lots of specific guidance about what it should say or what it should not say. Certainly we had to imagine what a bot representing Radiohead would be like and what it would want to say.


I love Radiohead, actually, so it was really exciting to be involved in anything connected to them. While it was going on, all I could think of when we were working on it is trying to put myself in Thom Yorke’s shoes, if such a thing is possible, and thinking, “Is this thing embarrassing and dumb, or is it kind of cool and interesting?”

Villany: We loved diving into the subject matter and the story and what the personality could be of this bot. Let me tell you, there were some very heated nights and days of arguing over, “Should Googly respond in this way? What’s the logic behind that?” [Laughs.] Honestly. And I loved that. [There was] a lot of passion with that team.


Bird: It was a new concept for a record label to know and embrace the fans and give them tools to promote the album and to let them contribute to the design and content in the technology.

Beam: I think whoever programmed [GooglyMinotaur] at ActiveBuddy did a great job tapping into Radiohead. It was very creative. It fits with them, too—because most of the ways they operated online were ahead of the curve.


Villany: [GooglyMinotaur] was the first time we really went from being a tech-centric start-up company with technology at the center and thinking very much in terms of natural language query and providing responses. This shot us into a whole different way of thinking.

We [had] some creatives on the team who really had to think, “What did it mean to be creative about bots?” Not only in our responses, but what kind of games were we going to put in there? What kind of content were we going to leverage? How were we going to use different characters in the responses that would—interestingly enough, even though it was text-based—break the platform? How are we going to think about what was the creative aspect in messaging and tone for this bot and give it this personality? It challenged us as a start-up.


GooglyMinotaur’s launch and growth

“Well, of course I’d like to sit around and chat”

GooglyMinotaur launched on June 4, 2001, and was particularly popular on AOL Instant Messenger. Besides sharing information about the band and album, it also became known for its non sequiturs, sassy responses, and generally Radiohead-esque attitude. When it was retired in March 2002, the bot had “sent about 60 million IM messages to nearly one million different people,” according to a Medium post penned by Sloan Bechtel, Bird, and Lauren Goldberg.


Marketing copy written by Lauren Goldberg, Capitol Records, New Media Coordinator: “Like peanut butter and jelly, like ebony and ivory… For the complete Radiohead experience, you can’t have one without the other, and you’ll need both to access a secret webcast due to broadcast June 9th. Stay tuned for more clues. May the minotaur be with you…

In case you’ve been stuck on an island or whittling away in the Australian Outback, the GooglyMinotaur has now entered the world of instant messaging. Add him to your buddy list today. Also, there’s currently a 4-for-1 special on minotaur icons. Collect ‘em all and trade ‘em.”


Bird: He had a menu that he would serve up to guide you. In that menu, you could find out about the band and its members’ history, current news, upcoming tour dates. At first, it had a fun surprise element: Since no one knew what a chatbot was, the marketing was hush-hush—“Just put this in your buddy list.”

Sloan Bechtel: [GooglyMinotaur] spoke French. He asked you nicely not to swear. He cursed at you and made you apologize. He was amused by boxers and briefs. He was snarky and clever; he didn’t take any B.S. He would challenge you to a game of hangman. He had millions of natural responses—not unlike Siri today. Reporters would interview it for articles. He did not like Megadeth! He was quick to point out that no one is perfect, when users would tell him his responses were wrong.


Villany: We would program the bots, and we would get a lot of the mechanisms going, and then we would create a massive Word document transcript of the majority of the responses, so that Capitol could review those. Almost like a copy desk, if you will, for them to review, and then they could give some feedback. Maybe we did, like, two or three rounds of that and then they blessed it and then we’d release and test.

It almost speaks to the democracy of the internet, [that] even though we were deploying on the network, we had very little restriction from AOL, Yahoo, or MSN, which were our three prime networks. Googly was actually released on the three, and so Capitol was definitely very close to the marketing and the overall approval of the bot, to make sure that we weren’t encroaching on anything that was not favorable to the band.


Hoffer: The most communication we had with the band was they liked us and they sent us a guitar stool. “Thank you for doing this,” kind of thing. Capitol looked to us for guidance, they had a PR person involved. Basically, we told them what we could build and what we could do, and all they really wanted to do, to be truthful, was to get the PR. Which they did.

Bird: For us the funniest thing is that Googly, a chatbot, got physical mail from AOL, which was their big tagline back then: “You’ve got mail!”

SmarterChild explodes in popularity

“Everyone wants to be a friend”

In June 2001, SmarterChild also launched to the public—and it, too, exploded in popularity, especially via AOL Instant Messenger. In the still embryonic world of online chatting, and the wild west of bots, there was no precedent.


Hoffer: We were having trouble negotiating carriage for SmarterChild, because SmarterChild was literally crushing America Online’s service. It was never designed to deploy a 100 percent online, 24-by-7 entity. We were 5 percent of total traffic on all instant messaging—and growing. We were the first internet meme, ever. We were growing faster than Twitter ever grew. We were adding tens of thousands of new users each and every single day.

What happened was the guy that was in charge of instant messaging at America Online is a friend of mine by the name of Raul Mujica. And Raul didn’t want SmarterChild on the network at all and was shutting us down every couple of days. We were fighting with them to stay online.


What happened was Radiohead showed up. It turns out that Raul was a Radiohead fanatic. [Because of that] I was able to negotiate carriage permanently for SmarterChild on the AOL Instant Messenger network. That gave SmarterChild a reason to live. Like piggyback legislation, I snuck SmarterChild into the contract with AOL. (Mujica did not respond to an email requesting an interview.)

Guiney: I could spend the entire day thinking of new ways of answering people’s questions. We saw the raw logs of what our users were actually asking, and of course a lot of it was profanity, insults, just horrible racist [things]. And also sexual propositions. But also really interesting questions, out-of-the-blue philosophical questions.


A lot of people were lonely. There was just anything you could imagine. With that information, I tried to respond to everything that we could in ways that were funny and interesting and also that made people think that maybe they were talking to—not necessarily a person, but that they were talking to something that had a mysterious understanding of them or could actually engage in a conversation.

Kay: We kept track of everything that people said that we didn’t know how to answer. We would sort that. The ones that were popular came to the top. For example, it knows how to answer, “What is the air speed velocity of an unladen swallow?” Why? Because that was what people asked.


Hoffer: [SmarterChild] was the most popular instant messaging bot in history. People don’t like Siri. [Laughs.] They loved SmarterChild.


When users started messing with SmarterChild

“There are many things to talk about / Be constructive”

Hoffer: We have the world’s most sophisticated profanity filter.

Kay: Every conversation in the world with this new thing was the same. They’d all add it to the Buddy List and say hi. It would say, “Hi, I’m SmarterChild built by ActiveBuddy! Type home to find out what I can do.” Everybody would type home and it would give you a list of all these options, so people would pick one. They would do sports scores or the weather or something. Then every user on the planet had the same next query: They all said, “Fuck you.” Everybody! It’s unbelievable.


The guys who had poured their hearts into the application and making sure the stock quotes were good and all that stuff were really frustrated. They were saying, “What’s wrong with people?” I said, “Obviously, its not what’s wrong with the people, because if everybody is doing it, it’s what’s wrong with the product.”

Then Chris Bray went home one weekend, and in a fit of frustration he wrote the profanity handler. Here’s the thing: We had this deal with AOL, and one of the terms of the deal was we really had to be careful. We couldn’t be cursing at our customers, right? They might be 10-year-olds.


[Bray] wrote every possible [profane] expression. I mean, I don’t even know what half of this stuff meant. He came in Monday morning, and he said, “Ach, these idiot users are all cursing at it. This will take care of that.” We deployed it. And then the next person that cursed at it, it said, “Are all humans so crude?” It was that instant that our traffic took off like a rocket.

Hoffer: A friend of mine at the time, who was the very first investor in ActiveBuddy, is a guy by the name of Doug Frankel. He is the only guy still at Pixar who was a 2-D animator at Disney who made the transition to 3-D animation digitally. He is the youngest man to ever have animated Mickey Mouse.


He came to visit us after we built SmarterChild. When Doug saw our audience numbers, he said, “You are bigger than The Simpsons.” We had more audience every day talking to SmarterChild than The Simpsons had by far—we were way bigger. Then he identified the reason: Our bot had a personality. It was snarky.

And that is the reason that the big companies can’t do it properly. It’s because when you have a character with a personality, you’re not facing a technical problem maybe, you’re facing an editorial problem. And you have to be willing to piss off exactly 50 percent of your audience.


Kay: If you cursed at [SmarterChild] enough, it would demand an apology. It would say, “Look, until you apologize I’m not going to talk to you anymore.” And it would be silent. You could show up and say hi the next day, and you’d get nothing back. Three weeks later, you’d say sorry, and it would say, “Well, I am glad that we got that over.” Then it would start talking to you again.

Hoffer: Even though [Microsoft] bought our company in ’06 for $50 million, they forgot that they had this technology when they launched the Tay bot. It’s on the shelf up there, and they just derped it again. Even though they had the experience in 2006 and ’07 of how to do that, they forgot that they knew. When they launched Tay bot, they didn’t launch it with a curse-word filter, and it was immediately co-opted by end-users, and became awful.


The first thing Microsoft did when they picked up our company was they absolutely lobotomized the bot. They neutered him—they took out the personality.

Post-GooglyMinotaur growing pains

“There was nothing to fear and nothing to doubt”

As SmarterChild evolved, so did ActiveBuddy. The company began building bots for movies, magazine launches, and even Major League Baseball and the BBC. It also started exploring ways to branch out with its technology.


Villany: After GooglyMinotaur, we really started to become more sophisticated in how we were thinking about things.

Hoffer: Radiohead was like a gift from god. They showed up, and poof, we found ourselves doing not only Radiohead, but dozens of bots for artists—dozens. Madonna. Artists that came and went, artists that you’ve never heard of. We had bots for all of them. Like, [Grateful Dead’s] Phil Lesh, we did a bot. It became a big deal for the business.


Villany: [Capitol was an] awesome client. Through that relationship, we started doing things for New Line Cinema. I built a bot for Austin Powers, working actually with Mike Myers and the team at New Line. We also did the bots for Lord Of The Rings. That franchise had just been released. And then SmarterChild was just this playground for all this different stuff that we were going to test. We would have so many Easter eggs in there. But Smarter just had such a mass audience. And that thing was a total mistake.

Kay: That’s one of the businesses we were in, was these marketing bots. It never really paid the bills, so we also went into customer support. We were doing customer support for Comcast, for example. If you went to their support site, you’d see that you can call us on the phone, you can send us email, or you can talk to our chat bot. That was us.


That was a great customer, and they actually paid a decent amount of money for that. The problem is that they don’t release numbers. They would not act as a marquee customer for us. That made it difficult for us to then move on to selling the next one.

Villany: We were creating the bots, but we were also creating the software for BuddyScript, SDK, so that we could get into something that was a little more user-friendly. Eventually, we wanted to license our software to other marketers or companies so that they would be enabled to build bots. We loved building bots and they were fun, but we as a company felt we would have much more success and reach if we allowed other people to build bots.


ActiveBuddy deals with conflicting priorities

“You don’t live in a business world”

Hoffer: We were at the top of our game, we had plenty of money in the bank. SmarterChild was growing. The only real pressure was, “How do you mitigate this idea of who we’re going to be? Are we going to be a company that builds bots on behalf of other companies?” There was managerial pressure at the top of the company. There was a real rift in terms of who we wanted to be. Half of the management want to not do that, and half did.


We had really done a great job from an engineering point of view. The quality of the language processing and tools and technologies that we built back then still stacks up as good or better than even Siri, which is Apple’s largest software division. As certainly as good as Alexa is today. We had world-class technology 15 years ago. We had plenty of talent.

The issue was, “Do we do this? And if we do this, what’s our business model?” I was on the side of the fence that was like, “Forget the business model, guys, what we’re gonna do is drive the growth. And eventually, once we have a few hundred million people using it, then we’ll come up with a business model.” And that kind of thinking was not in vogue at the time, because we had just had the 2001 internet crash. The idea of attracting distribution first and then building a business model later had fallen out of favor.


Kay: The whole relationship between AOL and SmarterChild was tortured. Again, it speaks to a past where for a while there was an eyeball play. Up until the dot-com crash, everybody was saying, “Oh, it doesn’t matter if you have a business model. Eyeballs is what counts.” Then the dot-com crash came and now everybody says, “You’ve got to be making money.”

Our first deal with AOL was a year long, where we launched for free and we’d just see how it goes. Everybody said, “You gotta start making money with this,” including AOL. They wanted their money. It wasn’t enough that they had this hot property. And they were threatening to shut it down, and so on. We actually did take SmarterChild down because AOL wouldn’t sign an extension. We would be at their whim.


Then we brought SmarterChild back as a subscription product. I don’t think I was excited about that either. They were going to charge $10 a year. The engineering team put together the ability to do subscriptions, and I think maybe 20 subscriptions came in. They finally relented and gave the money back. AOL agreed to put SmarterChild back up without subscriptions.

Nobody knew what to do with a property like this. How do you make money at it? It was just media, and yet the whole internet’s been swinging back and forth between “content is king” versus “you need to have a real business model.” We were content at the time. Were we content or were we technology? That was also an interesting question.


We had these media properties with lots of traffic and no revenues associated with them. And we had this technology which was pretty awesome. You could internally account for how much revenue that was toward the technology—not much. And then we had things like that Comcast deal, which were reasonable-sized checks if we could have replicated it. There’s an argument to be a technology company.

How ActiveBuddy was eventually bought by Microsoft

“We are the dollars & cents and the pounds and pence”

Kay built an instant-messaging proxy server that would keep internal chatting secure, which he considered taking to market. Also on the table: splitting ActiveBuddy in two. Neither ended up happening.


Kay: I had a proposal in front of the board of directors, which they didn’t accept, which was to split the company in two, a technology company and a media company. SmarterChild would go off with Robert and do cool stuff—potentially. It was still a risk, because how do you make money at that? But at least it would have clarified focuses. And we would be a technology company, and we would license the BuddyScript platform to the media company.

GooglyMinotaur was the one who put us into this whole space where it even made that possible. SmarterChild alone was interesting, but then the fact that GooglyMinotaur caused us to create this whole splinter off of dozens of different product pointed the way to a possible business. Keep in mind it was dot-com times, where they were selling pet food over the internet. Those were big valuations without any realistic reasons for it, and then it all came crashing down.


We ended up not splitting the company and continuing to do a few of the GooglyMinotaur-like bots, but trying to focus on enterprise and developed a lot of technology that way. In the end, Microsoft came in and had us focus in a very different direction, which was to create Microsoft products within MSN messenger. They liked it enough to buy the company [in 2006].

Hoffer: We could do it again today, if we felt like it. We knew what we wanted to do was to grow that thing. Chris Bray said to me the other night, “Can you imagine if we had continued to work like you wanted to on SmarterChild for the last 15 years, how big a service that would be today?”


Siri doesn’t do even the basic natural language process. If you ask Siri, “What’s the weather in Miami?,” it’ll tell you the weather in Miami. But if you then say, “Siri, can you give me directions from there to Fort Lauderdale?,” it won’t know what “there” is. It forgets from one sentence to the next. But SmarterChild didn’t. We had a profile that lasted forever, because the mission was to be your best friend on the internet. We applied that technology to each bot we built.

Kay: There’s no doubt that if we made it through to the age of smartphones, it would have been awesome. But the company was long gone before the first iPhone shipped. That’s how crazy ahead of time we were.


Hoffer: Back then, and today, the big problem in instant messaging is that nobody has figured out how to monetize it. There’s never been any income generated, revenue generated, to companies. AOL never made a dime from Instant Messenger. They all have these great user bases, and the only people that ever make money is, like, when ICQ was sold to AOL. [Founder] Yossi Vardi made [millions of] dollars, but he never made a dime of revenue.

The idea of bots—which has now come around again—is a major thing. Fifteen years later, oh my god! Every major company—IBM, Microsoft, Facebook, Google—they’ve all announced bot initiatives to monetize the instant messenger window. They don’t have a business plan on how that’s going to happen.


Kay: Here we had this great technology, and the question was, “What’s our business? How do we make money at it?” [Laughs.] I don’t think we ever really answered that question.

The legacy of GooglyMinotaur and SmarterChild

“Think about the good times and never look back”

Hoffer: Radiohead is really brilliant. Even their most recent PR stunt, where they removed all of their junk from the internet to launch a new thing, was brilliant. They’re always sort of one step ahead. They saw there was a trend in instant message, and that’s where people were, and they went there.


Beam: It was always fun to be a fan back then—and even now, though they are less active online—because they were connecting with their fans in completely atypical ways. I’m a professor in communication [studies] at Kent State now. I primarily study online news, and I’m always fascinated by the ways new technologies and innovators bypass traditional gatekeepers to provide alternate routes to connection. Radiohead did that very well.

Villany: [We were] taking this idea and this platform that we had for bots, and we were thinking about it in a very tech-centric way, and giving it a very creative, user-centered, fan audience approach. Thinking in that way really brought us back to, “How do we think about all of that from a technology point of view?” That’s when we started to officially think about, “Could bots be self-learning? How would bots develop their personalities on their own? What kind of libraries should we be building?” We certainly don’t want to build all these patterns for jargon and slang every time we build a bot. [GooglyMinotaur] was the turning point for us as a start-up. Very much so.


Guiney: There was surprise and delight people had when they realized the thing they were chatting with wasn’t just this really dumb, lazy marketing ploy—where all it would do is give you the date that the album was being released, no matter what you said, and hammer you with a really dreary marketing statement.


When people discovered that you could really talk to it and get all kinds of answers, and it wasn’t necessarily going to be fixated on marketing something to you, people instinctively felt good about that and felt it was worth exploring and having repeated interactions with. Which is, of course, something that’s very valuable on the internet in general. People came back again and again, and they told their friends about it, and before you knew it, tons of people were using it.

Hoffer: There’s something uniquely intimate about the instant messaging phenomenon that just keeps being that way. And they were intimate at a level that you don’t get on the telephone or face to face. It’s this own unique medium. What we were able to do was to conversationally capture some of the magic of instant messaging that people adore and love, within the context and confines of a bot.


When you think about the future, everyone has a vision. But we all share that vision from when we grew up. It’s sort of like, “What’s your future going to be? Is it going to be the Dr. Seuss future? Is it going to be the Stanley Kubrick future? Is it going to be the Hanna-Barbera future?” In the Hanna-Barbera future, we’re all flying around like The Jetsons with flying cars, which fold up into suitcases. That future is being realized by Elon Musk—he is going in that direction, with electric cars, and flying rockets. We were chasing the Star Trek future. We were chasing the talking computer that everyone knows is going to happen. The question is when.

Some time along the line, there’s going to be artificial intelligence, and it’s happening right now. We were 15 years ahead of that curve, but it’s not like we started that curve. I came up with the idea for SmarterChild sitting in a movie theater watching Robin Williams depict a character called the Bicentennial Man. And that’s what I wanted to build.