Apr 17, 2016

Bots, Messaging and SmarterChild

If I gave you my love, I tell you what I'd do
I'd expect a whole lot of love out of you
- Al Green

Smart people believe that messenger type interfaces and associated “bots” will become a primary way people access new services going forward. We at USV certainly find truth in that, see both Kik and Koko and Libov’s thoughts here.

There are a bunch of reasons why this could be the case. First, the primacy of messaging - in all its various forms - appears to be growing as fast as ever. The conversational nature of messaging and chat also feels natural. Finally, the reduced load on a user from having to switch contexts to do different things appears as a major benefit to the different interface.

I’ve seen this before, as one of my first investments ever in 2000 was leading the first financing in SmarterChild, where that product and company faced many of these same questions. What did they see and do?

  • Given that it was 2000-2001, SmarterChild launched on the then dominant (only?) messaging platform - AOL’s instant messenger (AIM). In no way did we think this was a wholly original idea - it was inspired by Eliza, Alice, and many others. Instead, the implementation and technology were more unique.

  • SmarterChild was not designed to live as simply one of many names on your contact list delivering services. The goal was for it to be your primary portal into the world - to be the one “bot” that ruled them all.

  • As a result, the company spent an inordinate amount of time upfront building personality libraries. The idea was that since SmarterChild was interjected into a place that was exclusively where personal conversations occurred with friends, the service (or bot) had to act like your friend in addition to being a utility service. Otherwise it would too interruptive or, worse, easy to ignore.  

  • This required SmarterChild to be able to respond to not only informational queries but also, well, sex and profanity. Chris Bray, one of the original developers, was once asked how people were using SmarterChild. He replied:

  • The implicit assumption here was that the utility services would take time to perfect, so you had to give people a reason to stick around while that happened. A domain library that responded to profanity was thus a necessity, along with weather, sports scores, stock quotes, news, dictionary, yellow pages and search.

  • Making SmarterChild appear as a friend (albeit a strange one) would become the primary way people would then tell their other friends about it. And alot of people told their friends. At the peak, SmarterChild had tens and tens of millions of followers on the IM platforms. I am pretty sure this was - at the time at least - the most followed “user” in the history of AIM. People posted thousands of screenshots of their interactions with and feelings about SmarterChild:

Screen Shot 2016-04-16 at 3.38.39 PM.png

Screen Shot 2016-04-16 at 6.38.55 PM.png

  • The other area that the company invested heavily into was response time. A first principle was that unless the bot could answer in under 10 milliseconds to each and every request it would be a failure. Speed, speed, speed.

SmarterChild worked well until it didn’t work well. For one, at some early point it needed explicit agreements with the platform owners because it quickly ran into rate limits. The volumes were also so massive it needed technical cooperation with them. And, finally, there was the constant tug back and forth with the platforms. 

ActiveBuddy, the parent corporation of SmarterChild, lived virtually every day with both the excitement of massively increasing user numbers and fear that they would be shut down because they neither owned nor had any control over their distribution. Ultimately, that tension was too much to manage against.

blog comments powered by Disqus