Chatbots are an Interface, Not Your Best Friend

Chatbots seem to be the rage these days. Many entrepreneurs are pursuing them in an assumption that they are part of the “next big wave” of technology companies. First there was the PC, then the smartphone, now the chatbot. But the general approach has disappointed people greatly, with many just being lame to use. It becomes obvious after a few words that I’m not talking to a human being. So unless the chatbot is doing some procedural in processing, it’s doubtful that the interface is useful.

Except, that’s just the way people are currently building them.

See, chatbots are terrible as conversational UIs. I dislike talking to them whenever I have to and always prefer a human being. They are generally so terrible that many startups like Fin. and Facebook’s messenger use humans behind the scenes to try and cover up the uncanny valley effect.

But this is under the assumption that chatbots can be used like regular human beings. There is a saying at Y Combinator that a startup should require only one miracle to work. A company like Airbnb had to prove that people would use the product, not that the underlying technology would work. [1] Chatbots companies rely on achieving this miracle and then other miracles as well.

What would be far more useful is to think of chatbots as natural language parsers, not as virtual assistants. The Google search engine is the closest current example I know of as you can ask questions that are naturally parsed language . Thus I can type in “action movie where a german terrorist tries to destroy a building from the 1980s” and get Die Hard as a result. Ok, it’s not the first result, but it’s still really impressive.

If Google can do it, the technology is therefore proven. That’s not to say it wouldn’t be hard, but it is to say that it’s relatively solved already.

So what I think startups and technology companies should focus on is not conversational AI, which is much too hard, but instead focus on natural language parsing of commands. That is not only a doable miracle, it’s also one that is extremely useful to solve and thus becomes one startups can possibly seek to do.


[1]: Which is not to say it’s not hard, just that Airbnb’s technology is not what it’s selling. It’s simply part of the machinery required to sell that product. Think of the meat patty machines at McDonald’s: nobody would deny that they are a technical achievement, but McDonald’s uses them to sell a benefit to the consumer.

Like what you read? Give Vincent Alexander Saulys a round of applause.

From a quick cheer to a standing ovation, clap to show how much you enjoyed this story.