Sramana Mitra: Have you seen a use case that you could talk about where such conversational interfaces are being applied to a business problem?
Ken Fang: One of the areas that we are working on with USCS is a pilot project where we built an Alexa interface for people to report. They have a web form where you can go ahead and enter in earthquake information. It’ll ask for your name, address, when you felt the earthquake, and how you strong you believed it to be. What we did was we built an Alexa app where you can do that through Alexa.
In that interface, the key part is I don’t have to tell you who I am. You already know who I am. I don’t have to tell you where I am. You know where I am. In that way, USCS can get that kind of information more rapidly. Currently, they actually have sensors across the nation where they’re trying to do that. Here you can have millions of consumer devices and people can go ahead and report it through conversational information type of thing. That can move into much larger things like early notifications.
Sramana Mitra: Switching the line of questioning a bit, a lot of what we are discussing are one-off uses cases. Where do you see opportunities for productizing some of these trends and coming up with business ideas that would be the foundation of new companies?
Ken Fang: I think that there’s a lot of potential out there. One of the areas we’re seeing a lot of interest is in chatbots. On a website, something will popup and you’ll have your chat conversation with it. It’s an automated robot that’s trying to answer your question. A lot of those things are actually handcrafted where there’s a little bit of if-then logic. A lot of them are also using supervised and unsupervised learning techniques to build out that kind of knowledge base.
The idea would be that you have your conversation with a robot. At some point in time, the human is going to be talking about things that the bot has never heard of. It just can’t answer it. That has to trigger a human coming into play to take over that conversation and replace the automated system. Then that interaction from there should actually feed the system such that the robot can then go ahead and answer. It’s a very complex problem to do in a specific setting.
Sramana Mitra: Assisted selling in specific domains with constrained domain vocabulary is a good idea. I think people are working on stuff like that.
Ken Fang: People are definitely working on it. What we would say is the idea would be for a business to say, “Is there a SaaS that I can add to my site, but I want to be able to do that without having programmers programming it for me?”
Sramana Mitra: It’s a big decision to add a chatbot if you also have to staff that chatbot.
Ken Fang: Yes. That’s exactly true. If I don’t have the customer service staff to take over, is there an elegant way for that chatbot to say that it can’t answer.
Sramana Mitra: Excellent. Thank you for your time.