Changing the Game: Creating Intelligent, Conversational Interfaces
My dad got me my first computer for Christmas when I was 11 years old. It was a Texas Instruments 99-4a. He must have gotten it second hand because it had no box or any form of storage like a tape drive.
But that didn’t stop me. I had been obsessed with the movie War Games, so my first program on the TI was designed to replicate the computer in the movie which uttered the famous line “shall we play a game.”
My program gave static responses to a huge variety of programmed questions, and mostly followed the script from the movie. In other words, it was not very good and was easily tricked, but it still blew the minds of friends and family alike. When I got them to type questions and carry on a simple dialog with the T1, they thought I’d coded the computer to speak to them naturally. Most so-called “conversational” interfaces at the time used tremendous amounts of statically coded question/response approaches. And like my program, they simply weren’t very good.
Thirty years later, computers are finally able to carry on a real conversation.
My kids and I play a game in the car, which is to try to trick Siri with silly questions. I can always get them rollicking with laughter by throwing crazy accents at Siri and carrying on escalating silly conversations. “Siri, why don’t you love me?”, “How dare you speak to me in that tone of voice”, or “What kind of fool do you take me for?”. Siri often comes up with witty responses and occasionally surprises us.
But other than Siri and a small handful of others (Google Home, Cortana, etc), there are surprisingly few convincing conversational bots.
The rise of messaging apps, and the conversational bots which have followed, have given us a tremendous number of bots which are impressively bad at natural language conversation. They do OK with canned responses, but try to have a semi-unstructured conversation to get them to do something and the experience tends to be poor, frustrating and decidedly non-human.
Why is it that computers can think like us and program like us – but they can’t communicate like us?
At Cisco, we have been at the forefront of the messaging revolution with Cisco Spark, and have been seeing a boom in conversational bots for all kinds of purposes. We have also seen immense interest from our customer care customers, who see bots (rightly so) as the evolution of multiple choice interactive voice response (IVR) systems.
We realized that to really enable our customers to have more natural, conversational interactions in our enterprise collaboration tools, we’d have to do more for them, and do more of the heavy lifting.
So that’s why today, we are announcing the intent to acquire a company called MindMeld. MindMeld realized this same problem and has been at the forefront of Artificial Intelligence (AI) and Machine Learning (ML) research on creating lifelike, convincing conversational interfaces. Creating a high-quality conversational interface requires six distinct types of ML, including Natural Language Processing, Question Answering, Dialog Management and so on. MindMeld has written the book on these technologies and has built the world’s best conversational user interface platform.
This is the next step into a comprehensive AI-powered collaboration solution for Cisco. We are already leveraging AI/ML in new and exciting ways in existing products, from our SpeakerTrack to our VoiceTrack technologies. Bringing the MindMeld team to Cisco is a giant leap forward in helping our customers experience the next generation of interactive, conversational interfaces.
We’ve come a long way since the days of my silly first program on the TI 99-4a, but we’re really still at the beginning of the beginning of human-computer interfaces. Conversational interfaces are the next major step forward, and we’re thrilled to have MindMeld and their CEO, Tim Tuttle, join us to usher in the next era of AI-powered collaboration technology. What future do you see for AI in the enterprise? Let me know @rowantrollope.Tags: