CMU Team Aims To Create Chatty Software You Might Actually Want To Talk To

Feb 20, 2018

A team of students from Carnegie Mellon University was one of eight squads world-wide selected earlier this month by Amazon to compete in its Alexa Prize challenge.

The goal is to develop "socialbot" artificial intelligence software that can hold engaging conversations with people on a wide range of topics, like sports and current events. 

Many "bot" programs designed to interact with humans, like the interactive phone systems used by companies such as banks, are limited to rigid, narrow conversations on one topic. Others, like Apple's personal assistant Siri, can answer a wide range of questions, but can't really have an in-depth conversation.

"You ask for something like 'What time is it?' and it gives you the answer and that's the end of the conversation," said Alex Rudnicky, a professor in the Language Technologies Institute at CMU and the mentor of the university's Alexa Prize team, Tartan.

Rudnicky said their challenge is to develop a program with more sophisticated conversation skills.

"[There would be] no particular goal to the conversation. It's really just a social interaction... something you might have with somebody at a party or a bus stop," said Rudnicky. 

Each team in the challenge will receive $250,000 to develop its socialbot technology. Over the summer, finalists will be selected to have their socialbots compete for a $500,000 grand prize in November in front of a panel of judges. 

If the winning team's socialbot is able to sustain a conversation for 20 minutes that rates above a certain point on the judge's evaluation scale, the team's university will receive a $1 million research grant.