Amazon has launched a $2.5m competition for teams of students to create an artificial intelligence charming enough that people can chat to it about celebrity gossip and sports.
The Alexa Prize is named after the service that powers Amazon's voice-activated smart speaker, the Amazon Echo. The goal of the inaugural competition is to build a 'socialbot' for Alexa that will converse with people about popular topics and news events.
The team with the highest-performing socialbot will win a $500,000 prize, while there's a $1m for the team's university if their socialbot can chat "coherently and engagingly" with humans for 20 minutes. Up to ten teams of students will also win a $100,000 award, as well as Alexa-powered devices, free AWS services, and support from the Alexa team.
"Conversing for 20 minutes is difficult for most humans and an extraordinarily ambitious challenge for bots that are learning to converse like us," said Dan Jurafsky, professor of computer science at Stanford University.
Students will build their socialbots using the Alexa Skills Kit, which is used by developers to build voice-powered applications that can be accessed via the Alexa service.
Teams will be able to access conversational topic categories and digital content from sources, including The Washington Post (owned by Amazon CEO Jeff Bezos), which is making its news feed and comments available to the students for non-commercial use.
Let's chat about...
As part of the research and judging process, Alexa users will be able to converse with the socialbots on popular topics by saying: "Alexa, let's chat about..." with potential subjects including baseball playoffs, celebrity gossip, and scientific breakthroughs.
Alexa users will be able to give feedback on their experience of the conversation to help the students tweak their socialbot algorithms. This feedback from Alexa users will also be used to help choose the bots to advance to the live judging phase.
Amazon said the prize will help to advance several areas of conversational AI, including knowledge acquisition, natural language understanding, natural language generation, context modelling, commonsense reasoning, and dialogue planning.
The idea is to create 'socialbots' that can acquire knowledge and opinions from the web, and express them in context just as a human would in everyday conversations said Rohit Prasad, head scientist at Amazon Alexa.
"A socialbot that can converse coherently for 20 minutes is unprecedented and at least five times more advanced than state-of-the-art conversational AI," he noted.
"Human capacity for language is an instinct, but it's something that must be taught to machines," said Steven Pinker of Harvard University. "Everyday conversations that require context and understanding of the world come naturally to humans." Machines can't do that, which makes conversation like this much harder for them.
More on the Amazon Echo
- How we learned to talk to computers, and how they learned to answer back
- Amazon Echo: The four hard problems Amazon had to solve to make it work
- Amazon Echo voice-controlled smart speaker comes to UK and Germany
- Amazon intros cheaper Echo Dot, priced at $49.99
- Amazon Echo review: A perfect 10
- CNET: Amazon Echo (and Alexa) arrive in Europe, and Echo comes in white now too
- TechRepublic: Amazon Alexa will now talk to GE's connected appliances in smart home push