Onstage at the launch of Amazon’s Alexa Prize, a multimillion-dollar competition to build AI that can chat like a human, the winners of last year’s challenge delivered a friendly warning to 2018’s hopefuls: your bot will mess up, it will say something offensive, and it will be taken offline. Elizabeth Clark, a member of last year’s champion Sounding Board team from the University of Washington, was onstage with her fellow researchers to share what they’d learned from their experience.

What stuck out, she said, were the bloopers. This sort of misstep perfectly encapsulates the challenges of the Alexa Prize, a competition that will help shape the future of voice-based computing for years to come.

On the face of it, Amazon isn’t asking much: just create a chatbot using Alexa that can talk to a human for 20 minutes without messing up, and you get a $1.5 million prize (with $2 million in other grants and prizes). But as Clark’s anecdote illustrates, this is still beyond the capabilities of current technology.

There’s just so much that computers don’t know about the world, and there’s no easy way for us to teach them. “Don’t ruin Christmas for small children” isn’t a lesson that translates easily into code.

That’s why Rohit Prasad, chief scientist for machine learning at Alexa, compares the prize to the DARPA Grand Challenge, a series of competitions that was held by the US military agency to build self-driving cars in the mid-2000s. Early entrants failed to even finish the course, but the million-dollar rewards on offer galvanized research. Read more from theverge.com…

thumbnail courtesy of theverge.com