We are EVE, a team of researchers from the Brigham Young University Perception, Control, and Cognition Laboratory, and we’ve been accepted to the Amazon Alexa Prize Challenge! As one of eight university teams tasked with the challenge to create a conversational AI system for Amazon’s Alexa, we are building a fully functioning socialbot that you can hold a real conversation with.
EVE stands for Emotive Adversarial Ensembles, a combination of machine learning terms that describe her overall structure. Eve attempts to express and interpret emotion using a group of response generators, called an ensemble, that help her decide what to say. The word adversarial does not describe her personality, but rather is a reference to the way her component neural networks will be trained.
As of this post, Eve is still in her infancy. She was designed as part of the 2018 Alexa Prize, a 3.5 million dollar competition sponsored by Amazon, Inc. We trust that over time she will grow into, and perhaps beyond, her name.
Why do we need conversational AI? The Alexa you know today functions primarily as a taskbot: A system with preprogrammed responses and buttons you can push with your voice. “Alexa, set an alarm for 7:00am tomorrow morning.” “Alexa, play NPR news.” “Alexa, what’s the weather today?” While these tasks have the ability to simplify and improve our lives, we’d like to take Alexa’s abilities even farther.
“Alexa, why is it important for me to get an education?” “Alexa, how do I cheer up my sister?” “Alexa, it’s my mom’s birthday. Could you pull up her favorite recipe?”
These questions go beyond the capabilities of a normal chatbot. They require the ability to capture the intent of the user and improvise accordingly. They require a long-term memory that recognizes users and associates memories and personalities with them, so long as the user agrees to it. They require AI. We are leveraging these ideas to create the next generation of voice interfaces: no longer preprogrammed taskbots, but intelligent socialbots.
This is, of course, an incredibly difficult task. Researchers have been working on it for decades, and the industry has made unprecedented strides in recent years. We hope to build on the accomplishments of those who came before us, work in collaboration with those alongside us, and push the boundaries of AI to new levels.
We are not alone in this effort. A crack team of Amazon specialists is helping each of the university teams to interface with Alexa’s back end systems, making it possible for Eve to converse with thousands of people at a time without losing track of each conversation.
Over the next few months, we invite you to share our journey as we strive to perfect EVE. As Ashwin Ram, Sr. Manager of AI Science and Alexa Prize Lead once said, “Conversation is the hardest AI problem I know, and we’re looking forward to solving this together with you.” As EVE develops you’ll have the opportunity to chat with her on your own Amazon Alexa devices. And right now, in the early phases, we’d like your help in order to accelerate the development of AI! Let us know what you hope for, what you’re afraid of, where you see this going, and what you expect us to do about it.
The future begins with us. All of us. Which means it also begins with you.