Skip to main content

“Alexa, let’s chat!” With those three words, Alexa customers will soon be able to converse with one of 10 socialbots selected to participate in the Alexa Prize Socialbot Grand Challenge 3 — a multimillion-dollar competition among colleges to build the best socialbot. While this is the competition’s third year, it is the first time UCI has applied. In June, the UCI team learned it had been selected to participate and would be receiving $250,000 in research funding as well as Alexa-enabled devices, free Amazon Web Services (AWS) to support development efforts, and access to other tools, data and Alexa team support.

“When I first heard about the Alexa Prize last December, I immediately thought that UCI would be a perfect school,” says William Schallock, a third-year student double-majoring in computer science and physics. So he founded Alexa @ UCI, a campus club dedicated to organizing a competition team. “I view the Alexa Prize as an opportunity to work on the cutting edge of AI research and to develop something unique.”

Building a Multidisciplinary Team to Advance AI
Schallock, along with help from mathematics major Claire Urata, is now leading a team of 14 students advised by Assistant Professor of Computer Science Sameer Singh. The mix of undergraduate and graduate students come from not only the Donald Bren School of Information and Computer Sciences (ICS) but also the Samueli School of Engineering, the School of Physical Sciences, and the School of Social Sciences.

“Tackling many of today’s largest problems requires collaboration between individuals from a wide array of fields and with a wide array of specialties,” says Schallock. “Creating an engaging socialbot is no exception. I think we can leverage the fact that our teammates all come from different backgrounds to create something that is just as unique as our team.”

The ZotBot team: (from left, back row) Steven Hu, Daniel Agress, Ilene Do, Michelle Liang, Xuan Liu, (front row) William Schallock, Claire Urata, Sameer Singh, Yoshi Matsuburu and Dheeru Dua. (Team members not pictured: Brian Caulfield, Yao Du, Ananth Gottumukkala, Arseny Moskvichev and Moeez Qamar.)

The team is currently working on the socialbot, named ZotBot, and hopes to design it such that it makes personal connections with its users. The goal is to converse coherently and engagingly with a human for 20 minutes on a range of topics — from current events and politics to sports, technology and fashion — and receive a customer rating of at least 4 (on a scale of 1-5).

The competition aims to advance conversational AI through research into natural-language understanding and generation, context modeling, dialog management, commonsense reasoning, and knowledge acquisition. “The Alexa Prize challenge provides a valuable opportunity for students not only to familiarize themselves with research in conversational agents and large-scale NLP systems, but it is also often their first experience dealing with complex software engineering pipelines running on a massive scale that is directly accessible by hundreds of thousands of public users,” explains Singh, whose involvement in AI research includes a partnership between ICS and the Allen Institute for Artificial Intelligence (AI2). “In this regard, we’re lucky Amazon is providing this level of access to their products, software tools and consumer base.”

How You Can Help
The team welcomes feedback from fellow Anteaters! You can sign up to test ZotBot by filling out this form. You can also keep an eye out for their Alexa @ UCI booth in the Engineering Quad on campus.

Starting on Dec. 4, anyone with an Alexa device can simply say, “Alexa, let’s chat,” to open up the Alexa Prize Skill and start chatting with one of the socialbots being developed for this year’s competition. Users can then rate the conversation and provide feedback. “These ratings will be used to decide which teams advance into the later rounds of the competition,” says Schallock. “However, due to the fact that the ratings are taken into account, the bots will be anonymized, so you could be talking to ZotBot or any of the other nine socialbots.”

The Grand Challenge
The semifinals will take place in March and April 2020, and the finals will occur in May, with winners being announced in June. The first-place team will receive $500,000, while the second- and third-place teams will receive $100,000 and $50,000, respectively. A $1 million research grant will be awarded to the winning team’s university if the socialbot achieves the grand challenge of conversing coherently and engagingly with humans for 20 minutes with a 4.0 or higher rating.

Win or lose, Schallock points out that working on a project like this gives students real-world experience they can’t get in a classroom. “The Alexa Prize team from Amazon has been quite helpful,” he says. “It is an overall great experience to work side by side with some of the best engineers/computer researchers in the world.”

He also recognizes the larger significance of this work. “I think [conversational AI] is the next frontier in the world of computer science,” he says. “There is also potential here to develop intuitive systems that can be seamlessly integrated into our everyday lives to improve our quality of life.”

Shani Murray