Amazon Unveils Alexa Prize Socialbot Grand Challenge 5 Competitors
Amazon has announced the teams jockeying in the Alexa Prize Socialbot Grand Challenge 5. The $1 million prize for Alexa bots that can carry out complex conversations on different subjects and mimic how humans talk was first announced in June, with the finalists scheduled to be revealed in November, but it seems to have been delayed.
The nine competitors include five returning from previous contests, including the second and third-place winners of the last competition. This year’s returning competitors include the Czech Technical University, University of California, Santa Cruz, Stanford University, Universidad Politécnica de Madrid, and Carnegie Mellon University. The new names include the Stevens Institute of Technology, the University of California, Santa Barbara, the University of Illinois, Urbana-Champaign, and Virginia Tech.
“Creating a socially adept AI is a hard problem. This is because human-like social conversation is remarkably delicate and complex, and the open domain nature of the SocialBot dialogues makes it extremely challenging,” Alexa AI senior principal scientist Alexa Prize head Reza Ghanadan said. “You need to provide relevant and deep responses to a wide range of topics users may ask, maintain a natural and coherent exchange throughout a potentially long conversation, and accurately interpret the intent of the user by correctly picking up on names, topics, places and products while taking into account the context of each conversation turn. You also need to make the interactions with users lively and engaging, which is challenging given the diversity of topics and users interacting with Alexa.”
The Alexa Prize began in 2016, with Team Alquist from Czech Technical University winning the most recent iteration and its $500,000 prize. CTU was preceded by earlier winners Emory University and its Emora socialbot, the University of California, Davis’ Gunrock socialbot and the University of Washington’s Sounding Board program. This year’s winning team will earn $250,000, with $50,000 and $25,000 prizes for second and third place, respectively. All of the teams chosen to compete will get a $250,000 research grant, including Alexa devices, free Amazon Web Services (AWS) and other tools, and a connection to Alexa developers for support.
Amazon added a new facet to the socialbot competition this year, adding a separate prize for scientific invention and innovation. The same socialbots will compete for both prizes, with a chance of winning either or both. The teams will send in short summaries of scientific experiments and results at the end of the competition for review by Amazon scientists. After the socialbot finals, chosen teams will present their research to a Science and Innovation Review Panel before answering questions posed by Alexa AI scientists. As with the socialbot competition, the top science project team gets $250,000, and second and third place receive $50,000 and $25,000.
“We have learned that success requires researchers to create generalizable AI techniques and incorporate knowledge in appropriate and engaging ways,” Ghanadan said. “It also involves addressing open research problems in natural language understanding and multimodal language processing, contextual understanding, natural response generation, empathy, and commonsense reasoning, to understanding social norms, and dialogue management.”
The teams are competing to outperform each other, but winning the Grand Challenge requires the socialbot to earn a composite user rating of at least 4.0 out of 5 and sustain a minimum of 20 minutes of conversation with the judges where at least two-thirds of that conversation matches a human’s pace and coherence. No team so far has completed the Grand Challenge and its $1 million prize pot, though they have edged closer with each new competition.
The Alexa team updated Amazon’s Conversational Bot Toolkit, Cobot, in tandem with the new scientific competition. CoBot is built for scientific research with a streamlined setup and infrastructure combining the Alexa Skills Kit, Amazon Web Services’ cloud-computing tools, and the natural-language-processing models of the Alexa Prize Toolkit Service. The augmented version of CoBot includes several upgrades and offers more models, a modular setup, and plug-and-play software. You can try out previously submitted socialbots, by saying, “Alexa, let’s chat” to an Alexa-enabled device.
“Prize competitions provide an agile experimentation framework for researchers and students, encouraging them to explore transformational ideas at the boundaries of what is achievable,” Ghanadan said. “We have developed the CoBot platform and tools to lower the barriers to AI innovation for academic research community and students interested in conversational AI assistants. These tools allow students to quickly deploy their solutions in the real world with Alexa, then observe, evaluate and enhance their research results using feedback from Alexa customers.”