Amazon launches Alexa Prize TaskBot Challenge 2
University teams are invited to compete to build multimodal conversational agents that assist customers in completing tasks requiring multiple steps.
Following the success of the inaugural competition in 2021, Amazon is officially launching the Alexa Prize TaskBot Challenge 2. Starting today, university teams across the globe can apply to compete in developing multimodal conversational agents that assist customers in completing tasks requiring multiple steps and decisions. The first-place team will take home a prize of $500,000.
The TaskBot Challenge 2, which will begin in January 2023, addresses one of the hardest problems in conversational AI — to create next-generation conversational AI experiences that delight customers by addressing their changing needs as they complete complex tasks. It builds upon the Alexa Prize’s foundation of providing universities a unique opportunity to test cutting edge machine learning models with actual customers at scale.
“Alexa Prize competitions provide an agile experimentation framework in which industry and academia can collaborate to advance the science behind AI assistants, and to improve the dynamics of human-AI interactions leveraging real-time feedback from Alexa users,” said Reza Ghanadan, senior principal research scientist with Alexa AI. “The TaskBot Challenge introduces a unique opportunity for conversational AI researchers interested in making AI useful and beneficial to people to rapidly experiment and evaluate their innovations from concept to research to user-validation at scale using Alexa devices online.”
Teams are challenged to build taskbots that assist customers in tasks requiring multiple steps — such as removing a stain or cooking a turkey — while also adapting to the items and tools the customer has on hand. If, for example, a stain removal tip called for white vinegar and the customer doesn’t have any, the taskbot should adapt to offer an alternate approach based on the items the customer does have.
In last year’s competition, 10 university teams from three continents competed to build digital task assistants over the course of a year. In June of 2022, three of those teams — the University of Glasgow in Scotland, NOVA School of Science and Technology in Portugal, and Ohio State University — were announced as the top performers.
The Taskbot Challenge is a unique opportunity for student researchers to experience how their ideas and techniques can be deployed in the real world by interacting with Alexa customers
The top performers earned prize money to be divided among the student team members: $500,000 for first place, $100,000 for second, and $50,000 for third. In addition, all 10 teams participating in the competition published research papers detailing their work.
“The Taskbot Challenge is a unique opportunity for student researchers to experience how their ideas and techniques can be deployed in the real world by interacting with Alexa customers,” said Eugene Agichtein, a computer science professor at Emory University who served as the faculty advisor for two of Emory’s Alexa Prize teams before joining Amazon as an Amazon Scholar, helping to drive the definition and organization of the TaskBot Challenge. “Most students are never exposed to the demands and constraints of incorporating their ideas and innovations into a large production system. The TaskBot Challenge drives the students to see the bigger picture of how their specific research ideas fit together to impact an overall customer experience, often influencing their future research career.”
The TaskBot Challenge also incorporates screen-based interactions into the conversational experience, so in addition to verbal instructions, customers with Echo screen devices or a FireTV may be presented with step-by-step instructions, images, or diagrams that enhance task guidance.
Updates for TaskBot Challenge 2
This year’s challenge will be expanded to include more hobbies and at-home activities. Participating teams will be asked to propose innovative ways to incorporate visual aids into every conversation turn when a screen is available. Innovative ideas on improving the presentation of visual aids, as well as the coordination of visual and verbal modalities will be part of the team selection criteria.
“Together with our academic partners, we learned a ton during the first year of the Alexa Prize TaskBot challenge. This will help us ease the task — pun intended — of the academic teams when they build even smarter taskbots in the second year of the challenge,” said Yoelle Maarek, vice president of research and science, Alexa Shopping. “In this collaboration with talented graduate students, we will keep pushing further the boundaries of science for Alexa to delight and ease the life of millions of customers."
Each participating team will receive a $250,000 research grant, Alexa-enabled devices, free Amazon Web Services cloud computing services to support their research and development efforts, access to the TaskBot Toolkit, other data resources, and Alexa team support.
Similar to last year, teams will be able to utilize datasets from Whole Foods Market and wikiHow in developing their taskbots. These data sources will be readily available to teams via application programming interfaces (APIs), streamlining development efforts while allowing teams to deliver quality customer experiences.
Starting in the spring of 2023, customers will be able to interact with the taskbots by saying, “Alexa, let’s work together”. That prompt will initiate an interaction with one of the taskbots. After the interaction ends, the customer will be asked to rate how helpful that taskbot was in guiding the interaction, and will have the option to provide freeform feedback to help the teams improve their taskbots.
“Building on our experience with the first Taskbot Challenge, we know that success will require participating teams to address open research problems in identifying and representing relevant general and specialized knowledge for the task from a variety of sources, and incorporating this knowledge into the conversation in appropriate and engaging ways,” Agichtein said. “It will also require examining long standing challenges in commonsense and causal reasoning, and multimodal language processing and response generation.”
How to apply
The application period for the challenge begins on Sept. 22 and runs through Nov. 4, 2022. Up to 10 teams will be selected to participate in the challenge by Nov. 21, 2022, and the competition will begin on Jan. 9, 2023. Participating teams will be provided access to the APIs and the TaskBot toolkit during a bootcamp in January. The challenge will conclude with winners being announced in September 2023. For more information, visit the TaskBot FAQs.