Alexa TaskBot Challenge Award

A team of graduate students at the University of Glasgow have emerged as winners of Amazon’s inaugural Alexa Award competition with GRILLBot, a “task-oriented, multi-media digital assistant to guide users through complex real-world tasks”.

Amazon launched the Alexa Award in 2016 with the goal of building conversational agents who are able to talk coherently and engagingly with humans on popular topics. There is still a $1 million prize to be won by the team that can build a SocialBot that can hold a compelling conversation for the full 20 minutes. Meanwhile, two more competitions have been added under the banner of the Aleza Prize dedicated to accelerating the field of artificial intelligence and shaping the next generation of virtual assistants. SimBot Challenge:

Focus on helping develop next-generation virtual assistants that will help humans complete real-world tasks through continuous learning, and gain the ability to perform logical reasoning.

Currently in progress with 10 participating university teams.

TaskBot’s more modest goal is to help humans with real-world household tasks, while making use of audio and visual techniques. This challenge requires TaskBots to identify and understand user needs, define and integrate tasks and domain knowledge into interaction, and develop new ways to engage the user without distracting from the task at hand, among other challenges. As the first conversational AI challenge to integrate customer experiences, multimedia, voice and vision, in addition to receiving verbal instructions, customers can be presented with images or graphs using Echo display devices to guide them through the task. This challenge is set to run for three years, in parallel with the Socialbot Alexa Grand Prize competition, and the focus in the first year was cooking and home improvement.

According to Yoelle Maarek, Vice President of Research and Science, Alexa Shopping:

Success in the challenge will require participants to develop state-of-the-art technology in conversational AI, address challenging scientific challenges related to knowledge representation and inference, logical reasoning and causal reasoning, language understanding and generation, among others—requiring the synthesis of multiple domains and methods in AI.

Ten university teams were selected to participate in the inaugural TaskBot Challenge, and each received a $250,000 research grant in addition to Alexa-enabled devices, free Amazon Web Services (AWS) cloud computing to support research and development efforts, and access to the TaskBot Toolkit, In addition to other resource data, Alexa support team. All participating teams have published papers based on their research which are included in the Alexa Award process. Filter on the TaskBot Challenge to find them.

The competition started in June 2021 and went through several phases including an internal Amazon beta period, where Amazon employees tested TaskBots and provided feedback. Then, Amazon Alexa customers in the US region were able to reach Taskbots by saying, “Alexa, help me.”

Amazon customers are encouraged to try the 10 TaskBots with requests to help fix a leaky faucet, prop up and crumble apples to create a delicious meal with holiday leftovers.

Task Robot

At the end of the interaction, clients were asked to rate the usefulness of the TaskBot in the task on a scale of 1 to 5.

Five teams advanced to the finals and competed against the judges who scored TaskBots points for their ability to successfully complete the required task. The top performers receive a prize money that is divided among team members: $500,000 for first place, $100,000 for second place, and $50,000 for third place.

The first prize went to a team from the University of Glasgow for the GrillBot that had an average rating of 3.86/5.0 for the last month. Use the team’s solution TaskGraphs as a dynamic graph that unifies steps, requirements, and curated domain knowledge enabling detailed contextual explanations and adaptive task execution. Their paper says:

Automatic linking of multimedia elements helps the user to navigate through the task and enrich the experience with useful videos and images. Extensive use of neural language models makes for flexible chatting, contextual intent analysis, and accurate task retrieval.

In second place, a team from the NOVA School of Science and Technology in Portugal came up with their robot Twiz, which introduced a new element to the challenge by stating curious facts related to the tasks being performed, which increased customer engagement.

Ohio State University won third place with TacoBot, which has an average rating of 3.55/5.0 and is designed to offer an accessible, collaborative dialogue experience.

more information

Alexa TaskBot Challenge Award

Three top performers appear in the inaugural Alexa Award TaskBot competition

GrillBot (pdf)

Related Articles

Leave a Comment

Your email address will not be published. Required fields are marked *