Nine teams move on to semi -final for Alexa Prize Socialbot Grand Challenge

In November, Amazon announced that nine teams from around the world, we chose to participate in Alexa Prize Socialbot Grand Challenge 5 (SGC 5), a university challenge that focused on promoting interaction between people and computer and open dialogue conversation. As of today, all nine teams have moved on to the semi -finals based on their performance during the initial customer feedback periodback.

The teams selected for the challenge, which began in November, included five recurring competitors – including the two top finishers in the latest challenge – and four new universities.

“Sayte its beginning in 2016, Socialbot Grand Challenge has driven technical advances in neural response generation and the use of large language models on open domain dialogue,” said Michael Johnston, a used science director at Alexa AI, who leads the science and engineering teams supporting the Alexa award. “This year, SGC5 teams use and integrate and integrate a wide range of different large language models into social bots, and it is super exciting to see the kind of interactive experiences they can activate for Alexa customers”

The Compaeting teams are also facing a challenge that is recently intacted for SGC5: Their Socialbots must provide a competent multimodal user experience and integrate speech with visuals. Team pursues a wide range of approaches, including emotional avatars, synchronized graphics and multimedia, image generation and multimodal dialogue using tips and touch input.

“It will be incredibly interesting to see approaches that have been proven to be effective in the final,” Johnston added.

“Creating a socially skilled AI is a tough problem,” said Reza Ghanadan, senior main scientist at Alexa AI and head of the Alexa award. “This is because the human-like social conversation is to note delicate and complex, and the open domain of the Socialbot dialogues makes it extremely challenging.

“You have to give relevant and deep resorts to a wide range of topics, awareness of distinguishing between reality and imagination, maintaining a natural and coherent exchange through a potentially long conversation and accurately interpreting the intention by properly picking up names, topics considering the context of each conversation.

The Alexa Award is a unique partnership program for the Industry Academy that provides an agile real-world experimental framework and tools to accelerate scientific discovery. University students have the opportunity to launch innovations online and quickly adapt to feedback from Alexa customers.

“Prize competitions provide data, AI tools and a smooth experimentation framework for researchers and students to innovate on advanced topics in creating socially intelligent digital assistants, calling for them to explore transformation ideas on the bonds about what can be achieved in the real world,” Ghanadan said.

Alexa customers can interact with University Socialbots by saying “Alexa, Let’s Chat” on Amazon Echo or four TV devices. Customer assessments and feedback help the student teams improve their bots leading up to the competition final.

The ultimate goal is to accommodate the Grand Challenge: Earn a composite score of 4.0 or higher (out of 5) from a panel of Judes, and get these Judes to find that at least two -thirds of their social bot conversations in the final round of judge remain coherent and engage for 20 minutes. The first team that meets the Grand Challenge wins a $ 1 million research fellow for their university.

Updates for this year’s competition

Absed above is this the first iteration of the Socialbot Grand Challenge to incorporate multimodal customer experiences. In addition to verbal conversations, customers with Echo screeners or a brand -TV can be presented with images or text that improves the conversation experience. Teams have the opportunity to improve their customer interactions by including additional text and images that provide more different and meaningful information.

This year there are also two sets of prices: a set for the overall social interaction performance and a set for scientific innovation. Prizes for the overall performance in the competition will be $ 250,000 for the first plate team, $ 50,000 for the second and $ 25,000 for the third.

The new award for scientific invention and innovation allows teams to focus on promoting the area of ​​conversation AI through a deeper study of the basic elements of open dialogue conversations. Cash prizes for scientific contribution will be awarded students on the winning team- $ 250,000 for first place, $ 50,000 for second and $ 25,000 for the third.

A unique challenge

Socialbot Grand Challenge represents a unique opportunity for student researchers to experience and learn how their ideas work within a real environment.

“We have learned that successful demands researchers to create generalizable AI techniques and incorporate nowledge in appropriate and engaging ways,” Ghanadan said. “It also involves tackling open research problems in natural language understanding and multimodal language processing, contextual understanding, natural response generation, empathy and commonsical repairs, to understand social norms and dialogue management.”

Hvert universitet, der er valgt til udfordringen, modtager et forskningsstipendium på op til $ 250.000, Alexa-aktiverede enheder, gratis Amazon Web Services (AWS) cloud computing-tjenester til støtte for deres forskning og udviklingsindsats, adgang til Amazon-forskere, COBOT (Conversational Bot) værktøjssæt og andre værktøjsgenkendelse gennem Alexa, neurale detektion og genereringsmodeller, Conversational Data Sets and Design Guidance and Development Support from Alexa Price-Price.

In previous challenges, participating teams have improved the latest species for open domain dialogue systems by developed improved natural language understanding systems (NLU), neural response generation models, models of common sense and dialogue policies leading to smoother and more engaging conversations.

The “Assust” team from the Czech Technical University won the fourth Challen, with teams from Stanford and the University of Buffalo, who earned second and third location prizes, respective. The publications from this challenge can be found here.

Winning teams from previous years include Emory University, the University of Washington and the University of California, Davis.