Democratised and Co-creative AI

視線推定技術開発のデータセット生成ゲーム&ワークショップでAIの未来を考える機会のデザイン

AICOMは、学習データ獲得のためのゲームとワークショップがセットになった機会を提供することによって研究者と一般の方とのコミュニケーションの機会を作り、最先端AIテクノロジーを民主化するための新しい仕組みをデザインするプロジェクトです。

AICOMのデータ獲得ゲームでは2人のプレイヤーがペアになり、間に挟んだ透明の文字盤上で相手が見ている文字を単語を予想しながら当てることでポイントを競い合います。ゲーム中の顔画像が正解の文字とともに記録されるため、顔画像と視線方向がペアになった、視線推定AIの学習に用いることのできるデータセットが同時に出来上がることになります。

ゲーム参加と合わせて人工知能技術の紹介やワークショップツールを用いたアイデア発想ワークショップの機会も設けることで、参加者が一方的にデータを獲得されるのではなく研究者と一緒に未来を考える機会を提供します。

東京大学生産技術研究所DLX Design Labと、視線推定技術の開発を行っている同研究所の菅野研究室が共同で行っているこのプロジェクトでは、研究者側がAI研究を進める上で必要不可欠なデータ収集過程をゲーム化することで広く一般に開かれたものにすることを実現しながら、参加者側もAIや機械学習技術を学びながらデータが必要とされる理由や目的を理解できるような、相互にとって意義のある仕組みをデザインすることが目的になっています。

Project Website
https://aicomjp.editorx.io/aicom

The alternative science communication design for democratising latest AI research by dataset generation game and co-creation workshop

The AICOM project aims to create better communication between scientists and local communities for democratising latest AI technologies and building the future society together.AICOM provides a new gamified experience for collecting training datasets required for developing gaze estimation models. Two players are paired up during the game. One player looks at a character on a transparent board, and the second player has to guess which character the other player is looking at. They will compete with each other while enjoying the face to face communication during the game.

The player’s face images are recorded during the game, and the face images with their corresponding gaze direction annotation will be saved as a part of the dataset after the game. Such a large and diverse image dataset will contribute to gaze estimation research.Unlike previous data collection schemes, the AICOM project will also provide a workshop session after the game experience, to create a better understanding of the technology and imagine the future of AI together. Using the tools designed for the workshop, participants will think together about their own new AI ideas based on the game and lecture experience. AICOM is a collaboration project with the Sugano lab at IIS, which researches artificial intelligence and machine learning as a leading expert in gaze estimation.

Dataset collection is a common bottleneck of AI research. Instead of the data set has been recorded in shadow, we would like to provide an opportunity, users can understand the process of the dataset generation, at the same time AI researchers can receive the benefit to gain the better data sets which will help their latest AI research. We would like to provide an opportunity for better understanding and positive inspirations of the latest AI researches and increase the number of people in society who would like to make their own AIs. We are hoping to create communication and engagement between the researchers and local people for fostering advanced AI technologies and building a better society together.

Awards

CREDIT

Co-Director

Tomomi Sayuda (Design Researcher, DLX Design Lab, UTokyo-IIS )
Yusuke Sugano (Associate Prof., Sugano Lab, UTokyo-IIS,)

Product Developer

Tetsuya Yamamoto, Daisuke Kasuga

Project Advisor

Miles Pennington (Design Researcher, DLX Design Lab, UTokyo-IIS)

Yen Ching-Chiuan (Professor, National University of Singapore, NUS-Keio CUTE Centre)

Project founded by

UTokyo Ushioda Foundation