It is certainly true that many industries are trying to adopt AI in their workflow. However, there are many cases that still need human intervention even with the support from AI, such as the determination of recidivism rate and HR decisions. How can we stimulate the collaboration between AI and the domain experts?
In many cases, what holds back the adoption of AI is the lack of manpower equipped with both domain and AI knowledge. Because people don’t know how to describe their goal in an AI-interpretable way and how to interpret the result, AI is being ostracized! Can a system with proper UI solve this problem?
The system that we designed has 3 major components to solve problems. Task, Grammar and UI. First of all, we are going to collect tasks that are AI supportable yet common across the industries at the same time. After that, we are going to build a specific grammar structure which will help us to reduce tasks to certain AI workflows. Lastly, we are going to map these reduced tasks into specific UI components that will support users’ interpretation toward both intrinsic model features and the model results.
July 2021 ~ TBD
Massiv amount of ideation until now…🤯 (and much more to come!)
- This project is advised by Prof. Juho Kim