The Second Resiliency of Intelligent Automation Systems Challenge

Abstract

Intelligent Process Automation (IPA) systems such as UiPath, Microsoft Power Automate, Automation Anywhere and IBM RPA are a rapidly-evolving technologies that are transforming many domains by automating repetitive and labor-intensive tasks. Recent demonstrations of ChatGPT and Adept drive a new generation of these IPA systems. However, like any AI system, they are susceptible to unexpected changes (e.g., changes in the application layout), which may affect their performance and reliability. This is a major unsolved problem that prevents automation from scaling to the long tail of potential user needs and for enterprises to adopt and maintain AI-based automation systems. In addition, for humans to adopt this new form of AI, it must exhibit consistent behavior that humans can trust. A key aspect of this is resiliency, which is the focus of this challenge. Our proposed challenge is a unique learning opportunity for participants to explore the capabilities and limitations of state-of-the-art automation technologies. Through a series of tasks and activities, participants can gain practical experience with IPA technology and develop essential skills of problem-solving, critical thinking, and creativity. This challenge highlights the importance of robustness in automation systems, as participants will be asked to design and test solutions that can withstand various changes in the automated task. By participating in this challenge, participants will learn to develop new AI-based algorithms and gain valuable insights into the potential of IPAs to enhance processes and drive innovation in this rapidly growing field. This challenge is based on last year's IPA challenge at IJCAI23 ({https://iparesiliency-ijcai23.github.io/}). A demo paper has been submitted to the Demo track of IJCAI24 summarizing the results and experiments of last year's challenge. This year, we plan to extend the challenge such that the AI system will have to also detect the semantic state of UI elements and present resistance when they change.