You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The recent success of Large Language Models (LLMs) signifies an impressivestride towards artificial general intelligence. They have shown a promisingprospect in automatically completing tasks upon user instructions, functioningas brain-like coordinators. The associated risks will be revealed as wedelegate an increasing number of tasks to machines for automated completion. Abig question emerges: how can we make machines behave responsibly when helpinghumans automate tasks as personal copilots? In this paper, we explore thisquestion in depth from the perspectives of feasibility, completeness andsecurity. In specific, we present Responsible Task Automation (ResponsibleTA)as a fundamental framework to facilitate responsible collaboration betweenLLM-based coordinators and executors for task automation with three empoweredcapabilities: 1) predicting the feasibility of the commands for executors; 2)verifying the completeness of executors; 3) enhancing the security (e.g., theprotection of users' privacy). We further propose and compare two paradigms forimplementing the first two capabilities. One is to leverage the genericknowledge of LLMs themselves via prompt engineering while the other is to adoptdomain-specific learnable models. Moreover, we introduce a local memorymechanism for achieving the third capability. We evaluate our proposedResponsibleTA on UI task automation and hope it could bring more attentions toensuring LLMs more responsible in diverse scenarios. The research projecthomepage is athttps://task-automation-research.github.io/responsible_task_automation.
AkihikoWatanabe
changed the title
あ
Responsible Task Automation: Empowering Large Language Models as
Responsible Task Automators, Zhizheng Zhang+, N/A, arXiv'23
Jun 16, 2023
URL
Affiliations
Abstract
Translation (by gpt-3.5-turbo)
Summary (by gpt-3.5-turbo)
The text was updated successfully, but these errors were encountered: