人类人工智能团队合作从 1A 级迈向 2B 级.pdf

1、An Agency of the European Union1HUMAN AI Teaming and moving from Level 1A to 2BOr 2B or not 2B is that a question?EASA Concept Paper for AI2EASA AI trustworthiness frameworkSafety&Security Assessments3AI classification schemeEASA describes three general AI levels Based on industry regarding the type

2、s of use cases foreseen by AI-based systems,Considering a staged approach that industrial stakeholders are planning for the deployment of AI applications,Starting with assisting functions(Level 1 AI),then making a step towards more human-AI teaming(Level 2 AI)and at last seeking for advanced automat

3、ion of the machine(Level 3 AI).4AI Classification schemeAI level Function allocated to the system to contribute to the high-level task Authority of the end user Level 1A Human augmentation Automation support to information acquisition Full Automation support to information analysis Full Level 1B Hum

4、an assistance Automation support to decision-making Full Level 2A Human-AI cooperation Directed decision and automatic action implementation Full Level 2B Human-AI collaboration Supervised automatic decision and action implementation Partial Level 3A Safeguarded advanced automation Safeguarded autom

5、atic decision and action implementation Limited,upon alerting Level 3B Non-supervised advanced automation Non-supervised automatic decision and action implementation Not applicable IncreasingAutomation withcooperation or collaborationAdvanced automationNo automaticdecision-making at high-level taskI

6、ncreasingauthority of the machine5AI classification schemeAnticipated MOC Anticipated MOC-CL-01-2:The following considerations support the delineation of boundaries between AI levels:The boundary between level 1A and level 1B is based on the notion of decision-making.Level 1A covers the use of AI/ML