Workshop on Artificial Intelligence and the "Barrier of Meaning"

本次研讨会汇集了计算机科学、心理学、生物学、神经科学等领域的杰出学者，讨论人工智能中的“理解”问题。在这个活动中，参与者将考虑先进的计算机系统拥有类似人类的理解能力意味着什么，探索智能系统展示这种理解能力的必要性，并讨论赋予这些系统这种能力的方法。多学科社区的参与将为我们如何定义、设计、实施和控制克服这一意义障碍的复杂系统提供新的、可操作的见解。研讨会还可能产生成果和后续活动，以促进人工智能教育和公众对当前人工智能状况的认识，包括其局限性和潜在脆弱性。本次研讨会的方法是探索复杂系统如何从它们遇到的信息中提取意义。研讨会参与者将参与有关“理解”或“提取意义”在许多学科的复杂系统中的功能和机制的问题，并特别关注类人理解与创建可靠的、适应新情况的、对对抗性攻击具有鲁棒性的人工智能系统的相关性。理解理解的本质和必要性仍然是人工智能研究中最深刻的智力挑战之一。讲习班讨论将旨在澄清常见问题并确定可能的新途径来回答这些问题。组织者将发布技术和一般读者摘要，传达研讨会讨论的结果，这些讨论涉及理解或意义作为不同学科现象的概念，以及这些现象如何与现实世界中安全可靠的人工智能所需的鲁棒性相关或实现。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

This workshop brings together eminent scholars in the fields of computer science, psychology, biology, neuroscience, and others to address the topic of "understanding" in artificial intelligence. In this activity, participants will consider what it would mean for advanced computer systems to possess human-like understanding, explore how necessary it is for intelligent systems to exhibit such understanding, and discuss approaches to imbuing these systems with such a capability. Engagement of a multidisciplinary community will develop new and actionable insight into how we define, design, implement, and control complex systems that overcome this barrier of meaning. The workshop will likely also lead to outcomes and follow-on activities to benefit AI education and public awareness regarding the state of current artificial intelligence, including its limitations and potential vulnerabilities.The approach in this workshop is to explore how complex systems extract meaning from the information they encounter. Workshop participants will engage questions about the function and mechanisms of "understanding" or "extracting meaning" in complex systems across many disciplines, and focus specifically on the relevance of human-like understanding for creating artificial intelligence systems that are reliable, adaptable to novel situations, and robust against adversarial attacks. Understanding the nature and necessity of understanding remains among the deepest intellectual challenges in AI research. Workshop discussions will be aimed at clarifying common questions and identifying possible novel pathways to answering these questions. Organizers will publish both technical and general-readership summaries communicating the results of the workshop discussions concerning the notions of understanding or meaning as phenomena in diverse disciplines, and how these phenomena relate to, or enable, the robustness that will be needed for safe and trustworthy AI in the real world.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.