探索AI前沿：AI代理的演变和影响入门

W H I T EP A P E RD E C E M B E R2 0 2 4 Contents Foreword3 Executive summary4 Introduction5 1Definition of an AI agent6 2The evolution of AI agents8 2.1. Key technological trends92.2. Types of AI agents102.3. Advanced AI agents122.4. AI agent system132.5. The future of AI agents: Towards multi-agent systems14 3Looking ahead17 3.1. Key benefits183.2. Examples of risks and challenges183.3. Addressing the risk and challenges20 Conclusion22 Contributors23 Endnotes26 Disclaimer This document is published by theWorld Economic Forum as a contributionto a project, insight area or interaction.The findings, interpretations andconclusions expressed herein are a resultof a collaborative process facilitated andendorsed by the World Economic Forumbut whose results do not necessarilyrepresent the views of the World EconomicForum, nor the entirety of its Members,Partners or other stakeholders. ©2024 World Economic Forum. All rightsreserved. No part of this publication maybe reproduced or transmitted in any formor by any means, including photocopyingand recording, or by any informationstorage and retrieval system. Foreword Fernando AlvarezChief Strategy andDevelopment Officer,Capgemini Jeremy JurgensManaging Director, WorldEconomic Forum In the contemporary world, where technology israpidly reshaping every aspect of our lives,AI agents are emerging as transformative toolsthat are redefining human interactions and theoperation of our society. These agents, whichbegan as simple computer programs, haveevolved into sophisticated systems with thecapability for autonomous decision-making.This evolution signifies a major shift, positioningAI agents as active participants in crucial sectorssuch as healthcare, education, financial servicesand beyond. attention and proactive management. Ensuringthat AI development aligns with societal valuesand aspirations is paramount to its successfulintegration into daily life. The aim of theseinnovations is to amplify human ingenuity – not toreplace it – within our economy. This comprehensive overview serves as animportant resource for those involved in shaping thefuture of AI technology. By exploring the capabilitiesand implications of AI agents, stakeholders canbetter understand how to leverage the power ofthese systems to drive meaningful progress acrossvarious sectors. It is through this understandingthat we can ensure AI technologies are developedresponsibly and used in ways that enhance humanwell-being. With careful stewardship, AI agents canbecome invaluable allies in fostering innovation andimproving quality of life worldwide. The advancement of AI agents brings with it awealth of exciting possibilities and transformativepotential. Their ability to manage complex taskswith minimal human intervention offers the promiseof significantly increased efficiency and productivity. However, as we step into this AI-driven era, itis essential to not only harness the immensebenefits these technologies offer, but also toaddress the challenges they present. Issuessuch as ethical considerations require careful In partnership, the World Economic Forum andCapgemini have joined forces through the AIGovernance Alliance to advance this critical topic incollaboration with the AI community. Executive summary This paper examines the development andfunctionality of AI agents – and the implicationsof their use – amid rapid advances in largelanguage and multimodal models. Defined as autonomous systems that sense and actupon their environment to achieve goals, artificialintelligence (AI) agents are being deployed in a widerange of roles in different industries. This requiresthe adaptation of governance frameworks to ensureresponsible adoption. sectors such as healthcare, customer serviceand education. However, AI agents also presentnovel risks, including potential misalignment,2along with ethical concerns about transparencyand accountability. Future advances in the area are likely to involvemulti-agent systems (MAS), where AI agentscollaborate to address complex challenges such asurban traffic management. More advanced systemsintroduce new demands for interoperability andcommunication standards to function effectively,while these protocols still need to be debated andagreed upon by a wider community. AI agents, comprising components such assensors and effectors, have evolved from rule-based systems to advanced models capableof complex decision-making and independentoperation. Enabled by breakthroughs in deeplearning, reinforcement learning and the transformerarchitecture,1AI agents span applications fromworkflow automation to personal assistants. Thisprogression now encompasses more sophisticatedutility-based AI agents that incorporate memory,planning and tool integration, broadening theircapabilities and relevance. This paper highlights the need for robustgovernance, ethical guidelines and a cross-sectoralconsensus to integrate AI agents safely into society.As more advanced AI agents