《ITU：2024年人工智能治理日-從原則到實施報告（英文版）（98頁）.pdf》由會員分享，可在線閱讀，更多相關《ITU：2024年人工智能治理日-從原則到實施報告（英文版）（98頁）.pdf（98頁珍藏版）》請在三個皮匠報告上搜索。

1、ITUPublicationsInternational Telecommunication UnionTelecommunication Standardization SectorAI Governance Day From Principles to Implementation2024 ReportPlease consider the environment before printing this report.ITU 2024Some rights reserved.This work is licensed to the public through a Creative Co

2、mmons Attribution-Non-Commercial-Share Alike 3.0 IGO license(CC BY-NC-SA 3.0 IGO).Under the terms of this licence,you may copy,redistribute and adapt the work for non-commercial purposes,provided the work is appropriately cited.In any use of this work,there should be no suggestion that ITU endorse a

3、ny specific organization,products or services.The unauthorized use of the ITU names or logos is not permitted.If you adapt the work,then you must license your work under the same or equivalent Creative Commons licence.If you create a translation of this work,you should add the following disclaimer a

4、long with the suggested citation:“This translation was not created by the International Telecommunication Union(ITU).ITU is not responsible for the content or accuracy of this translation.The original English edition shall be the binding and authentic edition”.For more information,please visit https

5、:/creativecommons.org/licenses/by-nc-sa/3.0/igo/DisclaimersThe designations employed and the presentation of the material in this publication do not imply the expression of any opinion whatsoever on the part of ITU concerning the legal status of any country,territory,city or area or of its authoriti

6、es,or concerning the delimitation of its frontiers or boundaries.The mention of specific companies or certain manufacturer products does not imply that they are endorsed or recommended by ITU in preference to others of a similar nature that are not mentioned.Errors and omissions excepted,the names o

7、f proprietary products are distinguished by initial capital letters.All reasonable precautions have been taken by ITU to verify the information contained in this publication.However,the published material is being distributed without warranty of any kind,either expressed or implied.The responsibilit

8、y for the interpretation and use of the material lies with the reader.The opinions,findings and conclusions expressed in this publication do not necessarily reflect the views of ITU or its membership.ISBN978-92-61-39291-8(Electronic version)978-92-61-39301-4(EPUB version)978-92-61-39311-3(Mobi versi

9、on)AI Governance Day-From Principles to Implementation2024 ReportiiAI Governance Day From Principles to Implementation 10.Executive Summary.11.Foreword by Doreen Bogdan-Martin,ITU Secretary-General.32.Preface by Prof.Robert Trager,moderator of AI Governance Day.43.Format of AI Governance Day.63.1 St

10、ructure of AI Governance Day.63.2 Run of show,morning session.73.3 Run of show,afternoon session.84.Morning sessions.104.1 Welcome speech by Doreen Bogdan-Martin,ITU Secretary-General.104.2 How can we move AI governance from principles to implementation?Multistakeholder panel.124.3 Theme 1:What is t

11、he landscape of AI governance and how will it evolve?.174.3.1 Examples of multilateral and national AI governance frameworks.174.3.2 The debate about national vs international AI governance.194.3.3 Insights from the breakout sessions:theme 1.204.4 Theme 2:how do we implement AI governance frameworks

12、?.224.4.1 Lagging laws,lagging tech:the AI governance paradox.224.4.2 Compute governance.234.4.3 Insights from the breakout sessions:theme 2.244.5 Theme 3:how do we ensure inclusion and trust?.264.5.1 Inclusion connecting the unconnected.264.5.2 Giving voice for the global majority.274.5.3 Open vs c

13、losed sourcing of(generative)AI models.284.5.4 Insights from the breakout sessions:theme 3.294.6 Update from United Nations high-level advisory body on AI .304.7 Governance pulse check.315.Public afternoon session.335.1 Welcome speech,Doreen Bogdan-Martin,ITU Secretary-General.33Table of contentsiii

14、5.2 Leaders speak:Insights and key findings on AI governance implementation.355.3 State of play of major global AI governance processes.405.4 Leveraging the UN system to advance global AI Governance efforts.475.5 The Governments AI dilemma:how to maximize rewards while minimizing risks?.535.6 The cr

15、itical conversation on AI safety and risk.595.7 To share or not to share:the dilemma of open source vs.proprietary large language models.675.8 Harmonizing high-tech:the role of AI standards as an implementation tool.735.9 From principles to implementation pathways forward.81Appendix 1:Essential voca

16、bulary for AI governance.84Appendix 2:List of participants of the morning session.85Appendix 3:Acknowledgments and sources.86ivList of figures and tablesFiguresFigure 1:Doreen Bogdan-Martin,Secretary-General,International Telecommunication Union(ITU).4Figure 2:Robert F.Trager,Professor,University of

17、 Oxford.6Figure 3:Anka Reuel,PhD student of computer science,Stanford University,co-moderator.7Figure 4:Roundtable participants during the morning breakout session.9Figure 5:Center Stage Welcome Address featuring Doreen Bogdan-Martin,Secretary-General,International Telecommunication Union(ITU).9Figu

18、re 6:The panel of speakers from a Center Stage discussion.10Figure 7:Speakers from the Multistakeholder Panel:How can we move AI Governance from Principles to Implementation?.12Figure 8:H.E.Ms.Rose Pola Pricemou,Ministre,Guinea(Ministre des Postes,des Tlcommunications et de lEconomie Numrique)during

19、 the Multistakeholder Panel.13Figure 9:Rumman Chowdhury,CEO,Humane Intelligence,during the Multistakeholder Panel.14Figure 10:Tristan Harris,Director,Executive Director of the Center for Humane Technology,speaking during the Multistakeholder Panel.15Figure 11:Stuart Russell,Professor of Computer Sci

20、ence,University of California,Berkeley,speaking during the Multistakeholder Panel.16Figure 12:Lane Dilg,Head of Strategic Partnerships,OpenAI,speaking during the Multistakeholder Panel.16Figure 13:Individuals using the Internet/Percentage of population vs time.27Figure 14:Carme Artigas,Co-Chair of U

21、nited Nations High-level Advisory Body on Artificial Intelligence,reporting live from Singapore.31Figure 15:Pre-pulse check on whether participants strongly disagreed(value of 1)or strongly agreed(value of 5)with statements regarding AI in general.The scale was from 1 to 5.32Figure 16:Pre-pulse chec

22、k on whether participants strongly disagreed(value of 0)or strongly agreed(value of 5)with statements regarding AI governance.The scale was from 0 to 5.32Figure 17:The panelists and moderator from the Center Stage session Leaders speak:Insights and key findings on AI governance implementation.36Figu

23、re 18:H.E.Ms.Emma Inamutila Theofelus,Minister of Information and Communication Technology,Namibia.38Figure 19:H.E.Mr.Mauricio Lizcano,Minister of Information Technologies and Communications,Colombia.38Figure 20:H.E.Mr.Zunaid Ahmed Palak,Minister of State for the Ministry of Posts,Telecommunications

24、 and Information Technology,Bangladesh.39vFigure 21:The panelists and moderator of the Center Stage session:State of play of major global AI Governance processes.40Figure 22:Thomas Schneider,Ambassador and Director of International Affairs,Swiss Federal Office of Communications,and Chair of the Coun

25、cil of Europe Committee on AI.41Figure 23:Juha Heikkil,Adviser for Artificial Intelligence,European Commission.42Figure 24:Alan Davidson,Assistant Secretary of Commerce for Communications and Information,Head of the National Telecommunications and Information Administration(NTIA).43Figure 25:Shan Zh

26、ongde,Vice Minister,Ministry of Industry and Information Technology,Peoples Republic of China.44Figure 26:H.E.Mr.Hiroshi Yoshida,Vice-Minister for Policy Coordination,Ministry of Internal Affairs and Communications,Japan.44Figure 27:H.E.Mr.Dohyun Kang,Vice Minister,Ministry of Science and ICT,Republ

27、ic of Korea.45Figure 28:Ebtesam Almazrouei,Founder and CEO of AI E3,United Arab Emirates.46Figure 29:The panelists and moderator from the Center Stage session:Leveraging the UN system to advance global AI Governance efforts.47Figure 30:Doreen Bogdan-Martin,Secretary-General,International Telecommuni

28、cation Union(ITU)speaking alongside Daren Tang(left),Director General,World Intellectual Property Organization(WIPO)and Reinhard Scholl,Program Chair,AI for Good.49Figure 31:Tshilidzi Marwala,Rector,United Nations University;Under-Secretary-General of the United Nations.49Figure 32:Daren Tang,Direct

29、or General,World Intellectual Property Organization(WIPO).50Figure 33:Gilbert Houngbo,Director-General,International Labour Organization(ILO).51Figure 34:Dongyu Qu,Director-General,Food and Agriculture Organization(FAO).52Figure 35:Reinhard Scholl,Programme Chair,AI for Good.55Figure 36:The panel an

30、d moderator from the Center Stage session:The Governments AI dilemma:how to maximize rewards while minimizing risks?.55Figure 37:H.E.Ms.Emma Inamutila Theofelus,Minister of Information and Communication Technology,Namibia.57Figure 38:Mercedes Aramendia Falco,President,Directorio Unidad Reguladora de

31、 Servicios de Comunicaciones(URSEC),Uruguay.58Figure 39:Niraj Verma,Additional Secretary Department of Telecommunications,Government of India.58Figure 40:Robert F.Trager,Professor,University of Oxford.60Figure 41:Panel and moderator from the Center Stage session:The critical conversation on AI safet

32、y and risk.61Figure 42:Amir Banifatemi:Co-Founder and Director of AI Commons.61viFigure 43:Stuart Russell,Professor of Computer Science at the University of California,Berkeley.62Figure 44:Lane Dilg,Head of Strategic Partnerships at OpenAI speaking alongside Professor Stuart Russell,Professor of Com

33、puter Science at the University of California,Berkeley;Rumman Chowdhury,CEO of Humane Intelligence and USA Science Envoy for Artificial Intelligence and Hakim Hacid,Acting Chief Researcher,Technology Innovation Institute(TII).63Figure 45:Rumman Chowdhury,CEO of Humane Intelligence and USA Science En

34、voy for Artificial Intelligence.64Figure 46:Hakim Hacid:Acting Chief Researcher,Technology Innovation Institute(TII).65Figure 47:Lane Dilg,Head of Strategic Partnerships at OpenAI.66Figure 48:Rumman Chowdhury:CEO of Humane Intelligence and USA Science Envoy for Artificial Intelligence.67Figure 49:Pa

35、nelists and moderator during the Center Stage session:To share or not to share:the dilemma of open source vs.proprietary Large Language Models.68Figure 50:Jim Zemlin,Executive Director of the Linux Foundation.69Figure 51:Melinda Claybaugh,Director of Privacy Policy at Meta.70Figure 52:Isabella Hampt

36、on,Policy Researcher at the Future of Life Institute.71Figure 53:Melike Yetken Krilla,Head of International Organizations at Google.71Figure 54:Chris Albon,Director of Machine Learning at the Wikimedia Foundation.72Figure 55:Panelists and moderator during the Center Stage session:Harmonizing High-Te

37、ch:The role of AI standards as an implementation tool.74Figure 56:Philippe Metzger,Secretary-General&CEO,International Electrotechnical Commission(IEC).75Figure 57:Sergio Mujica,Secretary-General,International Organization for Standardization(ISO).76Figure 58:Seizo Onoe,Director of the Telecommunica

38、tion Standardization Bureau(TSB),International Telecommunication Union(ITU).77Figure 59:Panelists and moderator during the Center Stage session:Harmonizing High-Tech:The role of AI standards as an implementation tool.78Figure 60:Bilel Jamoussi,Deputy to the Director and Chief of Telecommunication St

39、andardization Policy Department,International Telecommunication Union(ITU).79Figure 61:Gabriela Ramos(UNESCO)and Tomas Lamanauskas(ITU)care closing AI Governance Day with the session”From principles to implementation pathways forward”.80Figure 62:Tomas Lamanauskas,Deputy Secretary-General,Internatio

40、nal Telecommunication Union(ITU),co-chair of the United Nations Interagency Working Group on AI.80viiFigure 63:Gabriela Ramos,Assistant Director-General for Social and Human Sciences,United Nations Educational,Scientific and Cultural Organization(UNESCO),co-chair of the United Nations Interagency Wo

41、rking Group on AI.82Figure 64:Cover page of the United Nations system White Paper on AI Governance,produced by the Inter-Agency Working Group on AI,available at https:/unsceb.org/united-nations-system-white-paper-ai-governance.83TablesTable 1:Examples of national AI Governance initiatives.17Table 2:

42、Examples of multilateral AI Governance initiatives.181AI Governance Day-From Principles to ImplementationAI Governance Day From Principles to Implementation0 Executive SummaryThe first ever AI Governance Day subtitled From Principles to Implementation was held in Geneva,Switzerland,on 29 May 2024.Co

43、nvened by the United Nations International Telecommunication Union(ITU),the event brought together a kaleidoscope of participants from around the world.This included government leaders,policymakers,researchers,and technologists from both developed and developing countries.The events multistakeholder

44、 composition aligned with ITUs mission to provide an inclusive,neutral,and globally representative platform for artificial intelligence(AI).The day was marked by vibrant discussions and collaborations aimed at transforming AI governance principles into actionable frameworks.In the area of AI,the tra

45、nsition from principles to actionable governance is a challenge.These principles,while vital,have remained at a high level,often too abstract in their application for direct application in the ever-evolving AI landscape.Recently,a significant shift has been observed,as regulatory bodies worldwide ha

46、ve begun to codify these principles into concrete regulations,creating foreseeable regulatory pressure on the development of AI.For example,China has instituted an Algorithm Registry in December 2022,and in October 2023,the President of the United States issued an Executive Order on the Safe,Secure,

47、and Trustworthy Development and Use of Artificial Intelligence.The European Union AI Act has perhaps gone furthest:it is set to enter into force in the second half of 2024 and to become progressively enforceable over a 36-month period.In the rapidly evolving landscape of AI,an AI governance paradox

48、has emerged:while regulation is often seen as lagging behind technological advancements,there is an equally critical yet less recognized issue where the current state of technology and tools available does not allow for monitoring,checking,and controlling AI systems.This gap poses risks and undersco

49、res the need for advancing tools capable of ensuring effective governance.AI Governance Day tackled the step of moving from regulatory frameworks to implementation.How are countries and regions navigating the dual objectives of maximizing AIs benefits while minimizing its risks?Participants shared e

50、xperiences on what works,what does not work(yet),identified hurdles,and discussed what needs to happen next on the path towards effective regulatory implementation.The first half of AI Governance Day the morning sessions were interactive and dynamic discussions among 200 senior leaders from governme

51、nts,the private sector,international organizations,academia,the technical community,and civil society.Three themes were discussed:Theme 1:What is the landscape of AI governance,and how will it evolve?This report highlights the debate at the session about national vs.regional and international AI gov

52、ernance and includes a listing of the major multilateral as well as national initiatives as of the end of May 2024 please see Chapter 4.3.1.Theme 2:How do we implement AI governance frameworks?The session discussed how technical standards wuold be an important piece in measuring whether AI governanc

53、e frameworks are successfully implemented.As the report highlights,the current technical methods are often insufficient to measure the implementation of AI governance frameworks and to provide feedback.The report then goes deeper into compute governance,arguably an area where measurements can be mor

54、e straightforward and quantifiable as opposed to the governance of data or the governance of algorithms.2AI Governance Day-From Principles to Implementation Theme 3:How do we ensure inclusion and trust?Inclusion starts with connecting the unconnected.In 2023,according to ITU statistics,a third of th

55、e worlds population is still not connected to the Internet.Furthermore,many people and many countries that are being impacted by AI are not integrated in ongoing discussions to set AI policies.The report also provides details on the discussion of open vs.closed sourcing of AI models.The second half

56、of AI Governance Day,the public afternoon sessions,built on the morning discussions and featured panel discussions.ITUs Secretary-General Doreen Bogdan-Martin recalled that the conundrum of how to govern a technology if we do not yet know its full potential is not new:twenty years ago,the Internet w

57、as met with a similar mix of shock,awe,and skepticism.But what we have learned from the WSIS(World Summit on the Information Society)process was that we can take steps toward governance even if we are building the plane as we fly.From the discussions in the morning,Ms Bogdan-Martins learnings,based

58、on the fundamental premise of including every nation and every stakeholder group in governance efforts,included three key pieces which must be part of any AI governance effort:(1)development of technical standards;(2)putting core UN values at the heart of AI governance;and(3)development through capa

59、city building.In the session Leaders speak:Insights and key findings on AI governance implementation,three Ministers,from Namibia,Colombia and Bangladesh,underscored the importance of a coordinated,inclusive,and human-centric approach to AI governance.State of play of major global AI governance proc

60、esses featured the Council of Europe,the European Commission,USA,China,Japan(G7 Hiroshima process),and Republic of Korea(follow-up host of the UK AI Safety Summit in November 2023).Leveraging the UN system to advance global AI governance efforts spotlighted the Executive Heads of the International L

61、abour Organization(ILO),the Food and Agriculture Organization(FAO),the World Intellectual Property Organization(WIPO),the United Nations University(UNU),and the ITU.In The governments AI dilemma:How to maximize rewards while minimizing risks?,government leaders from Namibia,Uruguay and India stresse

62、d the benefits of AI such as in healthcare,education,digital literacy training,the need for ethical implementation and data privacy via a national AI strategy,but also the downsides such as the increase in cybercrime.The critical conversation on AI safety and risk saw researchers and company executi

63、ves underscore the complexity and urgency of AI safety and risk management.The discussion in the session To share or not to share:the dilemma of open-source vs proprietary large language models,hosting the Executive Director of the Linux Foundation and executives of Google,Meta,the Wikimedia Foundat

64、ion and a policy researcher of the Future of Life Institute,emphasized that open source plays a critical role in fostering innovation,ensuring transparency and preventing market consolidation,while also recognizing the need for responsible governance to address risks.Harmonizing high-tech:The role o

65、f AI standards as an implementation tool welcomed the Executive Heads of the worlds leading international standards organizations ITU,the International Organization for Standardization(ISO),and the International Electrotechnical Commission(IEC).The three organizations have been coordinating and coll

66、aborating on standards development for decades.(They announced new initiatives on AI at the AI for Good Global Summit,following AI Governance Day.)In the closing session From principles to implementation pathways forward,the two co-chairs(from ITU and UNESCO)of the United Nations Interagency Working

67、 Group on AI,highlighted ongoing UN System-wide coordination efforts on AI,and the United Nation Systems White Paper on AI governance(May 2024).3AI Governance Day-From Principles to ImplementationWe hope that the discussions and new connections formed will assist in moving AI governance forward from

68、 principles to implementation.We are looking forward to welcoming you at the next AI for Good Global Summit(including AI Governance Day)in Geneva in the week of 7-11 July 2025.1 Foreword by Doreen Bogdan-Martin,ITU Secretary-GeneralIt was a tremendous pleasure to welcome a mullti-stakeholder group o

69、f senior leaders from governments,the private sector,international organizations,academia,the technical community,and civil society to the AI for Good Global Summit on 30-31 May 2024 and to AI Governance Day on 29 May 2024.This inclusivity was at the heart of our mission at the International Telecom

70、munication Union(ITU)to create a neutral,global platform for artificial intelligence.AI Governance Day marks a truly global multistakeholder dialogue on AI governance,a topic that had gained significant importance over the past 12 months.Discussions during AI Governance Day focused on three themes:t

71、he AI regulatory landscape,implementing AI governance frameworks,and ensuring inclusion and trust in these frameworks.The invitation-only morning sessions involved surveying the AI landscape to understand the current state of regulations and identify areas of common interest.An ITU survey revealed t

72、hat 85%of our responding Member States had not yet established AI regulations,highlighting the urgent need for guidance and collaboration to avoid fragmented governance.We also had a report from the United Nations High-Level Advisory Body on AI,with a live connection to co-chairs of the United Natio

73、ns Secretary Generals Advisory Body from their final in-person meeting in Singapore.Discussions at AI Governance Day and at the AI for Good Global Summit set the stage for developing actionable,enforceable technical standards,essential for ensuring algorithm transparency,system safety,and security.T

74、he ITU had already published or was developing over 200 AI-related standards,but more are needed.Collaboration through mechanisms like the World Standards Cooperation,involving leaders from ISO and IEC,has been vital for advancing these standards globally.Inclusion and trust are deeply interconnecte

75、d and crucial for successful AI governance.Without trust,the adoption of AI would falter,potentially exacerbating digital divides.A survey by BCG highlighted that consumers in low and middle-income countries were more enthusiastic about AI than those in mature markets,viewing it as a means to leapfr

76、og technological gaps and drive innovation in critical sectors like education and healthcare.Reflecting on lessons from the World Summit on the Information Society(WSIS),we recognized the importance of a multi-stakeholder approach to governance.The WSIS process showed that governance could be iterat

77、ive and inclusive,accommodating diverse perspectives and needs.This approach was deemed equally necessary for AI governance,ensuring that power was not concentrated in the hands of a few and that all stakeholders had a voice.Our focus on human rights,inclusion,and capacity building was emphasized as

78、 paramount.It was essential that AI governance frameworks reflected core UN values and addressed the needs of all countries.The ITUs initiatives,in collaboration with UNDP and other partners,aimed to 4AI Governance Day-From Principles to Implementationupskill nations with low technological capabilit

79、ies,helping them navigate the challenges and opportunities of AI.I urge everyone to be ambitious and visionary.The goal is to make AI governance the cornerstone of a better digital future for all.Governance is a continuous process of listening,exchanging ideas,and implementing practical solutions.To

80、gether,we aimed to harness the power of AI for the greater good,ensuring it served as a force for positive global change.I hope that the outcomes of the discussions held in Geneva will resonate beyond AI Governance Day and the AI for Good Global Summit.Doreen Bogdan-Martin ITU Secretary-GeneralFigur

81、e 1:Doreen Bogdan-Martin,Secretary-General,International Telecommunication Union(ITU)2 Preface by Prof Robert Trager,moderator of AI Governance DayAt the 2023 edition of the AI for Good Global Summit,the ITU Secretary-General called for guardrails to ensure AI is safe,responsible and inclusive.As th

82、e speed of innovation races forward in AI,the transition from principles to actionable governance is a challenge.These principles,while vital,have predominantly remained at a high level,often too abstract in their application for direct application in the ever-evolving AI landscape.5AI Governance Da

83、y-From Principles to ImplementationRecently,a significant shift has been observed,as regulatory bodies worldwide have begun to codify these principles into concrete regulations,creating foreseeable regulatory pressure on the development of AI.For example,China has instituted an Algorithm Registry,an

84、d in October 2023,the President of the United States issued an Executive Order on the Safe,Secure,and Trustworthy Development and Use of Artificial Intelligence.The European Union AI Act has perhaps gone furthest:it is set to enter into force in the second half of 2024 and to become progressively en

85、forceable over a 36-month period.What is AI Governance Day?AI Governance Day spotlighted the status of worldwide efforts to shape the future of AI regulation and hosted discussions on implementation strategies that are key to ensuring the AI race leaves no one behind.It spanned a full day of interac

86、tive sessions,insightful discussions,and networking opportunities.With a focus on fostering collaboration among prominent stakeholders,AI Governance Day provided a platform for government ministers,regulators,industry leaders,academicians,civil society,and UN representatives to engage in meaningful

87、dialogue and chart the course for effective AI governance frameworks.As the day opened,ministers and regulators tackled these questions:What is the landscape of AI governance,and how will it evolve?How do we implement AI governance frameworks?How do we ensure inclusion and trust?Bringing AI governan

88、ce discussions into focusAfternoon sessions,open to the public,featured distinguished speakers and thought leaders,including high-level government officials to representatives from leading international organizations,reporting on the mornings outcomes and providing the state of play of major global

89、AI governance processes.Panel discussions included these topics:The governments AI dilemma:How to maximize rewards while minimizing risks?Leveraging the UN system to advance global AI governance efforts The critical conversation on AI safety and risk To share or not to share:the dilemma of open-sour

90、ce vs proprietary large language models Harmonizing high-tech:The role of AI standards as an implementation toolWe hope that the results of the discussions and the many new connections being made among the participants will assist in moving AI governance forward from principles to implementation.Rob

91、ert F.Trager,University of Oxford;Director,Oxford Martin AI governance initiative;senior research fellow,Blavatnik School of Government;International governance lead,centre for the governance of AI.Professor Trager moderated AI Governance Day.6AI Governance Day-From Principles to ImplementationFigur

92、e 2:Robert F.Trager,Professor,University of Oxford3 Format of AI Governance Day31 Structure of AI Governance DayAI Governance Day took place on Wednesday 29 May 2024 at the Geneva International Conference Center(CICG).AI Governance Day was structured into a morning session from 9:00 to 13:00,followe

93、d by a networking lunch.The morning session was by invitation only and for those who purchased a Leaders Pass.The morning session was focused,with 200 senior leaders engaging in the discussions.Chatham House Rule were used,i.e.,participants we free to use the information received,but not the identit

94、y or the affiliation of the persons.About two weeks prior to AI Governance Day,briefing notes prepared with the support of independent experts were distributed to the participants of the morning session of AI Governance Day.The briefing notes were structured according to the three themes discussed i

95、n the morning:Theme 1:What is the landscape of AI governance and how will it evolve?Theme 2:How do we implement AI governance frameworks?Theme 3:How do we ensure inclusion and trust?The text of the briefing notes has been incorporated into this report.7AI Governance Day-From Principles to Implementa

96、tionThe afternoon session was public and live-streamed and featured panel discussions,available here.Master of Ceremony and Moderator of AI Governance Day was Professor Robert Trager,University of Oxford.He was assisted in the morning session by co-moderator Anka Reuel,a computer science PhD student

97、 at Stanford University whose research focuses on enabling organizations to develop and deploy responsible,socially beneficial AI systems.Figure 3:Anka Reuel,PhD student of computer science,Stanford University,co-moderator32 Run of show,morning session Opening by Professor Robert Trager,moderator We

98、lcome by Doreen Bogdan-Martin,ITU Secretary-General AI Governance pulse check pre-discussion poll(see chapter 4.7)Multistakeholder panel:How can we move AI governance from principles to implementation?Panelists:H.E.Ms.Rose Pola Pricemou,Ministre,GUINEA(Ministre des Postes,des Tlcommunications et de

99、lEconomie Numrique)H.E.Mr.Dohyun Kang,Vice Minister,Ministry of Science and ICT,Republic of Korea Lane Dilg,Head of Strategic Partnerships,OpenAI Stuart Russell,Professor of Computer Science,University of California,Berkeley Rumman Chowdhury,CEO Humane Intelligence Tristan Harris,Director,Executive

100、Director of the Center for Humane Technology Daisy McGregor,Deputy Director,International AI Policy,Department for Science,Innovation and Technology,United Kingdom8AI Governance Day-From Principles to Implementation Moderator:Robert Trager,Professor,University of Oxford Update from the United Nation

101、s High-level Advisory Body on AI(HLAB)Carme Artigas,Co-chair of United Nations High-level Advisory Body on AI Amandeep Singh Gill,UN Secretary-Generals Envoy on Technology Breakouts Each table discussed one of three themes:Theme 1:What is the landscape of AI governance and how will it evolve?Theme 2

102、:How do we implement AI governance frameworks?Theme 3:How do we ensure inclusion and trust?Format of breakout session:20 tables with 10 people each.Each table had Ministers/Regulators UN representatives Industry leaders,Academia,Civil Society Expert Advisor Rapporteur(=Notetaker)The ITU Secretariat

103、made a seating plan which tried to balance geography and gender.Participants received their seating assignment as they came into the room.Participants were kindly asked to write on cards at their tables what they consider noteworthy/useful/innovative.About 150 cards were received and incorporated in

104、to this report.Coffee break Insights from breakouts Moderators:Robert Trager,Professor,University of Oxford Anka Reuel,PhD student,Stanford University AI Governance pulse check post-discussion poll(see Chapter 4.7)Same questions as in pre-discussion poll.Did opinions change?Closing Networking lunch3

105、3 Run of show,afternoon session 14:00 14:10 Welcome address,Doreen Bogdan-Martin 14:10 14:30 Insights and key findings on AI Governance Implementation 14:30 15:15 State of play of major global AI Governance processes 15:15 15:45 Leveraging the UK system to advance global AI Governance efforts 16:00

106、16:30 The Governments AI dilemma:how to maximize rewards while minimizing risks?16:30 17:00 The critical conversation on AI safety and risk 17:00 17:45 The dilemma of open source vs.proprietary Large Language Model 17:45 18:15 The role of AI standards as an implementation tool 18:15 18:25 From princ

107、iples to implementation pathways forward9AI Governance Day-From Principles to ImplementationFigure 5:Center Stage Welcome Address featuring Doreen Bogdan-Martin,Secretary-General,International Telecommunication Union(ITU)Figure 4:Roundtable participants during the morning breakout session10AI Govern

108、ance Day-From Principles to ImplementationFigure 6:The panel of speakers from a Center Stage discussion4 Morning sessions41 Welcome speech by Doreen Bogdan-Martin,ITU Secretary-GeneralIt is a tremendous pleasure to welcome you all to our first ever Governance Day and a series of roundtables.This is

109、kind of a different format for many of us.So,I am looking forward to having interactive,dynamic discussions today.It is inspiring to see so many senior leaders with us this morning,early morning,and ready to roll up their sleeves.And of course,we have representatives from governments,the private sec

110、tor,international organizations,academia,the technical community,and civil society,all here to discuss and help shape AI governance for the future.I am particularly pleased to have so many representatives from both developed and developing countries.This representation aligns perfectly with ITUs amb

111、ition to provide the most inclusive,diverse,neutral,and global platform for artificial intelligence.I am also pleased that this morning we will be linking up with Singapore.We will have a connection with the Secretary Generals co-chairs of the United Nations High Level Advisory Body on AI(HLAB).They

112、re meeting in Singapore this week,and we will hear from them live this morning.Their work,of course,is also very important as we look to the future and the global conversations around AI.Excellencies,ladies and gentlemen,this morning is really a sort of ground zero for AI governance innovation.These

113、 discussions will help us to get a better grasp of the current AI landscape,whos 11AI Governance Day-From Principles to Implementationdoing what,and where we are heading,how we can implement AI governance frameworks,and how we can ensure inclusion and trust are at the heart of all of our efforts.The

114、se are the three topics that were going to be discussing in the breakouts this morning,and I would like to briefly say a few words about each.The AI regulatory landscapeFirst,from the AI regulatory landscape.We have seen quite a fast response,in particular,over the past year,with governments activat

115、ing regional groups on the global level as well,starting to discuss this issue.But I think many are still contemplating what to do.The ITU actually conducted a landscaping survey among our 193 Member States,and we found that about 85%of our responding Member States hadnt yet put in place regulations

116、 or policies.Some are just beginning to think about these issues.But I think one important result of the survey was that it demonstrated that all countries are eager to learn.And so thats also part of the objective of the discussions this morning.Through dialogue,we can help to avoid fragmentation w

117、hen it comes to AI governance.Collaboration around areas of common interest and priorities comes next.A good example is standards,and were joined this morning by many of our standards friends.Standards are definitely central to this debate and,of course,to the entire summit.Its critical for all AI g

118、overnance initiatives.How to implement AI governance frameworks?That brings me to my second point,which is how to implement AI governance frameworks.From algorithm transparency to safety and security of systems,standards really serve as a prerequisite for the effective implementation of guardrails.S

119、imply put and I look to Gabriela Ramos from UNESCO the best ethical or human rights guidelines would be incomplete without being translated into actionable,enforceable technical standards.ITU already has more than 200 AI-related standards,either published or currently under development.Our standardi

120、zation process ensures that all voices are heard,including those from the developing world.But more standards are needed to address the pressing challenges around artificial intelligence.And,more importantly,we need to be developing them in a coordinated way,using established mechanisms like the Wor

121、ld Standards Cooperation.I recognize the leaders of ISO and IEC who are with us this morning.Inclusion and trustThats a natural segue to the third topic this morning,which is inclusion and trust.These two elements are deeply interconnected.Without trust,people will hesitate to engage with AI,potenti

122、ally creating yet another digital divide in an already unequal digital world.I think this risk is real.For the first time this year,adverse outcomes of AI actually entered into the top 10 rankings of the World Economic Forum Global Risks Perception Survey.Its important to note that not everyone feel

123、s the same way about AI.A recent survey carried out by BCG revealed that consumers in low and middle-income countries were actually much more excited than consumers in mature markets.Many see this as an opportunity,an opportunity to leapfrog technological gaps and accelerate innovation in vital area

124、s such as education and healthcare,and I would say all the SDGs.12AI Governance Day-From Principles to ImplementationExcellencies,ladies and gentlemen,the outcomes of our discussions today will resonate throughout this summit and beyond.I ask you this morning and throughout the day,to be ambitious.L

125、ets use AI to tackle our most pressing global challenges and achieve the SDGs.I would also ask you to be visionary.Lets make AI governance the cornerstone for a better digital future for all.But,above all,I dare you to have hope.These are not ordinary times.These are times when we need to show audac

126、ity,solidarity,and a shared sense of responsibility.Thank you very much.Its a really a great pleasure to welcome you all to our governance day and a series of roundtables.This is kind of a different format for many of us.So looking forward to having lots of interactive dynamic discussions today.42 H

127、ow can we move AI governance from principles to implementation?Multistakeholder panelPanelists:H.E.Ms.Rose Pola Pricemou,Ministre,Guinea(Ministre des Postes,des Tlcommunications et de lEconomie Numrique)H.E.Mr.Dohyun Kang,Vice Minister,Ministry of Science and ICT,Republic of Korea Lane Dilg,Head of

128、Strategic Partnerships,OpenAI Stuart Russell,Professor of Computer Science,University of California,Berkeley Rumman Chowdhury,CEO Humane Intelligence Tristan Harris,Director,Executive Director of the Center for Humane Technology Daisy McGregor,Deputy Director,International AI Policy,Department for S

129、cience,Innovation and Technology,United KingdomModerator:Robert Trager,Professor,University of OxfordFigure 7:Speakers from the Multistakeholder Panel:How can we move AI Governance from Principles to Implementation?13AI Governance Day-From Principles to ImplementationInvestment in AI science There i

130、s an urgent need for increased investment in AI science,surpassing the funding allocated to other scientific fields.The potential for AI to surpass human intelligence through Artificial General Intelligence requires preemptive measures to maintain control over these powerful entities before their wi

131、despread deployment.Regulatory frameworks and safety measures Historical precedents,i.e the uncontrolled risks of nuclear power illustrated by the Chernobyl disaster,underscore the importance of rigorous safety measures and oversight in AI development.AI applications must be restricted to sectors wh

132、ere safety can be unequivocally ensured,similarly to the stringent safety requirements in the pharmaceutical and nuclear industries.The principle of proving safety before deployment should be paramount.Current international and domestic policy frameworks are inadequate to address the rapidly evolvin

133、g AI landscape.There is a need for new institutions and regulatory tools to manage AI risks effectively.International cooperation and summits Past and upcoming international summits on AI governance and safety in the UK(2023),Korea(2024),and France(2025)are important in establishing global standards

134、 and interoperability of AI safety measures.These summits aim to foster cooperation among heads of state,government leaders,UN agencies,and other stakeholders to address the AI divide,promote inclusivity,and ensure that safety measures are integrated into AI development.Figure 8:H.E.Ms.Rose Pola Pri

135、cemou,Ministre,Guinea(Ministre des Postes,des Tlcommunications et de lEconomie Numrique)during the Multistakeholder Panel14AI Governance Day-From Principles to ImplementationFigure 9:Rumman Chowdhury,CEO,Humane Intelligence,during the Multistakeholder PanelChallenges in AI governance One of the prim

136、ary challenges in AI governance is the nascent understanding of AI models and their implications.Balancing the benefits and risks of AI requires a deep understanding of these technologies and their potential impacts.Interoperability at the international level is crucial to enable small companies to

137、scale across borders and participate in the global AI ecosystem.This requires detailed workstreams and ongoing collaboration among countries.Risks of the AI race The competitive race among AI companies often prioritizes market dominance over safety,leading to the use of unlawful shortcuts.Negative c

138、onsequences include the development of AI models that can hack into potential safety limits implemented.Governance frameworks must shift the incentives towards safety and responsible development to mitigate these risks.The historical failure to anticipate the externalities of social media serves as

139、a cautionary example.Preparedness and science-based governance Effective AI governance must be rooted in scientific principles,with preparedness work based on reliable scientific data.This includes leading in safety and capability while ensuring models are not prematurely released to the public.Ther

140、e is a need for standardized evaluations and methodologies to assess AI models,considering their probabilistic outputs and the challenges in creating consistent tests.15AI Governance Day-From Principles to ImplementationInclusivity and capacity building AI governance must address the specific needs

141、of developing countries,ensuring that AI solutions are tailored to local realities and cultural specifics.Inclusivity,for instance of young women in African societies,is vital for equitable AI development International cooperation on standards,pooling financing resources,and avoiding fragmented effo

142、rts are essential for effective capacity building and governance.Education and transparency Transparency in AI interactions is critical,including the right to know whether one is interacting with a human or a machine.The burden of proving the safe and ethical use of AI should rest on developers.Educ

143、ation on AI technologies and their implications should be a priority to ensure that society is adequately prepared for the changes brought by AI.Moving from summits to action Transitioning from discussions at summits to actionable plans through AI Safety Institutes and other mechanisms is crucial fo

144、r implementing effective AI governance.Emphasizing the definition and regulation of AI safety is necessary to address the risks associated with AI technologies.Global digital compacts and resolutions,such as the WTSA Resolution for the ITU to coordinate standards efforts within the UN,are steps towa

145、rds cohesive international governance frameworks.Figure 10:Tristan Harris,Director,Executive Director of the Center for Humane Technology,speaking during the Multistakeholder Panel16AI Governance Day-From Principles to ImplementationFigure 11:Stuart Russell,Professor of Computer Science,University o

146、f California,Berkeley,speaking during the Multistakeholder PanelFigure 12:Lane Dilg,Head of Strategic Partnerships,OpenAI,speaking during the Multistakeholder Panel17AI Governance Day-From Principles to Implementation43 Theme 1:What is the landscape of AI governance and how will it evolve?4.3.1 Exam

147、ples of multilateral and national AI governance frameworksGovernments around the world are starting to establish rules with respect to AI.Here is a list of multilateral and national initiatives as of the end of May 2024(the list is indicative and non-exhaustive).Table 1:Examples of national AI Gover

148、nance initiativesExamples of National AI Governance Initiatives1.BrazilAI Strategy(April 2021),Draft AI Bill(proceeding through Parliament May 2024)2.ChinaInternet Information Service Algorithmic Recommendation Manage-ment Provisions(March 2022),Provisions on the Administration of Deep Synthesis Int

149、ernet Information Services(November 2022),Measures for the Management of Generative AI Services(July 2023),Global AI Governance Initiative(Oct 2023)3.EstoniaNational AI Strategy(2022-23)4.FranceNational AI Strategy(April 2024)5.IndiaNational Strategy for AI(June 2018)Principles for Responsible AI(Fe

150、b 2021)6.JapanAI Guidelines for Business(April 2024)7.KazakhstanDigital Transformation Partnership(Oct 2023),AI in Government Services(Feb 2024)8.KenyaNational AI Strategy(April 2024)9.Republic of KoreaNational Strategy for AI(Oct 2019)10.RwandaNational AI Policy(April 2023)11.SingaporeProposed Mode

151、l AI Governance Framework for Generative AI(Jan 2024),AI Verify Platform(June 2023)12.UAENational Strategy for AI(Oct 2017)13.UKA Pro-Innovation Approach to AI(August 2023),Inspect Platform on AI Safety(May 2024)14.USAExecutive Order on the Safe,Secure,and Trustworthy Development and Use of AI(Oct 2

152、023)18AI Governance Day-From Principles to ImplementationTable 2:Examples of multilateral AI Governance initiativesExamples of Multilateral AI Governance Initiatives1.African UnionMultistakeholder consultations on a Continental Strategy on AI(April 2024),AUDA-NEPAD White Paper:Regulation and Respons

153、ible Adoption of AI in Africa(June 2023)2.AI Safety SummitThe Bletchley Declaration on AI Safety(Nov 2023)(UK host with follow-up by the Republic of Korea and France),Seoul Declaration(May 2024),France-China Joint Declaration on AI Gover-nance(May 2024)3.ASEANASEAN Guidance on AI Governance and Ethi

154、cs(Feb 2024)4.BRICSAgreement to establish a BRICS AI Study Group(Aug 2023)5.Council of EuropeFramework Convention on AI,Human Rights,Democracy and the Rule of Law(17 May 2024)6.Digital Cooperation OrganizationGenerative AI Shaping the Digital Economy(Jan 2024)7.European UnionEU AI ACT(April 2024 ver

155、sion of text)8.EU-US Trade and Technology CouncilJoint Statement(April 2024)Joint Roadmap(Dec 2022)9.Latin America and the CaribbeanThe Santiago Declaration to Promote Ethical AI(Oct 2023)10.G7 Hiroshima ProcessG7 International Guiding Principles and Code of Conduct for Organizations Developing Adva

156、nced AI Systems(Oct 2023)11.G20Endorsement of OECD Recommendations on AI(June 2019)and G20 AI principles(June 2019)CAIPD AI Statements(Dec 2023)12.GCCEthical Framework for AI in the work of Attorneys-General and Public Prosecutors(Oct 2023)13.Global Partnership on AIMultistakeholder Initiative hoste

157、d by OECD(June 2020)Ministerial declaration(December 2023)14.ITUAI for Good Summit 2024:AI Governance Day(May 2024)15.ISOTwo key AI standards agreed.AI Management system(Dec 2023),AI Guidance On Risk Manage-ment(Feb 2023)16.IEEERecommended Practice For Organizational Governance Of AI(Feb 2020),Algor

158、ithmic Bias Considerations(Sept 2023)17.League of Arab StatesArab AI Working Group to Develop an Arab AI Strategy(Feb 2021)18.Nordic Council of MinistersDeclaration on AI in the Nordic-Baltic Region(May 2018)19.Southern AfricaWindhoek Statement on AI in Southern Africa(Sept 2022)20.OECDRecommendatio

159、ns on AI(May 2019),OECD AI Policy Observatory(Feb 2020),OECD Frame-work For The Classification Of AI Systems(Feb 2022),AI Language Models Technological,Socio-Economic And Policy Considerations(April 2023)21.MERCOSURDeclaration on Principles of Human Rights in the field of AI(Nov 2023)22.Responsible

160、AI in the Military Domain Summit(REAIM)Call to Action(Netherlands and Republic of Korea led process)Political Declaration(US initia-tive)(Feb 2023)23.UNESCOUNESCO Recommendation on Ethics of AI(Nov 2021)24.UN High-Level Advisory Board on AIInterim Report on Governing AI for Humanity(Dec 2023)25.UN I

161、nter-Agency Working Group on AITerms of Reference(March 2021),UN System White Paper in AI Governance(May 2024)26.UN General AssemblyResolution A/78/256 Seizing the opportunities of safe,secure and trustworthy AI systems for sustainable development(March 2024)27.UN Security CouncilHigh-level debate o

162、n AI convened by UK(July 2023),Arria-formula meeting on AI,peace and security convened by UAE and Albania(Dec 2023)28.UN Human Rights CouncilResolution 42/15 The Right to Privacy in the Digital Age(7 Oct 2019)29.WHOEthics and governance of AI for health Guidance on large multi-modal models(Jan 2024)

163、19AI Governance Day-From Principles to Implementation4.3.2 The debate about national vs international AI governanceAs AI regulations start to take shape,questions arise as to which areas of AI governance should be addressed at the national or regional level,and which at the international level.Items

164、 to be regulated internationally might include:AI in warfare AI and Human Rights:with due consideration given to the guidance by UNESCO and the Human Rights Council in this regard.Data privacy and cross-border data flows Interoperability and standards:The International Telecommunication Union(ITU),I

165、nternational Organization for Standardization(ISO),and International Electrotechnical Commission(IEC)are collaborating through the World Standards Cooperation(WSC)framework to develop international AI standards.The ITU has published over 100 AI-related standards,with 120 more in development as of 20

166、24,multiple of them in collaboration with other UN agencies.ISO and IEC have formed the joint subcommittee ISO/IEC JTC 1/SC 42 to advance AI standardization,developing foundational standards,reference architectures,frameworks,and guidelines for trustworthy AI systems.Shared research resources,in par

167、ticular on AI Safety.Some countries have set up,or are considering setting up,an AI Safety Institute,for example:United Kingdom:The UK has established the AI Safety Institute,which is the first state-backed organisation focused on advanced AI safety for the public interest.The UK government publishe

168、d on 18 May 2024 an up-to-date,evidence-based International Scientific Report on the Safety of Advanced AI.United States:The US has created an AI Safety Institute within the National Institute of Standards and Technology(NIST),following the Executive Order on AI.Singapore:Singapore has established a

169、 Generative AI Evaluation Sandbox to bring together industry,academic,and non-profit actors to evaluate AI capabilities and risks.Canada:The Canadian government has included in its 2024 budget funds to create an AI Safety Institute of Canada to ensure the safe development and deployment of AI.Japan

170、established an AI safety institute in February 2024.The European AI office called for in the European AI Act will cover AI safety but has a wider scope than just safety and also include research,innovation,deployment aspects,and international engagement.Items to be regulated nationally might include

171、:Sector-specific AI applications:National regulations,reflecting local needs,values,and legal systems,can address the deployment of AI in sectors such as healthcare,transport,finance,manufacturing,human resources,critical infrastructure(gas,water,electricity),law enforcement,administration of justic

172、e,education,national employment.Consumer protectionThere may also be a hybrid approach,where international guidelines provide a broad framework while allowing for national or regional specificity.Intellectual Property Rights(IPR)in AI Interoperability and standards Environmental sustainability20AI G

173、overnance Day-From Principles to ImplementationThe United Nations System published on 2 May 2024 a UN System White Paper in AI Governance:An analysis of the UN systems institutional models,functions,and existing international normative frameworks applicable to AI governance.Potential discussion ques

174、tions What are the learnings that can be drawn multilateral and national AI governance efforts?Which AI standards and regulations need to be agreed upon at the international level?Within each category,with which priority should the various items be tackled?Which process should be used to determine t

175、he prioritization of the various items?What role should the UN take in international AI governance efforts?How should participation in international AI governance decision-making be structured?How do we address cross-border risks of AI systems?How can we safely share knowledge about what is working

176、and what is not working well?How can regulation adapt and keep pace with technological advancements?4.3.3 Insights from the breakout sessions:theme 1The landscape of AI governance is intricate and evolving,with various approaches and development stages across countries and regions.Key themes include

177、 human-centric development,leveraging existing frameworks,inclusion,global coordination,private sector involvement,and balancing governance with regulation.Human-centric AI development:AI development should prioritize human welfare and societal betterment,focusing on ethical principles and social go

178、od.Leveraging existing frameworks:using existing regulatory frameworks from industries like automotive,pharmaceuticals,and cybersecurity can streamline AI governance.International organizations like the UN can provide unified principles to prevent fragmentation.Inclusivity and capacity building:addr

179、essing biases in AI and ensuring inclusive models are essential.Enhancing AI capabilities in underrepresented regions,particularly in the Global South,involves improving data collection and usability.Multilateral and coordinated efforts:global coordination is crucial to prevent big tech companies fr

180、om setting standards.A multilateral approach is more effective than regional or national efforts alone.Reducing fragmentation and consolidating efforts within strong institutions is vital.National and local applicability:implementing international AI governance frameworks locally is challenging due

181、to different adoption levels and needs.Smaller countries need support to build necessary institutions.Private sector involvement:private companies,especially in healthcare,must incorporate value-based AI governance in their operations.The private sector is key in ethical AI development.Governance ve

182、rsus regulation:governance encompasses broader objectives than regulation.Initiatives by UNESCO and ITU on ethics and standards are important.Regulations must adapt quickly to keep pace with AI innovations.Addressing linguistic and cultural divides:global frameworks must be balanced with local conte

183、xtualization to ensure inclusive AI development.Resolving linguistic and cultural divides is crucial.Data governance:data governance is fundamental in AI governance to prevent digital colonization.Ensuring quality and accessible data and establishing cross-sector standards are necessary.21AI Governa

184、nce Day-From Principles to Implementation Cross-sector collaboration and standards:interoperability is essential for AI governance,ensuring sectors can keep up with advancements.Regional digital regulatory sandboxes allow collaborative testing and refining of AI frameworks.Current Landscape Early st

185、age regulation:many countries are just beginning to develop AI governance frameworks,with regulation often driven by companies rather than comprehensive policies.Global leadership and inclusive representation are needed.Country-specific approaches:countries take unique approaches based on their need

186、s and development stages.The EUs AI Act sets a significant example.Balancing innovation with regulation is a critical challenge.Inclusivity in regulation:every nation must contribute to inclusive AI regulations to prevent dominance by a few powerful countries or corporations.AI governance must consi

187、der cultural,ethical,and religious values.Global and local balance:a combination of global principles and local adaptations is needed.Embedding technical language and ethical considerations into policies is essential.Future evolution Global coordination:international bodies like the UN are expected

188、to establish global AI governance documents,balancing standards with local regulations.Ensuring developing countries participate in AI advancements is critical.Comprehensive frameworks:learning from existing frameworks,such as nuclear regulation,can help create robust AI governance structures.AI lit

189、eracy should include understanding AIs implications,ethics,and governance.Dynamic and adaptable regulations:regulations must be dynamic and adaptable to keep pace with AI innovations.A blend of global standards and local adaptations will ensure inclusive and equitable access to AI.Intersection of ci

190、vil society and industry Government lag and civil societys role:governments often lag in adapting to AI,with civil society remaining reactive.Fragmented approaches lack integration with data governance and cybersecurity.Bottom-up vs.top-down approaches:bottom-up approaches risk duplicating efforts,w

191、hile top-down approaches may lack detailed roadmaps.Establishing clear definitions for robustness and safety is crucial.Geopolitical approaches:different regions have distinct AI governance approaches:the EU focuses on rights,China on economic development,and the US on maintaining leadership.Identif

192、ying applications needing strict regulations is essential.Learnings from multilateral and national efforts Variance between countries:advanced countries have varying AI strategies,presenting challenges for Least Developed Countries(LDCs).Startups in developing countries often adopt AI rapidly withou

193、t sufficient scrutiny.Standardized benchmarks can guide AI adoption.Cross-border risks and regulation absence:lack of regulation across borders presents risks.Governance of high-risk AI applications,like Generative AI,is crucial.Existing standards from organizations like WHO and ISO can provide reso

194、urces.Governance models and international cooperation UN values and principles:existing legal instruments for AI regulation should be leveraged.Increasing global awareness and establishing regional AI innovation centers are essential.22AI Governance Day-From Principles to Implementation Government r

195、oles and responsibilities Cabinet-level oversight:there is debate on whether AI should have dedicated oversight or be integrated into existing frameworks.Ensuring government responsibility for AI harms and mandating transparency is critical.Cybersecurity and AI:the growth of cyberattacks underscores

196、 AIs role as a defense system.Addressing severe,state-sponsored attacks and determining responsibility is necessary.44 Theme 2:how do we implement AI governance frameworks?4.4.1 Lagging laws,lagging tech:the AI governance paradoxIn the rapidly evolving landscape of AI,a paradox has emerged:while reg

197、ulation is often seen as lagging behind technological advancements,there is an equally critical yet less recognized issue where technology fails to keep up with regulatory demands.While existing regulatory frameworks struggle to adapt to the pace of AI innovation,simultaneously,the current state of

198、technology and tools available does not allow for monitoring,checking,and controlling AI systems.This gap poses risks and underscores the need for advancing tools capable of ensuring effective governance.AI governance includes the governance of data,the governance of algorithms,and the governance of

199、 computing resources(compute).Governance of data:The governance of data refers to the policies,procedures,and standards necessary to manage the lifecycle of data within AI systems.Data governance ensures data quality,integrity,and security,which are essential for the reliable operation of AI technol

200、ogies.This includes establishing robust protocols for data collection,storage,and sharing,and implementing privacy and security measures to protect sensitive information.Data governance frameworks must address issues such as consent,transparency,and accountability to maintain public trust and comply

201、 with regulatory requirements.Moreover,data governance plays a crucial role in mitigating biases in AI systems by promoting diverse and representative datasets,thereby enhancing the fairness and accuracy of AI outcomes.Governance of algorithms:The governance of algorithms focuses on the ethical and

202、responsible development,deployment,and oversight of AI models and their decision-making processes.This aspect of AI governance aims to ensure that algorithms operate transparently,fairly,and without discrimination.It involves creating standards and guidelines for algorithmic accountability,which inc

203、lude regular audits,performance evaluations,and the ability to explain AI decisions to stakeholders.Algorithmic governance also emphasizes the importance of ethical considerations,such as avoiding unintended harmful consequences and ensuring that AI applications align with societal values.Governance

204、 of compute:The governance of compute addresses the management and oversight of computational resources required to develop,train,and deploy AI systems.As AI models become increasingly complex and resource-intensive,the need for sustainable and equitable access to computing power grows.Compute gover

205、nance involves setting policies for the efficient and fair allocation of computational resources,ensuring that these resources are used responsibly and do not disproportionately favor well-resourced entities over smaller or less funded organizations.Additionally,it includes considerations for enviro

206、nmental sustainability,as the energy consumption of AI training and operations has a significant ecological impact.By implementing strategies to optimize energy use and reduce carbon footprints,compute governance aims to balance the advancement of AI with the imperative of environmental stewardship.

207、23AI Governance Day-From Principles to ImplementationWhere technology lags behind AI governanceIn the realm of AI governance,several areas reveal a significant lag in technological advancements needed to meet regulatory requirements.These gaps span across data management,computational oversight,mode

208、l integrity,and deployment practices,highlighting the urgent need for innovation and robust methodologies to support effective governance.Data Methods are lacking to clearly define and implement fairness in AI systems.There are no robust ways to fully safeguard privacy,especially in large language m

209、odels.It is difficult for users to know if their data was used to train a model and its influence on model behavior.Compute There are only nascent methods for verifying if a large model is being trained in order to enforce compute-based reporting requirements.Model Comprehensive and reliable methods

210、 to evaluate AI systems for safety,ethics,and reliability are lacking.Securing model weights from unauthorized access is challenging.Deployment Approaches to constrain model outputs to certain values are imperfect and can be circumvented.Detecting AI-generated content,especially non-watermarked cont

211、ent,remains difficult.4.4.2 Compute governanceCompute or computing power(often used interchangeably)refers to the necessary computational resources required to run software.Compute governance is arguably an area where measurements can be more straightforward and quantifiable.The training process of

212、AI models requires intensive compute,puting resources.and training a leading AI model tends to take months on a complex computing infrastructure involving specialized computer chips.In contrast,running a single inference query(e.g.having an AI model respond to a single question)requires much less co

213、mpute,but the total amount of compute used for inference is still very large,since large AI companies need to run millions of user queries per day.Compute is essential not only for training AI models,but also for deploying(i.e.operating)them.Just as operating expenses outpace initial fixed costs for

214、 many large-scale projects,the majority of available AI compute resources are used for operating AI models rather than training them.Frontier AI models are so large that they cannot be efficiently operated at large-scale with household amounts of typical consumer hardware.Instead,for models in high

215、demand,inference requires thousands of AI chips housed in specialized data centers to adequately serve the needs of thousands of users.The wider the deployment of AI systems(which requires more compute),the more impact they will likely have(both beneficial and harmful).24AI Governance Day-From Princ

216、iples to ImplementationCompute governance,the setting of rules on computing resources to achieve governance,can be an attractive tool for AI governance.This is because compute is detectable and quantifiable,allowing for effective monitoring and control.For example,energy-intensive,specialized data c

217、enter infrastructure is an indicator of compute activity.In contrast,while data and algorithms are also essential ingredients of AI,it is much more challenging for governments to quantify them.Many are of the opinion that using compute providers(e.g.Microsoft Azure,Amazon Web Services(AWS),Apple,Byt

218、edance,Meta,Oracle,Tencent,and Google Cloud)as intermediary regulators would be most effective in addressing risks associated with large-scale AI training to prevent bad actors from training advanced AI models,rather than addressing all AI-related risks.This is because non-compute-intensive AI model

219、s are often feasible to train and run on widely available customer hardware,so cloud providers have less ability to oversee such activities.Compute providers can therefore play an essential role in AI governance via four key functions:Securers:protecting AI systems and critical infrastructure Record

220、 keepers:improving transparency for regulators Verifiers:monitoring customer activities Enforcers:taking actions against breaches of rulesInternational cooperation is essential to handle cross-border supervision and data challenges(e.g.ensuring that personal data is protected according to different

221、regional standards and regulations),as it reduces the risk of compute providers and AI developers moving to jurisdictions with less regulatory oversight.In addition to its potential role in regulation,compute has the potential to advance international cooperation on AI,by enabling states and compani

222、es to demonstrate their adherence to their commitments without leaking sensitive data.States may be able to show that approximately all of their AI compute was used consistently with their commitments,meaning significant compute would not have been available for other purposes.These approaches could

223、 leverage Privacy-Enhancing Technologies(PETs)to enable assurance while preserving confidential data.Potential discussion questions How does compute governance differ from data governance and algorithm governance?Are there real-world examples of effective compute governance?How can compute resources

224、 be effectively monitored and controlled to ensure compliance with governance policies?How can compute providers improve transparency for regulators and stakeholders?What are the potential frameworks for international cooperation on compute governance?What are the potential risks of over-regulation,

225、and how can they be mitigated?How might compute governance evolve with advancements in AI and computing technologies?4.4.3 Insights from the breakout sessions:theme 2Implementing an AI governance framework involves addressing challenges across data,compute,models,and deployment.Here is a direct and

226、concise approach summarizing the key points and elements discussed to implement an adequate framework.25AI Governance Day-From Principles to Implementation Inclusive and Collaborative Governance Stakeholder Involvement:Include government,private sector,NGOs,scientists,and academia;create an internat

227、ional AI Governance Agency with an advisory board of experts and citizens.Education and Legislation:Integrate AI education into schools;develop AI laws and policies at national and international levels.Regulation and Enforcement Global Standards:Cooperate internationally to set common standards;esta

228、blish national institutes for ongoing regulatory discussions.Decentralized Decision-Making:Empower specific units for quicker regulatory decisions.Data Governance Fairness:Develop and integrate standardized fairness metrics.Privacy:Implement differential privacy and federated learning;set regulatory

229、 standards for privacy.Transparency:Mandate clear data usage policies.Compute Governance Verification:Create reporting standards for compute usage.Monitoring:Use compute providers for real-time monitoring and reporting.Model Governance Evaluation:Standardize safety,ethics,and reliability evaluations

230、;establish third-party certification bodies.Security:Implement encryption and secure access protocols for models.Deployment Governance Output Control:Develop systems to monitor and constrain harmful outputs.Detection:Invest in tools to detect AI-generated content.Interoperability and Common Definiti

231、ons Technical Standards:Ensure systems are interoperable globally;establish common definitions for software,hardware,data,and resources.Enforcement Models:Create audit models for balance between innovation and control.Practical Implementation and Testing Testing Tools:Develop practical testing tools

232、 for AI systems;propose certification frameworks.Continuous Testing:Emphasize ongoing testing for AI safety and efficacy.Monitoring and Enforcement Frameworks:Develop interoperable assessment frameworks considering diverse regulatory capacities and application levels.26AI Governance Day-From Princip

233、les to Implementation Standards and Global Processes Standards Organizations:Leverage existing standards organizations like IEC,ISO,and ITU;use established processes to develop AI standards.Balancing Risks and Opportunities Risk Management:Balance AI risks and opportunities,increase investment in AI

234、 safety research.Proactive Measures:Ensure AI safety before deployment;involve government to balance safety and competition.Education Investment:Invest in AI education and mental health;promote public understanding and ethical awareness of AI.Governance Pace Acceleration:Speed up AI governance to ma

235、tch technological advances;establish liability frameworks and encourage whistleblowers.Adaptability:Develop adaptable legislative frameworks;encourage open-source tools for AI testing and safety.Transparency and Accountability User Awareness:Ensure users know when they interact with AI;place proof o

236、f safety on AI developers.Incentives:Link AI governance to incentives;include diverse perspectives in AI development.AI Safety Institute Establishment:Create an AI Safety Institute for enforcing safety and transparency standards;invest in interoperability to aid integration.Flexible and Responsive L

237、egislation Adaptability:Develop adaptable legislative frameworks;encourage open-source tools for AI testing and safety.Standards and Certification:Implement ubiquitous standards;offer third-party and self-certification.Risk Management Representative Data:Ensure unbiased data sets;invest in computing

238、 power,including quantum computing.Mitigating Impact:Create indemnity and risk mitigation strategies.45 Theme 3:how do we ensure inclusion and trust?4.5.1 Inclusion connecting the unconnectedITU has been providing statistics on global connectivity,particularly in terms of Internet access,for many ye

239、ars.27AI Governance Day-From Principles to ImplementationFigure 13:Individuals using the Internet/Percentage of population vs timeFor the year 2023,global Internet usage is up to approximately 67%of the worlds population(8+billion people),or 5.4 billion people,marking a 4.7%increase since 2022,up fr

240、om the previous years growth rate of 3.5%.The offline population in 2023 decreased to an estimated 2.6 billion,about 33%of the global total.70%of men worldwide use the Internet,compared to 65%of women.This results in 244 million more men than women using the Internet as of 2023.Gender disparity is s

241、till a distant prospect in regions with low Internet use.4.5.2 Giving voice for the global majorityMany people and many countries that are being impacted by AI are not integrated in ongoing discussions to set AI policies.Many share the concern that foreign companies supplying AI systems might captur

242、e a significant part of their economic value and knowledge.Since AI shapes our perception of the world,how can cultures be preserved if tools are built on foreign references?This issue is even more pressing for less digitized societies and non-English-speaking countries,as generative AI models are m

243、ainly trained on English data and perform best in that language.Potential discussion questions How do we reach inclusive participation from individuals representing diverse backgrounds,cultures,and regions in AI research,development,and decision-making processes?How do we enhance access to AI resour

244、ces to achieve sustainable development goals?How do we address bias and discrimination in AI algorithms and systems to prevent inequalities and to ensure fair outcomes for all?How do we provide capacity building,education,and skill development in AI among not represented,or underrepresented,communit

245、ies to enhance their participation in and contribution to the field?28AI Governance Day-From Principles to Implementation How do we uphold human rights and ethical considerations in AI development such as transparency,accountability,privacy and others for all individuals globally?How do we have the

246、global majority take part in shaping the future of AI technologies and AI governance?4.5.3 Open vs closed sourcing of(generative)AI modelsThe debate surrounding the open-vs closed-sourcing of increasingly capable AI models highlights the tension between the benefits of transparency,external oversigh

247、t and rapid innovation against the potential risks of misuses and unintended consequences.Open-sourcing AI models for our purposes means that the AI model architecture and the associated weights are freely and publicly available to anyone to use or modify.Closed,or proprietary,refers to AI systems,p

248、articularly foundation models,where the underlying algorithms,model architecture,datasets,and training methodologies are kept confidential by the entity that developed them.Arguments in favor of open sourcing of(generative)AI models:Promotes innovation:By making the models public,a broader range of

249、developers can contribute to and enhance the technology.Increases transparency:Open-sourcing allows for community auditing of the models,which can lead to improvements in model safety and ethics.More eyes spot more bugs.Fosters collaboration:Open-source AI systems encourage collaboration among resea

250、rchers and developers,leading to faster innovation and improvement of technology.Increases access:Open-source AI systems are more accessible to a wider range of users,including those from resource-constrained environments,promoting inclusivity and access to AI technology.This can reduce the knowledg

251、e and resource gap between large corporations and smaller entities,counteracting the centralization of power in AI companies.Prevents vendor lock-in:Users of closed-source AI systems may become dependent on a single vendor,making it difficult and costly to switch to alternative solutions in the futu

252、re.Reduces monopolistic practices:Provides opportunities for smaller entities to participate in and benefit from advanced AI without the prohibitive costs of developing proprietary models.Arguments against open sourcing of(generative)AI models:Security risks:There is an increased risk of misuse as m

253、ore actors can access powerful AI tools,potentially leading to harmful applications.Quality control:It may be challenging to maintain high standards of quality and reliability when control over modifications is decentralized.Irreversibility:If a model has been released with a flaw that would allow g

254、rave misuse,or inherent safety risks of the model,there is no straightforward way to prevent someone from continuing to use the model or to ensure that users install patches to fix the model.Regulatory challenges:Open-source models could complicate efforts to enforce compliance with ethical standard

255、s and legal regulations due to their widespread and uncontrolled distribution.Potential Discussion Questions How has open-sourcing of AI models benefited startups,academic institutions,and developers in resource-constrained environments?Are we seeing the advantages/disadvantages from already open-or

256、 closed-sourced models?29AI Governance Day-From Principles to Implementation Are there other ways of pursuing open-source objectives while keeping models closed-source?How do we decide if models should be open-or closed-sourced?4.5.4 Insights from the breakout sessions:theme 3The following outlines

257、key strategies and actions essential for ensuring inclusion and trust in AI.Differentiation and guidelines Develop guidelines to distinguish between AGI and specific AI models.Tailored oversight and regulation are essential for managing each AI type effectively.Incentives and policies Implement comp

258、ute taxes and investment requirements to incentivize businesses to prioritize inclusivity and trustworthiness in AI development.Transparency and Oversight Establish sandbox environments for AI testing with oversight from civil society,academia,and public/private sectors.Ensure testing processes and

259、outcomes are transparent to build community trust.Certification and ethical standards Create sector-specific certifications with diverse global input.Facilitate international dialogue to establish ethical guidelines that respect local regulations and cultural differences.Education and connectivity P

260、romote digital literacy and expand connectivity to underserved communities.Support global participation in AI by improving digital skills and access to AI resources.Governance and regulatory frameworks Develop clear ethical guidelines and regulatory frameworks with global consensus.Ensure governance

261、 processes are open and inclusive,allowing for public input and scrutiny.Addressing bias Urgently address biases in AI algorithms to ensure fairness and prevent inequalities.Open Ecosystem and Collaboration Encourage an open AI ecosystem to foster innovation and trust.Promote collaboration between d

262、evelopers,governments,and NGOs.Data hubs and connectivity Develop data hubs and improve connectivity,especially in developing regions,to support AI infrastructure and access.Social dialogue and diverse perspectives Facilitate regular discussions between stakeholders to understand and adopt AI.Ensure

263、 diverse perspectives are included in AI development and decision-making processes.30AI Governance Day-From Principles to Implementation Transparency and accountability Make AI systems transparent and accountable.Users should be informed when AI is involved in decisions.Cybersecurity and surveillanc

264、e Protect AI systems from cyber threats.Balance AI surveillance benefits with privacy and civil liberties.46 Update from United Nations high-level advisory body on AI Reporting live from Singapore where the United Nations High-level Advisory Body on AI(HLAB)met for its third and final meeting in Sin

265、gapore,28-29 May 2024,were:Carme Artigas,Co-Chair of United Nations High-level Advisory Body on Artificial Intelligence Amandeep Singh Gill,UN Secretary-Generals Envoy on TechnologyCarme Artigas,one of the two co-chairs of the United Nations High-level Advisory Body on AI,recalled that the AI Adviso

266、ry Body has been working since last October to address the global governance deficit,especially concerning inclusivity and diversity.The Advisory Body published an interim report in December 2023,and its final report was aimed to be published in the early summer of 2024.She noted that the Advisory B

267、ody had conducted more than 40 consultations involving more than 1000 experts and some 20 deep dives into thematic areas.Carme Artigas highlighted the importance of increasing developing countries capacities to benefit from AI and the need for reliable,timely scientific information for policymakers.

268、Ms Artigas outlined six key conclusions from recent meetings:1)providing reliable,scientific and timely information for policy makers on AI capabilities,opportunities and risks;2)linking existing governance efforts through regular dialogue;3)promoting international cooperation around standards;4)sup

269、porting the global South to avoid an AI divide;5)pooling financial resources,expertise,data and compute to leverage AI for the Sustainable Development Goals;6)establishing a coordinating mechanism to avoid fragmented efforts and support a safe,inclusive AI future.“I want to share what were feeling h

270、ere:there is a momentum,and the momentum is now.This is the moment when we need to connect all global efforts and take bold decisions because what we are seeing when were talking about governance is not a hindrance,it is a capacity,an enabler for AI development for good.”(Carme Artigas)Amandeep Sing

271、h Gill,the UN Secretary-Generals Envoy on Technology,added that the upcoming United Nations Summit of the Future(22-23 September 2024)offers a chance to include AI governance in a global digital compact.31AI Governance Day-From Principles to Implementation“There is a concentration of AI capacity in

272、some areas,and we need to address this urgently so that some of the harms that were highlighted during the panel discussion can be addressed,and we can direct the arc of investments,political interest,and talent in the direction of AI for the SDGs.”(Amandeep Gill)Mr Gill said that the United Nations

273、 were ready to assist the international community in harnessing AIs benefits while tackling its risks,ready to help building capacity to reinforce the vital work on standards,to help producing assessments that assist policymakers in the public and private sectors,and to assist researchers and scient

274、ists in making the right decisions to improve the quality of their decision-making.Figure 14:Carme Artigas,Co-Chair of United Nations High-level Advisory Body on Artificial Intelligence,reporting live from Singapore47 Governance pulse checkA real-time poll was taken among the participants at the beg

275、inning of the morning session(pre-pulse check)and at the end of the morning session(post-pulse check).Five statements related to AI in general and five on AI governance were presented to the participants.Participants were asked:Do you agree with these statements?.Answers could be given on a scale fr

276、om strongly disagree to strongly agree.Please note also the distribution curves overlaid for each of the ten statements.32AI Governance Day-From Principles to ImplementationFigure 15:Pre-pulse check on whether participants strongly disagreed(value of 1)or strongly agreed(value of 5)with statements r

277、egarding AI in general.The scale was from 1 to 5.Pre-pulse checkPost-pulse checkFigure 16:Pre-pulse check on whether participants strongly disagreed(value of 0)or strongly agreed(value of 5)with statements regarding AI governance.The scale was from 0 to 5.Pre-pulse checkPost-pulse check33AI Governan

278、ce Day-From Principles to Implementation5 Public afternoon session51 Welcome speech,Doreen Bogdan-Martin,ITU Secretary-GeneralGood afternoon,everyone.Welcome to Day Zero of the AI for Good Global Summit.Our eagerly anticipated Governance Day is off to a running start.Weve already put our AI experts

279、and government leaders to work this morning.Weve spent the entire morning exchanging ideas on three critical topics:surveying the AI landscape and understanding how it might evolve,looking at how to implement AI governance frameworks,and,perhaps most importantly,discussing how we can ensure inclusio

280、n and trust as we implement those frameworks.This morning,we heard about various governance efforts,the areas they have in common,as well as some of their differences.Crucially,we learned from developing countries because we want to ensure that they are not left out of the process.This challenges th

281、e argument that governments lack initiative when it comes to tech regulation.In just a few moments,youll be hearing from some of our amazing roundtable participants who will be sharing the outcomes of their work.But first,let me tell you why were doing this.Why are we here today?What is AI Governanc

282、e Day all about,and why are we at the ITU going to keep doing it?As many of you know,ITU is the UN agency for digital technologies,and we have been working to harness AI for good for the past seven years.Weve been convening the UN system around AI,and weve been co-leading an interagency coordination

283、 mechanism with UNESCO since 2021.Through our AI for Good platform,a multi-stakeholder community of 28 000 people from over 180 countries,our focus has been on putting artificial intelligence at the service of the Sustainable Development Goals.Thats been our compass.Whats new is this much sharper,st

284、ronger focus on governance.Its not the benefits but the risks of artificial intelligence that keep us all awake at night.Much has been said about AI governance in the media,academic circles,startups,tech giants,and from local governments all the way to the United Nations,which recently adopted a his

285、toric resolution recognizing AIs potential to advance the SDGs.Governance and technology we have been here beforeLadies and gentlemen,at the heart of all of this is a conundrum:how do we govern a technology if we dont yet know its full potential?There is no one answer to that question,but we do know

286、 that we have been here before.Twenty years ago,the Internet was met with a similar mix of shock,awe,and skepticism.It raised the same questions about how our economies,societies,and environment would transform for better and for worse.Were still grappling with those questions two decades later.In f

287、act,we still dont know the full potential of the internet because a third of humanity has never connected.But before we could even realize the potential,generative AI came along.Yet even with the convergence of these world-changing,interdependent technologies,governance efforts have emerged.They may

288、 not be perfect,but were not starting from scratch.The Internet Governance Forum and the WSIS Forum were born out of the World Summit on the Information Society.Some of you,like me,were there when this all happened 20 years 34AI Governance Day-From Principles to Implementationago.I remember how then

289、,as now,we didnt even have the vocabulary to describe what we were dealing with,but that didnt stop us from moving forward.What weve learned from the WSIS(World Summit on the Information Society)process is that we can take steps toward governance even if were building the plane as we fly it.We can c

290、ome together as a community,share experiences,practices,lessons learned,barriers,and challenges,knowing that once more,there is no one-size-fits-all approach to balancing benefits and reconciling regulatory risks.We must look at governance from many different angles and the only way forward is throu

291、gh a multi-stakeholder approach.Thats why Im so glad that today,gathered in this room,we have our WSIS community with us.Welcome to the WSIS community.We hope that you will help guide us through these many complex questions and challenges.After listening closely to this mornings discussions,I believ

292、e there are three key pieces that must be part of any AI governance effort.Development of technical standardsThe first piece,and obviously very relevant to the ITU,is the development of technical standards.As we heard this morning,those working on AI governance already recognize how technical standa

293、rds can help implement effective guardrails and support interoperability.This is where ITU has such a key role to play as an international standards development organization.We already have over 200 AI-related standards that we have either developed or are in the process of developing.As part of the

294、 World Standards Cooperation,a high-level collaboration between IEC,ISO,and ITU,we are advancing the development of global standards that can make AI systems more transparent,explainable,reliable,and secure.This provides certainty in the market and eases innovation for both large and small industry

295、players everywhere,including in developing countries.Putting core UN values at the heart of AI governanceThe second element is putting human rights,inclusion,and other core UN values at the heart of AI governance.All stakeholders deserve a voice in shaping AIs present and future.But who can afford t

296、he compute resources that go into producing AI applications?Who is on the teams that design the foundational models?Right now,the power of AI is concentrated in the hands of a few.This is risky and ethically precarious for humanity.We must work towards an inclusive environment where diverse perspect

297、ives,including those on gender,are reflected in policies that align with UN values.International AI governance efforts must account for the needs of all countries,and thats why the United Nations,together with governments,companies,academics,civil society,and the technical community,must play a key

298、role in ensuring that power is distributed equitably.This will not happen automatically.Development through capacity buildingThe third element is inclusive development through capacity building.ITU has a long history of bringing the voices of the global south to the emerging technology table.Part of

299、 this means making sure that every workforce in the world can deal with the challenges and risks brought about by artificial intelligence.Thats why weve been integrating AI capacity support in our digital transformation offerings.Well continue to roll out these initiatives with many of our UN 35AI G

300、overnance Day-From Principles to Implementationpartners,including UNDP,focusing on countries with low technological capabilities to help upskill them no matter where they are in their AI journey.Are you AI ready?Ladies and gentlemen,governance is not a given.An AI readiness survey that ITU recently

301、conducted among its 193 member states demonstrated that a majority of our responding Member States,actually 85%,dont have any AI regulations or policies in place.Today,some might at least start thinking about the policy elements and what to do next.This makes the work were doing today and beyond abs

302、olutely fundamental and essential.All good governance starts with listening listening to experts,exchanging ideas and experiences with peers,identifying gaps,and building on potential areas of convergence.Governance is never a one-and-done process;it is an iterative,sometimes frustratingly slow,but

303、ultimately necessary multi-stakeholder process.Taking stock of the landscape and facilitating deep discussions,as we did this morning,is the first step in transforming principles into practical implementation.Implementation,ladies and gentlemen,is what today is all about.I know everyone in this room

304、 has a stake in seeing AI used as a force for good in this world.As we heard from the UN Secretary-Generals high-level advisory body,we need to take bold decisions and view governance not as an inhibitor but as an enabler an enabler for AI for Good.Thats why today,Im calling on all of you to get inv

305、olved,take action,and participate actively in the AI governance activities happening here at the ITU.Lets harness the power of this AI community to govern AI with and for the world.Lets show them what it looks like,how its done,and lets do it together.Thank you very much.52 Leaders speak:Insights an

306、d key findings on AI governance implementationPanelists:H.E.Ms.Emma Inamutila Theofelus,Minister of Information and Communication Technology,Namibia H.E.Mr.Mauricio Lizcano,Minister of Information Technologies and Communications,Colombia H.E.Mr.Zunaid Ahmed Palak,Minister of State for the Ministry o

307、f Posts,Telecommunications and Information Technology,BangladeshModerator:Robert F.Trager,Professor at the University of OxfordThe three ministers summarized the discussions in the morning with 200 attendees.The panel discussion underscored the importance of a coordinated,inclusive,and human-centric

308、 approach to AI governance.The ministers highlighted both the opportunities and challenges that lie ahead in the realm of AI.As Minister Palak aptly quoted his Prime Minister,If you want to go fast and far,innovate together.This sentiment encapsulates the spirit of global cooperation 36AI Governance

309、 Day-From Principles to Implementationneeded to navigate the complexities of AI governance and ensure that its benefits are shared equitably across the world.Figure 17:The panelists and moderator from the Center Stage session Leaders speak:Insights and key findings on AI governance implementationMin

310、ister Theofelus,NamibiaMinister Emma Theofelus kicked off the discussion by highlighting the productive nature of the mornings conversations.She outlined several critical areas of focus Global coordination:The necessity of coordinating diverse efforts at a global level to reduce fragmentation among

311、UN agencies,governments,and regional bodies.Governance is surely broader than regulation,and therefore we need to reduce the fragmentation to ensure that we have standards that we can all comply with around AI governance.(Emma Inamutila Theofelus)Human-centric AI:Ensuring that AI remains focused on

312、human rights and maintaining a human element in its applications.Leveraging existing instruments:Building on existing frameworks and regulations,rather than reinventing the wheel.We dont necessarily need to create or build new institutions,we can already build on existing capacities and existing ins

313、titutions.(Emma Inamutila Theofelus)Inclusive data governance:Addressing data governance as a fundamental step to inclusive AI governance.37AI Governance Day-From Principles to ImplementationWe need everybody around the table,whether its global north or south,to ensure that nobody is left behind.(Em

314、ma Inamutila Theofelus)Local applicability:Considering national capacities,cultural,and linguistic differences to avoid biases in AI systems.Keeping up with technology:Involving AI developers in governance discussions to keep pace with the fast-evolving nature of AI.Balancing regulation and innovati

315、on:Ensuring that regulation does not stifle innovation but instead sparks it.We must strike a balance between regulation and innovation and ensure that regulation actually sparks innovation(Emma Inamutila Theofelus)Parallel discussions:Promoting continuous dialogue among all stakeholders to avoid is

316、olated discussions between producers and users of AI.Minister Lizcano,ColombiaMinister Mauricio Lizcano of Colombia provided a reflection on the risks and opportunities AI presents.He stressed the importance of human-centric AI and the need for global cooperation.Minister Lizcano highlighted two maj

317、or risks:Human-serving AI:Ensuring that AI serves humanity and not the other way around.Competition:Managing the competition between companies and countries to prioritize safety over winning the AI race.He called for a democratization of AI,ensuring that it is not limited to a select few but is acce

318、ssible and understandable to all.AI doesnt have to be a discussion among elite people we need real people to understand what is happening with AI.(Mauricio Lizcano)Mr.Minister Lizcano also shared Colombias ambitious initiatives to democratize AI through education and capacity building.In Colombia,we

319、 are training secondary students in machine learning and python coding creating the first faculty of AI in Latin America1.(Mauricio Lizcano)1 Colombias government to contribute 50 billion pesos to new AI center in Bogot(The Bogota Post Article).38AI Governance Day-From Principles to ImplementationFi

320、gure 18:H.E.Ms.Emma Inamutila Theofelus,Minister of Information and Communication Technology,NamibiaFigure 19:H.E.Mr.Mauricio Lizcano,Minister of Information Technologies and Communications,ColombiaMr.Minister Palak,BangladeshMinister Zunaid Ahmed Palak of Bangladesh introduced the development of G

321、Brain,a government AI aimed at enhancing service delivery.39AI Governance Day-From Principles to ImplementationOur most ambitious initiative in Bangladesh is the development of government brain,in short G Brain.(Zunaid Ahmed Palak)Minister Palak recounted a recent demonstration where a startup creat

322、ed his digital twin using AI,showcasing AIs potential in education and training.This experience was both fascinating and eye-opening.(Zunaid Ahmed Palak)Figure 20:H.E.Mr.Zunaid Ahmed Palak,Minister of State for the Ministry of Posts,Telecommunications and Information Technology,BangladeshHowever,Min

323、ister Palak also warned of the dangers of AI-generated misinformation,citing a recent incident involving a fake photograph of a cyclones aftermath.This incident highlights a serious risk associated with AI defects,spray of fake information and the consequences of which can be dire,affecting social h

324、armony,public safety and the credibility of our institutions(Zunaid Ahmed Palak)He called for robust measures to counteract the misuse of AI,emphasizing the need for ethical guidelines and regulatory frameworks.40AI Governance Day-From Principles to ImplementationWe must also develop robust measures

325、 to safeguard against this misuse through ethical guidelines and regulatory frameworks to ensure AI is used responsibly(Zunaid Ahmed Palak)53 State of play of major global AI governance processesPanelists:H.E.Mr.Hiroshi Yoshida,Vice-Minister for Policy Coordination,Ministry of Internal Affairs and C

326、ommunications,Japan Thomas Schneider,Ambassador and Director of International Affairs,Swiss Federal Office of Communications,and Chair of the Council of Europe Committee on AI H.E.Mr.Dohyun Kang,Vice Minister,Ministry of Science and ICT,Republic of Korea SHAN Zhongde,Vice Minister,Ministry of Indust

327、ry and Information Technology,Peoples Republic of China Alan Davidson,Assistant Secretary of Commerce for Communications and Information,Head of the National Telecommunications and Information Administration(NTIA)Juha Heikkil,Adviser for Artificial Intelligence,European Commission.Moderator:Ebtesam

328、Almazrouei,Founder and CEO of AI E3,United Arab EmiratesFigure 21:The panelists and moderator of the Center Stage session:State of play of major global AI Governance processes41AI Governance Day-From Principles to ImplementationMr.Ambassador Thomas Schneider on the Council of Europe treatyMr.Ambassa

329、dor Thomas Schneider highlighted the complexity of AI governance and the necessity for a context-based approach.He drew parallels to engine regulation,emphasizing that AI should not be regulated as a monolith but in terms of its functions and impacts.We regulate the people that are driving the engin

330、es.We regulate the infrastructure.We regulate or protect people affected.Its all context-based.Its not the engine;its the function of the engine,the effect of the tool.The same logic should be applied to AI.(Thomas Schneider)He also spoke about the Council of Europe treaty,designed to ensure that ex

331、isting human rights protections apply to AI.The Council of Europe treaty was drafted not in a spirit to create new human rights,not to reinvent the wheel,but actually to make sure that the existing human rights and protections for democracy and the rule of law are applied in the context where AI is

332、used.(Thomas Schneider)Figure 22:Thomas Schneider,Ambassador and Director of International Affairs,Swiss Federal Office of Communications,and Chair of the Council of Europe Committee on AI42AI Governance Day-From Principles to ImplementationMr.Juha Heikkil on the European AI actMr.Juha Heikkil provi

333、ded an overview of the European AI Act,described as the first comprehensive,horizontal,and legally binding regulation globally.He detailed the phased implementation of the Act and its risk-based approach.The implementation itself is based on a strong pre and post market system of enforcement and supervision.Its a decentralized system of national notified bodies checking compliance with the AI Act