《ISC：為人工智能做好國家研究生態系統的準備-2024年戰略與進展報告（英文版）（76頁）.pdf》由會員分享，可在線閱讀，更多相關《ISC：為人工智能做好國家研究生態系統的準備-2024年戰略與進展報告（英文版）（76頁）.pdf（76頁珍藏版）》請在三個皮匠報告上搜索。

1、1Preparing National Research Ecosystems for AISTRATEGIES&PROGRESS IN 2024 International Science Council,2024.“Preparing National Research Ecosystems for AI:strategies and progress in 2024”is published by The International Science Council,5 rue Auguste Vacquerie,75116 Paris,FranceTo cite this documen

2、t:Title:Preparing National Research Ecosystems for AI:strategies and progress in 2024URL:https:/council.science/publications/ai-science-systemsPublisher:International Science CouncilDate:March 2024DOI:10.24948/2024.06 Acknowledgments:The project was developed with critical inputs and guidance from S

3、imon See,NVIDIA,and Hayley Teasdale and Sage Kelly from the Australian Academy of Science.The Centre also acknowledges the support and help of Australian Academy of Science and Academy of Sciences in Malaysia in the organization of a workshop in Kuala Lumpur.The Centre also extends its thanks to Dav

4、id Arranz,European Commission,and Alastair Nolan,OECD for their feedback on the key issues framework.The Centre also thanks Dong Liu who helped in bringing the case studies to their finalization.The Centre is grateful for the financial support from the International Development Research Centre(IDRC)

5、for this working paper and future projects.About the Centre for Science Futures:The ISC Centre for Science Futures explores where changes in science and the organization of science are taking us in the future.The Centre for Science Futures works to improve our understanding and provide options and t

6、ools for impact and action.futures.council.science About the International Science Council:The International Science Council(ISC)works at the global level to catalyse change by convening scientific expertise,advice and influence on issues of major importance to both science and society.The ISC is a

7、non-governmental organization with a unique global membership that brings together more than 245 international scientific unions and associations,national and regional scientific organizations including academies and research councils,international federations and societies,and young academies and a

8、ssociations.council.scienceCover Photo:GarryKillian/FreepikDesign by:Mr ClintonPreparing National Research Ecosystems for AISTRATEGIES&PROGRESS IN 2024David Castle(project chair),Mathieu Denis,Dureen Samandar EweisWith contributions from:Australia:Emma Schleiger,Hayley Teasdale and Alexandra Lucchet

9、ti Benin:Ministry of Digital Economy and Communications Brazil:Mariza Ferro,Gilberto Almeida Cambodia:Siriwat Chhem Chile:Rodrigo Durn China;Gong Ke,Liu Xuan India:Moumita Koley,Jibu Elias Malaysia:Nurfadhlina Mohd Sharef Mexico:Dora-Luz Flores Oman:Hamdan Mohammed Al Alwai Uruguay:Lorena Etcheverry

10、,Guillermo Moncecchi Uzbekistan:Abduvaliev Adbulaziz Abduvalievich4TABLE OF CONTENTS MESSAGE FROM THE HEAD OF THE CENTRE FOR SCIENCE FUTURES 5INTRODUCTION 6LITERATURE REVIEW 8COUNTRY CASE STUDIES 161.Australia:Preparing for human-centric use of artificial intelligence.17 2.Benin:Anticipating the imp

11、acts of artificial intelligence on West Africas aspiring digital services hub.22 3.Brazil:Reaping the benefits of artificial intelligence with some cautionary notes .274.Cambodia:Seeking artificial intelligence approaches to national research missions .325.Chile:Finding possibilities to apply artifi

12、cial intelligence in an existing research financing ecosystem .366.China:Promoting the Artificial Intelligence for Science approach.40 7.India:Gaining insights into transformative technologies and their social integration.45 8.Malaysia:Enabling the Fourth Industrial Revolution .49 9.Mexico:Creating

13、a national lead agency for artificial intelligence.55 10.Oman:Fostering innovation through an Executive Program.59 11.Uruguay:Following a roadmap to prepare national science systems for artificial intelligence.64 12.Uzbekistan:Building the right conditions and skills for artificial intelligence .69N

14、EXT STEPS 745Message from the Head of the Centre for Science Futures In late 2023,the International Science Council(ISC)released a discussion paper on evaluating rapidly evolving artificial intelligence and related technologies1.This new working paper on how countries are preparing their research ec

15、osystems for AI confirms the engagement of the ISC to explore the impact of AI on science and societies.Additional studies and initiatives by the ISC will develop in the coming months and years.This working paper addresses a gap in ongoing discussions about AI policies,namely the implications of the

16、se policies for national science and research ecosystems.This is a critical issue for the future of science globally.Yet,very little has been published on these issues thus far and the information on countries plans is hard to find.Our ambition with this paper is to increase our knowledge of current

17、 initiatives toward the integration of AI in national research ecosystems,of what has been achieved so far,and the possible roadblocks.To these ends,this paper provides a literature study and twelve country case studies.By the end of 2024,we will release a second,more comprehensive edition of this p

18、aper,incorporating additional case studies,and putting forward recommendations for more coordinated and collaborative science policies for AI.We had a few different but overlapping audiences in mind when developing this work.If you are a STI policymaker involved in integrating new AI technologies in

19、 your countrys research ecosystem,you will find in this paper first-hand evidence on the issues that are of importance for your work,as well as examples of the initiatives taken by other countries.It is likely that you will find examples of countries from your region,with research ecosystem of a sim

20、ilar size as that of your country.If you work with a granting council or a philanthropy,this paper will give you a sense of the priorities that countries have identified for the uptake of AI in science.If you work with an AI company and you are concerned with the specific technological and infrastru

21、ctural needs of science and research institutions,this paper will give you a primer on the challenges identified by countries as they roll out their AI strategy for research.If you are a scientist or a science journalist,and your main interest is on the impact of AI on science in general,you will di

22、scover in this paper the extent to which countries are currently actively adapting their science system for AI.This is the beginning of a conversation.We invite science leaders involved in preparing the uptake of AI in their institutions and countries to engage with us in the coming months and beyon

23、d.We ask that you share your approaches,experience,and questions.Your inputs will be critical in the further development of this project and help us all better prepare for this critical technological transformation of our science systems.1 ISC,2023.A framework for evaluating rapidly developing digit

24、al and related technologies:AI,large language models and beyond.International Science Council.DOI:10.24948/2023.11 https:/council.science/publications/framework-digital-technologies/Mathieu Denis6Introduction The dominant notion communicated today on the influence of artificial intelligence(AI)is it

25、s capacity to change everything across all sectors,including science(Khalif et al.,2023;Nature,2023;Van Noorden and Perkel,2023;Miller,2024).Beyond the promises of new advances in different fields of research,a set of critical questions is emerging about the impact of AI on the documentation,funding

26、 and reporting of science:How is the increasing use of AI going to influence research funding allocation?What research data standards will evolve?How will AI change the nature of scientific outputs?How will scientific careers evolve with the increasing use of AI in research?What investments in infra

27、structures are required for the successful uptake of AI by the science sector?What legal adjustments are needed to enable the use of AI in research while ensuring high standards in the responsible conduct of science?How is AI going to affect international research collaborations?Discussions around t

28、hese questions are critical for the future of science and research systems.Research institutions and ministries are beginning to tackle them,although with limited resources to guide them.As this study will show,there remains a notable absence of comprehensive literature regarding the impact of AI on

29、 the structural aspects of science and research.Several countries have developed overall AI strategies to set out their plans and aspirations for AI development and implementation across different sectors.Despite the immediate and significant implications of these strategies for science and research

30、,these documents mostly offer broad statements on the involvement of science and research institutions in delivering the national plans without looking further into the concrete implications.This is not to suggest that countries are inactive.Quite the opposite:much is under way.Partnerships are bein

31、g formed,training initiatives launched,infrastructures put in place and policies implemented.However,people in governmental ministries,universities and consultancy firms tasked with spearheading the preparation of the research environment for AI are largely working with speculation on the key challe

32、nges and have limited insight into the approaches being adopted by countries of similar size and capacity.Frameworks outlining the key issues for countries to consider when planning the integration of AI into their research ecosystems can come a long way at this critical stage.This working paper off

33、ers one such framework derived from an analysis of the existing literature.To start establishing a knowledge baseline,the paper also presents 12 case studies from countries of different sizes and regions,authored by people directly engaged in these discussions in their respective countries.We intend

34、 to expand the number of case studies and achieve a more comprehensive representation of the different global regions in the upcoming and final edition of the paper by the end of 2024.7It is important to consider the circumstances of countries of varying sizes,which are also major contributors to sc

35、ientific advancements,rather than solely focusing on the AI powerhouses.We deliberately sought to gain insight into how small to medium-sized countries are preparing their research ecosystems for the uptake of AI.This working paper therefore seeks to:gather the basic knowledge and information about

36、the issues,and the current efforts to prepare science and research systems for AI;help countries as they develop roadmaps for the uptake of AI in their science systems;create regional and global networks of people involved in the reflections on adaptation and implementation of AI for science;raise a

37、wareness and help shape a critical discussion among the scientific and policy communities of the critical issues that AI raises for the organization of science and research.The development of the working paper benefitted from a workshop convened in October 2023 in Kuala Lumpur,Malaysia,bringing toge

38、ther participants from 12 countries in Asia and the Pacific.Contributions from some countries who participated in the workshop have been incorporated into version 1 of the paper.The coordination of the workshop was generously supported by the Australian Academy of Science and the Malaysian Academy o

39、f Sciences.The publication of this paper will be followed by similar regional workshops and consultations.A second version of the paper will be released later in the year featuring additional country case studies and a set of conclusions and recommendations.ReferencesKhalif,Z.N.,Mousa,A.,Hattab,M.K.

40、,Itmazi,J.,Hassan,A.A.,Sanmugam,M.and Ayyoub,A.2023.The potential and concerns of using AI in scientific research:ChatGPT performance evaluation.JMIR Med.Educ.,Vol.9,pp.e47049.https:/doi.org/10.2196/47049.Miller,A.2024.The top arguments against artificial intelligence in science.ReHack,19 January.ht

41、tps:/ will transform science now researchers must tame it.Nature,No.621,pp.658.https:/doi.org/10.1038/d41586-023-02988-6.Van Noorden,R.and Perkel,J.M.2023.AI and science:what 1,600 researchers think.Nature,No.621,pp.672675.https:/doi.org/10.1038/d41586-023-02980-0.8Literature ReviewWhat are the crit

42、ical issues for the integration of artificial intelligence in science systems?A bibliometric analysis.This working paper seeks to take stock of how countries are approaching and planning the uptake of AI by their science and research ecosystems.A bibliometric study was undertaken to identify publica

43、tions from different parts of the world exploring the impact of AI on national science and research ecosystems.The study was done in partnership with Nature Research Intelligence in September 2023.It combines academic journal and book content,conference proceedings,policy documents and grey literatu

44、re.The search strategy encompassed three steps:A high-precision keyword search(with more than 30 search keywords)generated abase document set.Over 1,600 documents were thus identified using the Dimensionsdatabase.A review of that initial corpus of documents and selection of the most relevant ones(18

45、0 in total)created a training document set.The refined training document set was used to identify similar documents.Additionalweb searches were also made.The resulting dataset comprises 317 documents publishedbetween 2018 and 2023.They are the documents used in this review.Classification of the 317

46、publications in the literature review2PUBLICATION TYPENUMBERJournal articles123Book chapters59Preprints51Web pages30Conference proceedings20Policy documents18Booksand monographs16While 317 publications dealing with national plans to integrate AI in science and research ecosystems may seem relatively

47、 low,there was a tenfold steady increase in numbers of publications published annually between 2018 and 2022(from 9 to 88).This increase suggests a growing attention to the issues relating to the uptake of AI in national science and research ecosystems.We can realistically expect the number of publi

48、cations to continue growing in the coming years,as more experience is gathered on the progressive integration of AI in national science and research infrastructures.2 The full list of publications is accessible on the Centre for Science Futures website:https:/council.science/publications/ai-science-

49、systems9Leading countries by publication volume across project dataset(20182023)2COUNTRY PUBLICATIONS%TOTAL PUBLICATIONSUnited Kingdom 3211.9%United States2810.4%Germany134.8%China103.7%Canada93.3%India83.0%Sweden72.6%Spain72.6%Switzerland62.2%Singapore51.9%The review of these publications allowed u

50、s to identify a core set of 45 issues and topics which experts and observers have posited as critical for the integration and uptake of AI in research and science systems.We tried capturing these issues using a simplified version of OECDsframework for technology governance,with three broad themes:re

51、search and development agenda setting,technology assessment,foresight and science advice;public engagement,science communication and public accountability;regulation,standards,private sector governance and self-regulation.Some of the issues listed here are not specific to science and research,such a

52、s those related to careers and employment,data quality and AI safety,and those having to do with the development and adoption of AI in general.We tried to limit the number of such issues in this exercise but included those with a particular significance for science(e.g.data quality)or that we expect

53、 to be increasingly discussed in relation to the uptake of AI in research(e.g.AI safety and employment).Theme AREASa.Topics Issues10THEME 1:R&D agenda setting,technology assessment,foresight and science advice PRIORITY SECTORSa.Priority-setting We must find ways to identify strategic sectors for AI

54、development and for its uptake by the scientific community.Mechanisms may include funding,infrastructure development and capacity building programmes.FUNDING PRACTICESa.Will AI capacity replace scientific merit in science funding decisions?AI intensity may become an inappropriate deciding factor in

55、determining the allocation of resources and hence the trajectory of scientific discovery.Its salience could close off areas of research that do not use it.Competition within research could become less a matter of merit and more a matter of access to AI.This risks poor decision-making and further con

56、centration of research funding.b.Use of AI in resource allocation AI relies on machine learning from existing material.It may produce reviews that are inherently conservative and which reproduce old biases.c.Impact of AI on evaluation panels AI-driven science tends to be interdisciplinary because AI

57、s do not know subject boundaries.Todays domain-led expert panels may be unable to review it adequately,despite the many recent calls for science to be more interdisciplinary.CAPACITY BUILDING AND RETENTIONa.Growing AI skills in the scientific community There is a need for broad but differentiated AI

58、 skills development for learners and practitioners at all levels.Important aspects include education in AI,training in domain-specific use,ethics,and interdisciplinary competencies.Teaching will have to recognise that this is a fast-moving topic.b.Diversity in AI research There is a need to ensure t

59、he gender,ethnic and cultural diversity of the AI workforce,in the interest of equity and to improve the quality of research and other outcomes.Machine learning can reproduce existing inequity.We have to develop the right incentives for disciplinary and interdisciplinary AI.c.Talent retention in the

60、 public science sector Public sector science,including universities and research centres,needs talent acquisition and retention,given the strong demand for AI skills from the private sector.Unusually,this is an area in which the private sector can offer interesting jobs as well as high salaries.11 I

61、NFRASTRUCTUREa.Development of cloud computing appropriate for science Uncertain funding for cloud computing and research data repositories constrains scientific advances.In the absence of public cloud capacity,wealthier research institutions are likely to contract private companies,limiting the shar

62、ing of their research data and leaving less wealthy institutions behind.b.The digital divide goes algorithmic We must determine how inequity in AI access between individuals,groups,academic disciplines,organisations and locations results in poorer research outcomes.c.Development of AI tools for scie

63、nce We must determine what kinds of partnerships will encourage the development of AI tools appropriate for specialized research institutions.How do we ensure that new AI technologies are not driven solely by the AI and machine-learning communities,but rather developed jointly with all research comm

64、unities?INTERNATIONAL COLLABORATIONa.Variation between legal systems We need to assess how jurisdictional variability in governance and data protection between countries impacts international research and research collaboration.b.Regional collaboration Countries must find out the extent to which the

65、y can cooperate to establish regional AI centres and research networks if they dont have the resources to do it on their own.JOBS,CAREERS AND EMPLOYMENTa.Impact on jobs in science and research There is a need to monitor how advances in AI affect the number and nature of jobs in science.b.Continuous

66、AI training There is a need to develop ways for scientists and research staff to keep up to date with AI in order to produce better research and minimise job losses.There may need to be specialist AI trainers and teachers,for example to help users understand the ethical issues raised by AI.NETWORK A

67、ND REPOSITORY SECURITYa.AI effects on scientific cybersecurity Science institutions must ensure the best possible network hygiene,ensure the security of partner organisations,and control cybersecurity risks from individual people.How do they secure facilities against intellectual property theft,acce

68、ss to private and sensitive data,and ransom attacks?The protection of data quality and integrity requires controls on access to repositories,as well as highly qualified personnel,strong partnerships and an appropriate built environment.12THEME 2:Public engagement,science communication and public acc

69、ountability SCIENTIFIC INTEGRITY IN THE CONDUCT OF RESEARCHa.Principles and values of current science AI may generate tensions between some of the core principles and values that define todays science.Such contradictions might include openness vs.rigour;privacy and confidentiality vs.open science;ma

70、ssive data vs.high quality data;or explainability vs.“black box”results.b.Reliability and explainability of results Lack of trust in AI,within science and in other activities,may create challenges for its uptake in science.But uncritical trust will lead to a potentially dangerous overreliance on AI

71、technology and the results it generates.AI tends to produce normative results rather than groundbreaking insights,because it is based in existing knowledge and existing opinion.c.Reproducibility Todays science already has severe reproducibility issues.How will AI worsen them or perhaps solve them?Fo

72、r AI to improve reproducibility it will need to be more transparent,providing more information about codes,underlying data and experiment design.This applies both to AI research and to research using AI.d.Explainability of results The scientific method requires scientific claims to be explainable an

73、d understandable.Some popular AI methods operate as a black box,making it impossible to say how they have reached their conclusions or to identify spurious correlations or causalities.e.Ethical data use The use of big data and AI complicates present-day notions of consent and of human research parti

74、cipants,as well as the ways in which data is collected and used.AI Ethics and Review Boards focus on human subjects.As well as carrying out their present vital role,they should be able to examine possible harms to wider society.f.Accountability We will have to determine who is responsible for fabric

75、ation,falsification,plagiarism and other bad practice when the faulty conduct can be traced back to an AI.The answer may be simple if the AI has an obvious owner,but in the future many may not.g.Conflict of interest We need to see whether new conflicts of interest arise as AI spreads.They may not be

76、 covered by current conflict-of-interest policies.ENVIRONMENTAL IMPACT AI development has to be made more sustainable(in relation to the use of computer chips and electricity in particular).More fundamentally,AIs may well not be attuned to environmental concerns if they have not learned from appropr

77、iate input materials.13 SCIENTIFIC PUBLISHINGa.Acknowledgment of contributors and authors Researchers have to explain how AI was used in the production of research outputs.b.AI for policing science Publishers have to determine whether AI should be used to detect non-AI generated fabrication,falsific

78、ation and plagiarism.THEME 3:Regulation,standards,private sector governance and self-regulation DATA QUALITYa.Accuracy Larger datasets are better for training AIs,yet they are also more likely to produce responses based too closely on the data available to them(overfitting)or to contain inaccuracies

79、 and biases that could result in wrong or misleading results.Incorrectly sourced data,Frankenstein datasets and biased datasets already have dangerous implications for science.This problem needs to be addressed at every level,from considerations of governance and management to operational use.b.Bias

80、 and exclusion While AI,and large language models in particular,use biases(statistical similarity)in data to produce results,it is important to curate training data to avoid further marginalization of particular groups and regions.Digital exclusion leads to gaps in data.Furthermore,how do we represe

81、nt those who are offline?c.Subject orientation of data vs.the interdisciplinary nature of AI research Most scientific knowledge comes from a specific subject.We need to encode and use it,while enabling communication between domains and allowing for the growing generation of interdisciplinary knowled

82、ge.d.Data coding and annotation AIs,and large language models in particular,require humans to code and annotate the data they use.These individuals must be aware of the risk of embedding cultural differences in the data during the annotation process.DATA MANAGEMENT AND GOVERNANCEa.Open data vs.AI sa

83、fety Access to high-quality data is crucial to the development of AI for science.But the public interest,as well as that of individuals,calls for governance structures to protect privacy and to guarantee the ethical use of data.b.Access vs.Advantage Much of the data required for the development of s

84、cientific AI will not fall within the scope of open data initiatives,for example data held by the private sector.The tension between enabling access and maintaining commercial advantage may result in high-quality data being kept confidential.c.Data infrastructures The development of AI for science w

85、ill require harmonization of practices and the development of communities of practice.Current norms and practices for the 14production and use of data differ between disciplines and institutions.As scientific organizations increase their data curation and storage capacity,they will need to increase

86、interoperability between repositories.DATA STANDARDSa.Data standards for provenance The sources of training data must be appropriately disclosed and evaluated.A specific concern is the ethical aspect of data and data sources,and its implications for bias in AI.b.Data standards for quality(see also d

87、ata quality above)Technical standards,certification and compliance should be imposed to ensure that data used in science is properly curated and stored.LAW,REGULATION AND POLICYa.Legal liability of research done with AI We have to reconcile traditional liability systems with AI processes and outputs

88、,with their varying degrees of autonomy and transparency.At what point does an AI,rather than its maker,become responsible for its actions?b.Copyright protection or patenting for machine-generated creations?Uncertainty about the eligibility and appropriateness of copyright protection for AI-generate

89、d creations may lead to the use of patenting or trade secrecy techniques to protect intellectual property.This would reduce public availability of the valuable results,positive and negative,of AI projects.c.Protection and use of digital data Text and data mining risk infringing copyright through the

90、 creation of unauthorized copies,and may violate the terms and conditions of websites and databases.The United Kingdom is creating a copyright exception rule for text and data mining,and other jurisdictions may follow.Works mined for data can be protected by copyright,but data themselves are usually

91、 protected only if they were part of original datasets.This may lead to the use of trade secret to protect data.The European Union protects data extracted from protected databases for scientific research.But the borderless character of digital data exacerbates tensions between jurisdictions.REGULATI

92、ONSa.The domestic regulatory environment Work towards domestic AI regulation will be a balancing act between different considerations and needs.In these arbitrations,countries must create beneficial conditions for their science and research sectors to thrive and work for the common good.b.Impact of

93、regulation in other jurisdictions Observation of other countries actions can lead to leap-frogging and the alignment of provisions;or,uncertainty about regulation may lead some legal regimes to seek competitive advantage through less rigorous regulation,to the detriment of the country where the crea

94、tion was generated.15It has become common sense to predict that AI will transform science and research.The encompassing set of considerations and issues identified through the literature review unpacks the many ways in which AI is influencing how science is made,organized and funded.They relate to c

95、onditions for good and responsible practices of science with AI.The list should therefore be of use to countries as they develop and implement roadmaps for the uptake of AI in their science and research systems.It reflects imperfectly,however,the considerations that are currently guiding countries.A

96、s will become clear in the case studies current plans for the uptake of AI in science are only partially driven by considerations such as those highlighted in the list.By and large,they are rather guided by a countrys overall approach to AI and seek to support the ambitions(in terms of economic grow

97、th,better governance,digital infrastructures,etc.)attached to AI more generally.This partial disconnect and the pre-eminence of national strategies is understandable.However,insufficient attention to the specific conditions for a successful uptake of AI in science and research will affect the qualit

98、y of science in these countries and everywhere.It will be measured in poor research data policies,strengthened epistemic biases,insufficient capacity and ineffective institutional and regulatory environments.It will lead,in other words,to bad science.16Introduction to the case studiesThe following c

99、ase studies were developed to help increase our collective knowledge and understanding of countries approaches towards the integration of AI in research ecosystems.These short essays were developed by people involved in the development and rolling out of their countrys AI strategy for science.The co

100、untries were selected somewhat opportunistically,using ISCs networks and connections to identify willing contributors from diverse global regions.The next iteration of this report will include more case studies and a more balanced geographical representation including Canada,France,Jordan,Malawi,Mor

101、occo,Nigeria,Norway,United Arab Emirates,United Kingdom,Panama,Romania,Rwanda,South Africa,United States.In our initial interactions with the authors,we introduced the projects goals and ambitions,and provided a set of guidelines.Case studies signed by authors reflect each authors perspective based

102、on their experiences in their positions and what they deem most pertinent and current at the time of writing.In line with the ambition of expanding our knowledge basis and initiating a discussion,authors were encouraged to provide factual information and refer to key documents.An internal review pro

103、cess within the projects core team was conducted upon receipt of the first draft from each author.Comprehensive feedback was provided on the first drafts from the project team,followed by a secondary discussion to address the feedback and refine the draft further.References to the key documents fram

104、ing countries approaches are included in each case study.The bulk of those documents are not findable in the international publication databases and were therefore not included in the literature review discussed previously.17AUSTRALIAPreparing for human-centric use of artificial intelligenceEmma Sch

105、leigerCommonwealth Scientific and Industrial Research Organisation Dr Hayley Teasdale and Alexandra Lucchetti,Australian Academy of ScienceKey takeaways Ethical principles and human-centric approaches to AI are informing Australias emerging framework for AI governance.The number of tertiary educatio

106、n offerings for AI have increased in Australia and are complemented by an initiative to attract and train job ready AI specialists.While active programs to enhance diversity in Australias STEM workforce exist,they are not specifically tailored to address AI.Additionally,there is a recognized need to

107、 enhance ethical competence and raise awareness of human rights in AI-related scientific endeavours.However,more customized resources for the science sector are required.Other challenges remain to be addressed such as the high-performance and data computing infrastructure needed for AI and AI-enable

108、d science and the implementation of FAIR and CARE data principles.Australias government,scientific organizations and universities are exploring the preparedness of the national science system to capture the opportunities and mitigate the risks of AI to accelerate scientific discovery.For example,the

109、 national science agency,the Commonwealth Scientific and Industrial Research Organisation(CSIRO),released the report Artificial Intelligence for Science Adoption Trends and Future Development Pathways (Hajkowicz et al.,2022).It examines the impact of AI on science and the imperative for research org

110、anizations to invest in mechanisms to harness the benefits and mitigate the risks of these technologies.The report outlines six future development pathways to enable the transition,including hardware and software upgrades,data capability uplift,improved education and training,the development of huma

111、n-centred AI,improved workforce diversity and ethical capability.Organizations throughout Australias national science system have begun expanding their capacity for AI uplift in these areas with recent research initiatives,activities,programs and guidelines.However,challenges remain to be addressed.

112、Hardware and software Scientific organizations seeking to uplift their AI capability must make decisions about hardware,software and computational infrastructure upgrades.The Australian Academy of Science recently held a national roundtable to discuss the Australian science sectors future supercompu

113、ting needs.The group highlighted the need for a national strategy and an exascale computing facility to secure Australias sovereign research capability and enable science to meet national and regional priorities into the future (Australian Academy of Sciences,2023).18Data Future AI capability uplift

114、 also requires investment in high-quality data which is fit for purpose,provenance assured,validated,up to date and ethically obtained.The Australian government is leading by example through its Data and Digital Government Strategy(Government of Australia,2023).This initiative focuses on adopting be

115、st-practice approaches to data collection,management and use to become a data-driven organization.In conjunction with the increasing use of AI,it is essential for Australia to better implement the FAIR(Findable,Accessible,Interpretable and Reusable)and CARE(Collective benefit,Authority to control,Re

116、sponsibility and Ethics)data principles.These and other principles and practices from open science,the Indigenous Data Sovereignty movement and participatory data stewardship all provide critical guidance for the creation,use and management of the data that will underpin AI in Australias science sys

117、tem.Education,training and capability There is an imperative for education,training and capability uplift across the science sector and into lifelong education.The number of tertiary AI courses offered in Australia almost doubled between 2020 and 2023,providing greater educational opportunities(37 o

118、fferings in 2020,69 in 2023)(OECD,2024).The Australian Human Rights Commission(2023)has recommended that professional development and training be provided to teachers and schools should introduce comprehensive digital literacy programs to provide students with the skills needed to engage with genera

119、tive AI tools in a responsible and ethical way.In 2021,AUD24.7 million was invested in establishing CSIROs Next Generation AI Graduates Program to attract and train job-ready AI specialists in Australia(CSIRO,2021).Currently,more than a thousand CSIRO researchers are working on a diverse range of AI

120、 and data science projects(CSIRO,a).Human-centric artificial intelligence HumanAI collaboration and human-centric AI is designed and implemented to ensure humans can work effectively with AI and benefit from the complementary strengths of humans and AI systems to carry out tasks to higher standards

121、than either can achieve alone.In 2023,Australia signed the Bletchley Declaration affirming that AI should be designed,developed and deployed in a human-centric,responsible and trustworthy manner.CSIROs collaborative intelligence(CINTEL)program of work is developing the science and technology to ensu

122、re AI systems support humans to solve scientific challenges,such as highly labour-intensive tasks like genome annotation(CSIRO,b).Annotation uses the genome sequence to create biological phenotypes critical for increasing crop yields through selective breeding.The group is developing a scalable appr

123、oach involving collaboration between a domain expert and AI that will allow for accurate and timely annotation of genomes.19Gender,ethnic and cultural diversity The AI workforce lacks gender,ethnic and cultural diversity,which limits the quality of outcomes.Improving this will contribute to an uplif

124、t in AI capability within research organizations.The Government of Australias(2020)Advancing Women in STEM Strategy Action Plan 2020 provides a national,coordinated approach to achieving sustained increases in gender equity in science,technology,engineering and mathematics(STEM).Programs such as Dea

125、dly Science(Deadly Science)and the Indigenous STEM Education Project(CSIRO,2021)seek to support and engage Aboriginal and Torres Strait Islander students in science-and STEM-related careers.Between 2014 and 2021,the Indigenous STEM Education Project reached over 23,000 participants in 603 schools,an

126、d Deadly Science has delivered 7,500 boxes of science resources to over 800 schools.Ethical capability Evolving standards and regulation of the design and implementation of AI require investment in ethical capability including technology,skills and cultures.In support of responsible innovation,the A

127、ustralian government has produced a framework of eight ethics principles to ensure AI is safe,secure and reliable (Dawson et al.,2019;DISR,a).This was followed by the 2023 discussion paper Safe and Responsible AI in Australia(DISR,2023)to support responsible AI practices and increase community trust

128、 and confidence through consultative government responses .The Australian governments January 2024 interim response to the consultation identified a range of legal,regulatory and governance measures that are needed to ensure AI is designed,developed and deployed safely and responsibly(DISR,2024).CSI

129、ROs Responsible Innovation Future Science Platform is a program of research that systematically and scientifically assesses the risks,benefits and uncertainties of future science and technology.Meanwhile,the Australian Human Rights Commission(2021)recommends that professional accreditation bodies fo

130、r STEM should introduce mandatory training on human rights by design as part of continuing professional development.However,no framework or strategies are in place for such upskilling in the science sector,and very few professional accreditation bodies exist.Other challengesAs well as impacting how

131、science is done,AI may impact how science is administered,governed,funded and assessed.Australias research councils,the Australian Research Council and the National Health and Medical Research Council,have created policies to account for the role of generative AI in their grant processes(ARC,2023;NH

132、MRC,2023).The use of generative AI is prohibited in assessing applications to preserve the confidentiality and integrity of the process.For applicants,the policies note the potential benefits and need for caution in using AI but do not list any specific restrictions on the use of AI by applicants.Th

133、e AI workforce lacks gender,ethnic and cultural diversity,which limits the quality of outcomes.Improving this will contribute to an uplift in AI capability within research.20References ARC.2023.Policy on Use of Generative Artificial Intelligence in the ARCs Grants Programs.Canberra,Australian Resear

134、ch Council.https:/www.teqsa.gov.au/guides-resources/higher-education-good-practice-hub/artificial-intelligence Australian Academy of Science.2023.Is Australia ready for our supercomputing future?News and Media Releases,29 November.Canberra:Australian Academy of Science.https:/www.science.org.au/news

135、-and-events/news-and-media-releases/is-australia-ready-for-our-supercomputing-future(2023).Australian Human Rights Commission.2021.Human Rights and Technology Final Report.Sydney:Australian Human Rights Commission.https:/humanrights.gov.au/our-work/technology-and-human-rights/publications/final-repo

136、rt-human-rights-and-technology Australian Human Rights Commission.2023.Utilising Ethical AI in the Australian Education System:Submission to the Standing Committee on Employment,Education and Training.Sydney:Australian Human Rights Commission.https:/humanrights.gov.au/our-work/legal/submission/utili

137、sing-ethical-ai-education-system CSIRO.a.Artificial Intelligence.CSIRO.Canberra,Commonwealth Scientific and Industrial Research Organisation.https:/www.csiro.au/en/research/technology-space/ai CSIRO.b.Collaborative Intelligence Future Science Platform.Canberra,Commonwealth Scientific and Industrial

138、Research Organisation.https:/research.csiro.au/cintel/.CSIRO.2021 c.Indigenous STEM Education Project.CSIRO.Canberra,Commonwealth Scientific and Industrial Research Organisation.https:/www.csiro.au/en/education/programs/indigenous-stem-education-project.CSIRO.2021.2021-22 Federal Budget Statement.CS

139、IRO,12 May.Canberra,Commonwealth Scientific and Industrial Research Organisation.https:/www.csiro.au/en/news/All/News/2021/May/2021-22-Federal-Budget-Statement.Dawson,D.et al.2019.Artificial Intelligence:Australias Ethics Framework.A Discussion Paper.Canberra,Commonwealth Scientific and Industrial R

140、esearch Organisation.https:/www.csiro.au/en/research/technology-space/ai/AI-Ethics-Framework.Deadly Science.https:/deadlyscience.org.au/about-us/.DISR.a.Australias AI Ethics Principles.Canberra,Department of Industry,Science and Resources.https:/www.industry.gov.au/publications/australias-artificial

141、-intelligence-ethics-framework/australias-ai-ethics-principles.DISR.2023.Safe and Responsible AI in Australia:Discussion Paper.Canberra,Department of Industry,Science and Resources.https:/consult.industry.gov.au/supporting-responsible-ai.21DISR.2024.Safe and Responsible AI in Australia Consultation:

142、Australian Governments Interim Response.Canberra,Department of Industry,Science and Resources.https:/consult.industry.gov.au/supporting-responsible-ai.Government of Australia.2020.Advancing Women in STEM 2020 Action Plan.https:/www.industry.gov.au/publications/advancing-women-stem-strategy Governmen

143、t of Australia.2023.Data and Digital Government Strategy.https:/www.dataanddigital.gov.au/(Accessed 21 February 2024).Hajkowicz,S.et al.2022.Artificial Intelligence for Science Adoption Trends and Future Development Pathways.Canberra,Commonwealth Scientific and Industrial Research Organisation.https

144、:/www.csiro.au/-/media/D61/AI4Science-report/AI-for-Science-report-2022.pdf NHMRC.2023.Policy on Use of Generative Artificial Intelligence in Grant Applications and Peer Review.Canberra:National Health and Medical Research Council.https:/www.nhmrc.gov.au/about-us/resources/policy-use-generative-arti

145、ficial-intelligence.OECD.2024.AI courses in English by degree type.OECD.AI Policy Observatory.https:/oecd.ai/en/data?selectedArea=ai-education&selectedVisualization=ai-courses-by-degree-type(Accessed 21 February 2024).22BENINAnticipating the impacts of artificial intelligence on West Africas aspirin

146、g digital services hub Ministry of Digital Economy and CommunicationsKey Takeaways:Digital infrastructures and platforms have been put in place since 2016 as part of the Beninese vision as the hub for digital services of West Africa.Institutes in the country have initiated AI training and education

147、programs for the young generation.Challenges around data collection,preparation,access,storage and governance need to be addressed for proper operation of AI systems.Data protection and fundamental rights as well as data governance also raise legal,regulatory and ethical challenges The Government of

148、 Benin,with its vision to transform Benin into the digital services hub of West Africa for accelerating growth and social inclusion(MDEC,2016)has implemented several structural reforms and deployment projects of digital infrastructure and platforms since 2016.This vision has been articulated in the

149、governments action programs,which focus on flagship projects,priority projects,and projects with rapid impacts for structural,economic,political and social transformation of the country.The operationalization of its vision has enabled Benin to establish a digital code,a national data centre,a nation

150、al portal for public services,a public key infrastructure,a national administration network integrating over 187 sites,and a network of over 2,500 kilometres of fibre-optic cables deployed throughout the national territory,among other projects.The use of Benins new infrastructure and platforms will

151、generate massive amounts of data that must be managed and valorised through the use of AI tools and technologies so that their value creation potential does not escape the Beninese economy.National Artificial Intelligence and Big Data StrategyIt is within this framework that the Government of Benin

152、adopted,in January 2023,a National Artificial Intelligence and Big Data Strategy(SNIAM 20232027).This strategy outlines a structured action plan around four programs,including one related to Support for training,research,innovation,the private sector,and cooperation(MDEC,2023).Through this program,B

153、enin aims to support training and research by equipping universities and promoting partnerships in AI.It also aims to develop financing mechanisms by strengthening institutional support to the structures that are responsible for entrepreneurship and innovation as they mobilize and sustain resources

154、allocated to startups.Lastly,it aims to strengthen sub-regional and international cooperation in this area.23The development of SNIAM 20232027 was carried out in two phases:a preliminary stage followed by the development of the document itself.It was during the preliminary stage that the government

155、prepared by providing Benin with its digital code,connectivity infrastructure,data storage and platforms conducive to strengthening digital trust.However,many challenges remain to be addressed.There are data challenges concerning the collection,preparation,access,storage and governance of the data n

156、ecessary for the operation of AI systems.There are also notable legal and regulatory challenges related to AI governance and regulation,and ethical challenges concerning data protection and fundamental rights.At the same time,the opportunities for Benin are manifold and relate to supporting the deve

157、lopment of priority sectors such as education,vocational training,health,the living environment and transportation.Financing and institutional arrangementsWith an estimated cost equivalent to USD 7.7 million over a period of five years,the main actions of SNIAM 20232027 will be implemented through a

158、 publicprivate partnership,at the national level,targeting specific areas of development.Various sources of financing are proposed to mobilize the resources needed to implement the actions outlined in the strategy.These include calls for national funding from both the government and the private sect

159、or;appeals for bilateral and multilateral foreign aid;and appeals for foreign private capital within the framework of the publicprivate partnership.The integration of AI in Benin will require the participation of all public bodies,the public sector and the population to achieve the desired effects.T

160、he effects in question include improving productivity and the quality of products and services in priority sectors and those presenting real opportunities for AI;a dynamic AI ecosystem driven by Beninese companies;technology and knowledge transfers between research laboratories and the private secto

161、r;and recognition of Benin in the field of AI.Stakeholders shaping readiness in researchOn one hand,AI readiness in research involves public innovation bodies,and on the other hand,it involves civil society organizations,academics,startups and the private sector in general.Several targeted skill dev

162、elopment programs have been identified to help workers prepare for AI transitions.These programs are either directly envisaged by the government or in collaboration with partners.Thus,the Ministry of Digital Economy and Communications,as part of the operationalization of the AI strategy action plan,

163、is collaborating with various digital ecosystem partners in Benin to implement awareness raising,networking,training activities,and more.The use of Benins new infrastructure and platforms will generate massive amounts of data that must be managed and valorised through the use of AI tools and technol

164、ogies so that their value creation potential does not escape the Beninese economy.24Basic AI awareness actions are also planned during the development of digital literacy modules.The Smart Africa Alliance has developed a capacity building reference document that has led to the implementation of seve

165、ral projects and initiatives,including the Smart Africa Digital Academy(SADA)project,which supports existing processes in various countries(SADA,no date).In Benin,a convention for the implementation of SADA was signed in 2022,and in 2023,actions began to support the Lever of Learning for Retraining

166、in the Digital Sector(LeARN),focusing on three modules:training of 25 Data Steward experts,training of 25 Data Developers,and AI training(Government of Benin,2021).Furthermore,there are initiatives by some non-governmental actors in the Beninese digital and AI ecosystem that are worth highlighting.T

167、he Odon Vallet Foundation has held a Summer School on Artificial Intelligence since 2021,where around a hundred young people receive pragmatic and high-quality training on basic AI concepts such as programming,machine learning and embedded electronics(including robotics and home automation).Since 20

168、20,the Francophone Agency for Artificial Intelligence has been organizing awareness conferences for young Beninese people,including women,on the challenges of AI,as well as online masters-level training in AI and big data in partnership with Francophone universities(AFRIA,2020).National scientific a

169、nd research community SNIAM 20232027 is the result of a synergy of actions stemming from both government sectoral departments and the private sector,as well as associations or academic organizations.In its development process,the national strategys aim was to have a consensus document that takes int

170、o account vital domains such as research,developments and innovations,applications,market placement and intersectoral dissemination,support,and guidance for deployment.In terms of local research institutions,Benin has a training and research centre,the Institute of Mathematics and Physical Sciences(

171、IMSP),established in 1988.With its specialized resources in AI,the IMSP constitutes a centre of competence at the national level in mathematics and AI computer science(at the PhD level),and it has a supercomputer with rare power for an institute in West Africa.The challenge for the IMSP today is to

172、maintain computing power and strengthen the means to take advantage of this infrastructure.The Institute of Training and Research in Computer Science,the Abomey-Calavi Polytechnic School and its Doctoral School of Engineering Sciences,and the Laboratory of Biomathematics and Forest Estimations at th

173、e University of Abomey-Calavi are also working on several projects implementing AI technology as well as blockchain.Furthermore,several capacity-building actions have been initiated and are ongoing to prepare human resources for the labour market transformations induced by AI and emerging technologi

174、es in general.In addition to teaching computer science(networking and engineering),the IMSP has been offering a data science masters program since 2020,having already trained about twenty graduates,with around forty students currently undergoing training in this field.About ten theses in AI or relat

175、ed fields have already been defended at the IMSP.Additionally,at the Institute of Training and Research in Computer 25Science there is a bachelors program in AI.Efforts are under way to create a masters program here to allow students to continue their studies in AI.The AI training provided in this f

176、ield will address the various challenges in AI skills.Several universities and schools are also initiating training programs in AI within the private sector.For example,the Sm City Development Agency,in partnership with Sorbonne University,launched in 2022 a cohort of professionals who benefited fro

177、m highly certified continuing education in AI.Operational steps for the strategySNIAM 20232027 aims to make AI and big data a lever for Benins development by 2027,with increased support for strategic sectors such as education,health,agriculture,the living environment and tourism in an opportunistic

178、approach.Ongoing actions are distributed within the programs,and their implementation will be based on prioritization considering three factors.The first factor is business impact:the extent to which the proposed solution will benefit the primary beneficiary or address the original problem.The secon

179、d factor is given complexity:the extent to which the data are available and exploitable right now.The third is technological complexity:the effort it will take to create,deploy or adapt an AI solution.In operationalizing the strategy,initiatives are under way to identify and execute associated actio

180、n plans.These include feasibility studies and project definition to operationalize SNIAM 20232027.They also extend to the development of application platforms for AI use cases.As part of this latter action,the Government of Benin has implemented GPT.BJ,an initiative to promote access to legal inform

181、ation in citizens lives(Le Matinal,2023).GPT.BJ is a chatbot developed by the Benin Agency for Information Systems and Digital and is designed to answer questions related to the general tax code,digital code,labour code and penal code of Benin.It was launched in 2023 during the second Digital Entrep

182、reneurship and Artificial Intelligence Fair.References AFRIA.2020.Forum francophone de lintelligence artificielle(FFIA).Francophone Agency for Artificial Intelligence.https:/afria.global/portfolio/forum-francophone-de-lintelligence-artificielle-ffia/.(In French)Government of Benin.2021.Secteur du nu

183、mrique au Benin:Le Gouvernment Lance le Programme LEARN.Digital Sector in Benin:The Government launches the LEARN Programme.https:/www.gouv.bj/article/1316/.(In French)Le Matinal.2023.Nouvelle plateforme pour rpondre aux proccupations des Beninois:Gpt.bj,la grande innovation du Senia 2023.New platfo

184、rm to respond to Beninese concerns:Gpt.bj,the great innovation of Senia 2023.https:/lematinal.bj/nouvelleplateforme-pour-repondre-aux-preoccupations-des-beninois-gpt-bj-la-grandeinnovation-du-senia-2023/.(In French).MDEC.2016.Declaration de Politique Sectorielle:Orientations Strategiques 2021 dans l

185、e Secteur de lEconomie Numerique Sector Policy Statement:2021 Strategic Orientations in the Digital Economy Sector.Ministry of Digital Economy and Communications.https:/pspdb.dev.gouv.bj/server/storage/app/PolitiqueFichiers/69_DPS.pdf (In French).26MDEC.2023.Stratgie nationale dintelligence artifici

186、elle et des mgadonnes.National Artificial Intelligence and Big Data Strategy.Ministry of Digital Economy and Communications.https:/numerique.gouv.bj/publications/actualites/le-benin-se-doted-une-strategie-nationale-d-intelligence-artificielle-et-des-megadonnees.(In French).SADA.No date.Smart Africa

187、Digital Academy.https:/sada.atingi.org/.Accessed 21 February 2024.27BRAZILReaping the benefits of artificial intelligence with some cautionary notes Mariza FerroProfessor of Computer Science,Ethical and Sustainable AI,Universidade Federal Fluminense(UFF),Head of Reference Group for Ethical and Trust

188、worthy Artificial Intelligence(Ncleo IA tica)Gilberto M.AlmeidaProfessor of Computer and Internet Law at the Pontifical Catholic University of Rio de Janeiro,co-Coordinator of Reference Group for Ethical and Trustworthy Artificial Intelligence(Ncleo IA tica)Key takeaways:The need to facilitate AI re

189、search and development has driven the Brazilian government to enact legislative reform and a key achievement is the partnership of the Ministry of Science with national funders and experts for the creation of AI Applied Research centres.Challenges in the country include a gap in AI literacy and educ

190、ation as well funding for AI research.There is also worry on the stagnation of the national AI strategy and bills of law that could hinder science and research priorities,foster uncertainty among researchers and limit international collaboration.Brazil has a significant history in promoting long-ter

191、m policies for digital development,beginning in the 1970s with improved digital infrastructure for data collection,storage,processing and sharing(for example,within the federal agencies SERPRO and DATAPREV).Since then,specific legislation has supported the creation of networks by bringing companies

192、and universities together for instance,IBM and the University of So Paulo,which have developed a joint initiative for long-term research projects on AI such as AI for sustainable agribusiness and food networks,climate decision-making with multiple criteria among other projects and accelerating deplo

193、yment of Internet usage,including by instituting an encompassing Civil Framework for the Internet(Government of Brazil,2014).However,according to a Berkman Klein Center research report(Fjeld et al.,2020)and the MIT Technology Review(Gupta and Heath,2020),despite those important steps Brazil was not

194、ranked well amongst Latin American countries,up until 2020,in terms of AI regulations and respective national strategies.It made some progress thereafter,with later initiatives positioning it on OECDs Policy Observatory map of AI regulation and national strategies,as well as in reports from private

195、organizations such as the Global AI Index and others(IAPP,2023).28Research centresBrazil reached important milestones between 2018 and 2021,particularly with the enactment of new laws(Government of Brazil,2018;2019a)that removed bureaucratic barriers to digital transformation driven by AI research a

196、nd development.That was the scenario when,in 2019,the Ministry of Science,Technology,Innovation and Communications(MCTIC)partnered with the Foundation for Research in the State of So Paulo(FAPESP)and the Brazilian Internet Steering Committee to launch a call for creation of eight AI Applied Research

197、 Centres.The targeted beneficiary sectors were health,industry,cities,agriculture(formally prioritized in the Science,Technology and Innovation national policy),information security(including the investigation and design of algorithms and mechanisms)and cyber defence systems.Six of these centres wer

198、e selected in May 2021(one for AI in smart cities,one for agriculture,two for industry and two for healthcare)and four in 2023(two for AI in industry 4.0,one for renewable energy and one for cybersecurity).Each centre involves dozens of senior researchers and dozens of students,and each centre recei

199、ves around USD200,000 a year for up to ten years from FAPESP.National strategyIn April 2021,MCTIC presented the Brazilian National Strategy for Artificial Intelligence(EBIA),which was linked to the AI Applied Research Centres as another MCTIC structuring action to prepare the Brazilian science and i

200、nnovation system for AI(MCTI,2021).The EBIA aimed at designing an AI development plan for the country by providing guidelines for the federal Executive Branch to encourage research,innovation and development of AI solutions as well as on ethical and reliability concerns.Although the EBIA is a genera

201、l and macro-level national strategy,and has not specified particular fields for research on AI,it has indicated strategic actions where references to research are made,especially regarding research aimed at developing ethical AI solutions.Bills of law In parallel to the structuring of an overall adm

202、inistrative strategy,the legislative attempts to legitimize the national plan were followed,from 2019 through 2021,by the introduction of three AI bills of law in Congress(Government of Brazil,2019b;2020;2021),which particularly envisaged the fostering of innovation and the safeguarding of harm mini

203、mization.None of those bills of law were approved.In 2023,therefore,the Senate invited a group of 40 jurists to conceive of a fourth bill(Government of Brazil,2023;Hilliard,2023).Its contents were inspired by the European Unions AI Act then considered as international standard of good practice on th

204、e matter and included the aim of sustaining a risk-based approach to AI regulation.Such a long sequence is indicative of concentrated efforts on legislative action The EBIA aimed at designing an AI development plan for the country by providing guidelines for the federal Executive Branch to encourage

205、 research,innovation and development of AI solutions as well as on ethical and reliability concerns.29so far.Finally,since the last quarter of 2023,Congress has debated on all bills of law in an attempt to consolidate them.Strategic goals and actionAt the other end,in the administrative sphere,EBIA

206、purports to be driving the Brazilian government to stimulate research,innovation and development of AI solutions in accordance with multiple considerations,including the assurance of reliable and ethical development and usage(Government of Brazil,2022).Such goals have drawn on Organisation for Econo

207、mic Co-operation and Development(OECD)concepts and principles as source of reference for key issues to be addressed,and inspired EBIAs structure with regards to areas of concern for instance,inclusive growth.In practice,EBIA is split into six primary goals,namely:education,training and workforce;res

208、earch development and innovation;application in productive sectors;application in public administration;and public security.However,although such EBIA axes point to strategic actions,they have been vaguely worded,so there is lack of clarity on concrete ways to set proper public policies.The goals do

209、 not get into prescribed instrumental actions(Filgueiras and Junquilho,2023).For example,in the education axis the development of digital literacy programs is generically advocated for all areas and levels of education,irrespective of the natural specificities of each such as the particularities for

210、 the teaching of AI in the context of fundamental schooling,or of academic advanced studies.Paradoxically,the Latin American AI Index has interpreted these generic terms as a strength,making the assumption that Brazil has effectively incorporated AI elements into its national school curriculum.The C

211、ommon National Curriculum Base has indeed been recently updated to add computational thinking and computer programming items,but the reality is that AI literacy has not been properly introduced,as there are neither qualified teachers nor a defined strategy.Research guidelines Mirroring such a scenar

212、io,in November 2023 the Brazilian Academy of Sciences published a set of recommended guidelines for the use and scientific advancement of AI in Brazil(ABC,2023).The recommendations stress the existing gap in AI literacy and education throughout civil society,especially for minors,and in fundamental

213、action to prepare the national system for AI in the long term.Alongside these and other issues,the group of researchers from the Academy highlights the need for an immediate increase in funding from the government for public research(as public universities lead AI research in Brazil),the creation of

214、 mechanisms for the private sector to also increase investments in this technology,and the need for a regulatory environment safe for teachers and researchers(ABC,2023).In essence,the development of a national science system for AI necessitates the implementation of public policies designed to coord

215、inate the various enabling factors involved.Consequently,it is expected that the examination of the current legislative and administrative landscape in Brazil,coupled with an analysis of select studies from both scientific and grey literature,will afford insight into Brazils efforts to establish its

216、 national science system for AI and the resultant impact on the national science and research framework.30EBIA and AI bills of law serve as the principal instruments guiding scientific research priorities and fostering a targeted innovation ecosystem in Brazil.Failure to contemporaneously advance th

217、ese instruments may engender negative impacts by creating an uncertain regulatory environment for researchers and professors.Moreover,such stagnation could restrict international collaboration and funding.Missing implementation However,generally speaking,other Latin American countries national AI st

218、rategies(Chiarini and Silveira,2022)propose circa a decade to be implemented,while Brazil has attempted to do it within a relatively short period from 2020 to 2022.There should be little surprise,then,that no specific goals have been substantially achieved so far,despite the magnitude they may repre

219、sent in the context of a country with a continental size and population.EBIAs missing cascade of detailed indications of opportunities and challenges to implementation (Chiarini and Silveira,2022)is therefore a serious and urgent problem for Brazil,and for everyone who would likely benefit from AI r

220、esearch for accelerated solutioning.Given all of the above,the fact that AI is quoted in the Brazilian Digital Transformation Strategy 2018 nine times,but very generically and disconnected from any effective action or concrete objective,seems like one more sign that Brazil has not properly set EBIAs

221、 goals and has struggled for too long to approve a legislative platform.Brazils insufficient preparation for AI and machine learning makes its national science system inconsistent with international good practices.Its national challenges and possibilities,and regional prominence,demand prompt action

222、 and support.References ABC.2023.Recomendaes para o avano da inteligncia artificial no Brasil Recommendations for Advancement of Artificial Intelligence in Brazil.Rio de Janeiro:GTIA da Academia Brasileira de Cincias.https:/www.abc.org.br/wp-content/uploads/2023/11/recomendacoes-para-o-avanco-da-int

223、eligencia-artificial-no-brasil-abc-novembro-2023-GT-IA.pdf.(In Portuguese.)Chiarini,T.and Silveira,S.M.da.2022.Exame comparativo das estratgias nacionais de inteligncia artificial de Argentina,Brasil,Chile,Colmbia e Coreia do Sul:Consistncia do diagnstico dos problemas-chave identificados Comparativ

224、e Analysis of Artificial Intelligence National Strategies of Argentina,Brazil,Chile,Colombia and South Korea:Consistency of the Diagnostics of Key Issues Identified.Discussion Paper 2805.Rio de Janeiro,Instituto de Pesquisa Econmica Aplicada.http:/dx.doi.org/10.38116/td2805.(In Portuguese.)Filgueira

225、s,F.and Junquilho,T.A.2023.The Brazilian(non)perspective on national strategy for artificial intelligence.Discov.Artif.Intell.,Vol.3,No.7.https:/ Spanish.)Fjeld,J.,Achten,N.,Hilligoss,H.,Nagy,A.and Srikumar,M.2020.Principled Artificial Intelligence:Mapping Consensusin Ethical and Rights-Based Approa

226、ches to Principles for AI.Research Publication No.2020-1.Cambridge,MA,Berkman Klein Center for Internet&Society.http:/dx.doi.org/10.2139/ssrn.3518482.31Government of Brazil.2014.Lei N 12.965 Law No.12,965.https:/www.planalto.gov.br/ccivil_03/_ato2011-2014/2014/lei/l12965.htm.(In Portuguese.)Governme

227、nt of Brazil.2018.Lei N 13.674 Law No.12,965.https:/www.planalto.gov.br/ccivil_03/_ato2015-2018/2018/lei/l13674.htm.(In Portuguese.)Government of Brazil.2019a.Lei N 13.969 Law No.12,965.https:/www.planalto.gov.br/ccivil_03/_ato2019-2022/2019/lei/l13969.htm.(In Portuguese.)Government of Brazil.2019b.

228、Projecto de lei N 5051 de 2019 Senate Bill of Law No.5051 of 2019.https:/www25.senado.leg.br/web/atividade/materias/-/materia/138790 (In Portuguese.)Government of Brazil.2020.Projecto de lei N 21 de 2020.Senate Bill of Law No.21 of 2020.https:/www.derechosdigitales.org/wp-content/uploads/Brazil-Bill

229、-Law-of-No-21-of-2020-EN.pdf.(Unofficial English translation.)Government of Brazil.2021.Projecto de lei N 872 de 2021.Senate Bill of Law No.872 of 2021.https:/www25.senado.leg.br/web/atividade/materias/-/materia/147434.(In Portuguese.)Government of Brazil.2022.Brazilian Digital Transformation Strate

230、gy(E-Digital).20222026 Cycle.https:/www.gov.br/mcti/pt-br/acompanhe-o-mcti/transformacaodigital/arquivosestrategiadigital/digitalstrategy_2022-2026.pdf.Government of Brazil.2023.Projecto de lei N 2338 de 2023 Senate Bill of Law No.2338 of 2023.https:/www25.senado.leg.br/web/atividade/materias/-/mate

231、ria/157233.(In Portuguese.)Gupta,A.and Heath,V.2020.AI ethics groups are repeating one of societys classic mistakes.MIT Technology Review.https:/ is Brazil leading South Americas AI legislation efforts?.Holistic AI.https:/ AI Legislation Tracker.23 August 2023 update.Portsmouth,NH,International Asso

232、ciation of Privacy Professionals.https:/iapp.org/resources/article/global-ai-legislation-tracker/.MCTI.2021.Estratgia Brasileira de Inteligncia Artificial EBIA Brazilian Strategy of Artificial Intelligence EBIA.Braslia,Ministry of Science,Technology and Innovation.https:/www.gov.br/mcti/pt-br/acompa

233、nhe-o-mcti/transformacaodigital/arquivosinteligenciaartificial/ebia-documento_referencia_4-979_2021.pdf.(In Portuguese.)32CAMBODIASeeking artificial intelligence approaches to national research missionsSiriwat ChhemStrategic Advisor at Asian Vision Institute Key takeaways The collective efforts in d

234、eveloping cloud-based services in the country has been supported by local actors across different sectors.The National Research Agenda 2025 has identified the national challenges and has put a plan to address these challenges.There is limited funding and capacity for research in Cambodia as well as

235、weak alignment between research work and national challenges.Cultural caution around uncertain technologies makes a part of why education is predominantly prioritized for engineering and accounting.Among the immediate priorities are strengthening of the infrastructure for data and computing power as

236、 well as upskilling and expansion of AI practitioners.In the rapidly developing landscape of technological advancement,Cambodia stands poised to integrate machine learning and AI into its national science systems.We delve into the strategic approach taken by Cambodia,exploring various facets ranging

237、 from the governments perspective on AIs impact to the institutional arrangements and stakeholder involvement essential for fostering innovation and economic growth.Human-centred policies At the heart of Cambodias strategy lies a keen awareness of the transformative power of AI across diverse sector

238、s.With a vision aligned with global AI trends,the Government of Cambodia is crafting human-centred policies aimed at driving responsible AI research and development(R&D).The Ministry of Industry,Science,Technology and Innovation(MISTI)has published the report AI Landscape in Cambodia:Current Status

239、and Future Trends(MISTI,2023a).This forward-thinking approach underscores Cambodias commitment to leveraging technological innovations to enhance its socio-economic development,as the Supreme National Economic Council outlined in its Cambodia Digital Economy and Society Policy Framework 20212035(SNE

240、C,2021).Institutional framework Institutional arrangements play a crucial role in facilitating Cambodias AI agenda,with the government spearheading initiatives to initiate and integrate changes.Collaborative frameworks and knowledge-sharing platforms are instrumental in fostering collaboration among

241、 multidisciplinary research and innovation sectors,paving the way for holistic development.Cambodias Science,Technology&Innovation Roadmap 2030(MISTI,2021)33emphasizes that the National STI Policy prioritizes five pillars:governance,human capital,R&D,collaboration and ecosystem building.In addition,

242、MISTI(2023b)developed the Digital Tech Roadmap,pinpointing machine learning and AI as key technologies for national digital technology development.According to the MISTI(2023c)Science,Technology&Innovation Report 2022,MISTI has the mandate as a government entity to oversee the STI sector,and is resp

243、onsible for promoting the network of AI,robotics and automation in Cambodia.National research missionsThe National Research Agenda 2025 detailed by MISTI(2022)identified eight national research missions:1)local food;2)reliable energy supply;3)quality education;4)electronic and mechanical spare parts

244、;5)cloud-based services;6)electricity and potable water;7)carbon neutrality;and 8)digitally-enhanced health.The key research areas to support mission5 on cloud-based services are infrastructure,software,cybersecurity and accessibility.These services would be provided to businesses in Cambodia to dev

245、elop their digital capacities and store their data locally.MISTI,the Ministry of Education,Youth and Sport and the Ministry of Post and Telecommunications are all leading institutions in implementing policy instruments ranging from legal and policy frameworks to human resources,infrastructure and co

246、llaboration in accomplishing the cloud-based services research mission,with the National Council of Science,Technology and Innovation as the guiding body.Currently,universities and research institutions such as the CamTech University,Royal University of Phnom Penh,Institute of Technology of Cambodia

247、,Cambodia Academy of Digital Technology and Kirirom Institute of Technology,as well as broadband networks and service companies,software producers and cybersecurity companies,have been producing research to accomplish the cloud-based services research mission.Challenges and pathways to research and

248、innovation in CambodiaThe National Research Agenda(MISTI,2022)highlighted five challenges facing the national research and innovation system,all of which are relevant to AI research:There is national underinvestment in R&D and limited policy support to promote research.There is limited alignment bet

249、ween research activities and national challenges,and insufficient contribution of academic research to private sector innovation activities and policy-making.There is limited research capacity in the public and the private sectors.Research institutions need strengthening and resources.There is need

250、for stronger universityindustry linkages and sustainable international collaborations.In response,the National Research Agenda developed four pathways to achieve the countrys national research missions:.1Invest in research to support the eight research missions.2Strengthen the role and capacities of

251、 public research institutions.3Support research careers.4Incentivize research activities and collaboration.34Missing piecesOne urgent area of concern for Cambodia is the significant data and computing power required for effective machine learning algorithms.Infrastructure limitations and a shortage

252、of skilled practitioners in the AI field present immediate barriers for Cambodia.The lack of available talent and financing hampers AI research and experimentation,hindering the countrys ability to fully capitalize on AIs potential benefits.Additional support in the form of publicprivate partnership

253、 and international collaboration will be required to address these challenges.Cultural challenges also loom large as Cambodia delves deeper into AI adoption.A cautious yet experimental mindset is essential to navigate the uncertainties and errors inherent in AI implementation.Furthermore,fostering i

254、nnovation,critical thinking,and science,technology,engineering,arts and math education is crucial to equip the workforce with the skills necessary for successful AI development and deployment.Cambodias current education landscape is skewed towards the context of a developing country,with civil engin

255、eering and accounting as predominant majors.Without a strong foundation and culture of scientific reasoning,the impact of AI research and applications will be limited.Opportunities aheadMISTI collaborated with the United Nations Educational,Scientific and Cultural Organization(UNESCO,2022)in develop

256、ing the report Mapping Research and Innovation in the Kingdom of Cambodia.UNESCOs Global Observatory of Science,Technology and Innovation Policy Instruments survey conducted in 2021 conveyed that R&D expenditure and human capital in Cambodia were both limited.On the positive side,Cambodia is taking

257、steps to integrate AI effectively into its science systems.Networking,matchmaking and/or partner search for R&D/innovation activities and support for infrastructure were the two highest-ranked types of R&D and innovation-related support or services provided,at 50 percent and 40 percent consensus res

258、pectively.In conclusion,Cambodia offers a compelling narrative of a nation poised to harness the transformative potential of machine learning and AI for sustainable socio-economic development.The median age of Cambodia is 27 years,with a large majority of the population integrating social media,e-co

259、mmerce and mobile banking applications into their daily lives.With the unique combination of a young,tech-savvy population and a lack of legacy technologies,Cambodia has the unique characteristics to leapfrog conventional technological and industrial revolutions.Although late to the game,the timing

260、is opportune for Cambodia to adopt AI at the national level,in an era where the power of AI is now more accessible than ever.Through strategic planning,stakeholder engagement and a commitment to inclusivity,Cambodia is charting a path towards a future where technological innovation drives progress a

261、nd prosperity for all.35ReferencesMISTI.2021.Cambodias Science,Technology&Innovation Roadmap 2030.Phnom Penh,Ministry of Industry,Science,Technology and Innovation.https:/misti.gov.kh/public/file/202108261629990117.pdf.MISTI.2022.National Research Agenda 2025.Phnom Penh,Ministry of Industry,Science,

262、Technology and Innovation.https:/misti.gov.kh/public/file/202302191676819231.pdf.MISTI.2023a.AI Landscape in Cambodia:Current Status and Future Trends.Phnom Penh,Ministry of Industry,Science,Technology and Innovation.https:/misti.gov.kh/public/file/202305301685426285.pdf.MISTI.2023b.Digital Tech Roa

263、dmap.Phnom Penh,Ministry of Industry,Science,Technology and Innovation.https:/misti.gov.kh/public/file/202307291690603726.pdf.MISTI.2023c.Science,Technology&Innovation Report 2022.Phnom Penh,Ministry of Industry,Science,Technology and Innovation.https:/misti.gov.kh/public/file/202307101688972531.pdf

264、.SNEC.2021.Cambodia Digital Economy and Society Policy Framework 20212035.Phnom Penh,Supreme National Economic Council.https:/mef.gov.kh/download-counter?post=7116.UNESCO.2023.Mapping Research and Innovation in the Kingdom of Cambodia.GO-SPIN Country Profiles in Science,Technology and Innovation Pol

265、icy,Volume 11.Paris,United Nations Educational,Scientific and Cultural Organization.https:/www.unesco.org/en/articles/mapping-research-and-innovation-kingdom-cambodia.36CHILEFinding possibilities to apply artificial intelligence in an existing research financing ecosystemRodrigo DuranCEO,Centre of A

266、rtificial IntelligenceKey takeaways:Challenges in Chile around AI for science are multifaceted;primarily there is a lack of funding,resources,infrastructure and capacity and skills for AI.Priorities for AI have not been identified at the national scale and universities may be working in silos.Whethe

267、r a unified vision for AI for science will exist in the near future in Chile is not yet clear.Chile gained a National Artificial Intelligence Policy in 2021,after a two-year formulation process in which more than 1,300 people participated(MinCiencia,2021).The policy is formulated in three pillars:en

268、abling factors,R&D,and governance and ethics.The proposed guidelines have a ten-year scope and involve several public and private agencies,which are coordinated by the Ministry of Science for these purposes.It must be noted,however,that the policy is not a binding instrument;the guidelines are not e

269、xplicit mandates but proposed courses of action,which implies certain enforcement difficulties.In this sense,the policy also does not define priorities in the area of R&D project financing in any significant way.The larger research financing ecosystemThe Chilean research and development(R&D)ecosyste

270、m is relatively small compared to the average in the Organisation for Economic Co-operation and Development(OECD,no date).The percentage of Chiles gross domestic product allocated to R&D amounts to 0.36percent,while in the OECD it is 2.68 percent,meaning the relative investment in Chile is seven tim

271、es lower.At the same time,the system is highly dependent on public funding,which represents 57 percent of total investment(MinCiencia,no date a).In nominal terms,in 2021 total investment reached USD1.138 billion,USD648 million of which was public investment.These amounts represent the total investme

272、nt in R&D,including talent training,basic and applied research and technology transfer.Forty-one percent of public investment is managed through the National R&D Agency(ANID),which reports to the Ministry of Science and Technology,while 15.5 percent is resources invested by the universities and come

273、s from the national budget through fiscal contributions or undergraduate university 37tuition subsidies(DIPRES,2023).The remaining 30 percent depends on various agencies with specific mandates,such as the Development Corporation or Public Technological Institutes in specific areas such as fisheries,

274、agriculture or aerospace research.International contributions,for example from observatories,are included in the ANID amount.Public funding to researchThe Chilean public funding system covers the whole researcher career,starting at the formation of advanced human capital,its insertion into industry

275、or academia,the development of long-term individual and associative research projects,as well as infrastructure for centres and universities(MinCiencia,no date b).All of the above is financed through competitive calls,with award rates that vary between 8 percent and 30 percent depending on the instr

276、ument(ANID,2022).The evaluation of the projects is carried out by national academic peers,grouped in study groups that are nominated by collegiate scientific committees representative of the different sectors that participate in the ecosystem(universities,research centres,scientific societies and ac

277、ademia).Currently about 1,500 national researchers participate in 52 study groups,and 120 international peer reviewers evaluate the largest competitions(over USD1 million)(ANID,no date).Local research,however,lacks significant targeting and prioritization mechanisms as well as mandates to prioritize

278、.A full 87 percent of public investment in R&D USD564 million is allocated to open skies projects,whether for the formation of advanced human capital or for individual or group research(MinCiencia,no date a).The remaining 13 percent of public R&D investment is mainly housed in the Public Technologic

279、al Institutes,which have specific mandates from the government.This freedom of research transcends public funding and is also a differentiating element of the university ecosystem,composed of 56 universities,where more than 80 percent of the national knowledge-generating community is concentrated(Mi

280、nCiencia,no date b).In summary,the Chilean R&D ecosystem is small compared with the OECD average,with little prioritization in the allocation of resources and high dependence on public funding.Nevertheless,it has solid and transparent mechanisms for the evaluation of highly competitive projects for

281、the entire trajectory of researchers development,oriented mainly to individual research projects.The impact of Chilean publications measures close to the OECD average,and thus the impact achieved per dollar of investment goes well beyond the average.The arrival of artificial intelligence In terms of

282、 prioritization of sectors and funding practices,the Chilean R&D ecosystem faces challenges from AI.Being a highly atomized system in terms of project evaluation,many evaluators are not trained to properly assess the impact that the use of AI or machine learning tools can have on research,so more or

283、thodox approaches outside of the science,technology,engineering and mathematics(STEM)disciplines are likely to be prioritized.On the other hand,in the absence of prioritization or targeting mechanisms in specific sectors,the development of these competencies in the academic community depends profoun

284、dly on what the host institutions mainly universities do.However,the lack of 38base funds for universities in this area means they need to prioritize other policies rather than the continuous training of their academic staff.There is no mandate for universities to move in this direction,nor are ther

285、e any competitive mechanisms to encourage work along these lines.In this sense,the integration of AI tools in interdisciplinary research depends on the ability and possibility of researchers to articulate around specific projects for particular funding calls which must be evaluated by peers who do n

286、ot have the tools to understand their impact or else focus on particular STEM study groups.This phenomenon means that interdisciplinary projects using AI compete for funds with AI-focused R&D projects,which ultimately may discourage the AI community from collaborating with other disciplines.Addressi

287、ng AI governance issues has led to more international collaboration which has encouraged academic collaboration.Training and talentIn terms of training and retention of talent,since 2019 there has been a relative increase of 15 percent in funding for the training of advanced human capital at the loc

288、al level,with a decrease of 12 percent in funding for masters and doctoral degrees abroad(ANID,no date).This is consistent with the maturation process of the local university system in general.However,in disciplines such as AI it represents a challenge,since the community is less mature and therefor

289、e there is less quality supply than in disciplines like astronomy or biochemistry.This means that the speed at which the community has been growing is decreasing,which limits the possibilities for interdisciplinary research.Similarly,the growing interest of the private and public sector in the adopt

290、ion of AI tools at the international level has generated a significant increase in the demand for advanced human capital,which means that the salaries offered by academic research careers are less competitive than five years previously.Consequently there is a shortage due to better working condition

291、s outside the academy.Although the talent gap that will be faced in the future seems evident,there are no concrete efforts on the part of the private sector to significantly promote talent development on a national scale.Infrastructure and dataIn terms of infrastructure,Chile lacks national laborato

292、ries or big facilities with open access to the academic community.The development of AI models requires access to computing infrastructure,either physical or cloud,which is increasingly expensive due to the generalized increase in demand.This lack can be a significant impediment to the adoption of A

293、I tools in an interdisciplinary manner,or a concentration of tools in university institutions with the resources to fund them.The growing interest of the private and public sector in the adoption of AI tools at the international level has generated a significant increase in the demand for advanced h

294、uman capital,which means that the salaries offered by academic research careers are less competitive than five years previously.39Data access and governance for AI systems is also a structural weakness of the local system.A policy of open access to state-funded research data started in 2022,but the

295、academic community is still reluctant to embrace this openness.There is no culture of standardization of data formats,which means that in many disciplines curatorial work is required prior to their availability.This lack of standards is also reflected in privacy and access policies,which depend on w

296、hat is established by each university or even faculty within the university.All of the above translates into a substantive challenge for the adoption of AI in an interdisciplinary manner.ReferencesANID.No date.ANID-GITHUB.GitHub.https:/ 21 February 2024).(In Spanish.)ANID.2022.Compendio Estadstico 2

297、0182022 Statistical Compendium 2018-2022)Santiago,National R&D Agency.https:/anid.cl/anid-publica-actualizacion-de-su-compendio-estadistico/.(In Spanish).DIPRES.2023.Presupuesto 2023 Budget 2023.Santiago,Budget Office.https:/www.dipres.gob.cl/597/w3-multipropertyvalues-15199-35324.html.(In Spanish.)

298、MinCiencia.No date a.Investigacin y desarrollo(I+D)Research and Development(R+D).Observa.Santiago,Ministry of Science,Technology,Knowledge and Innovation.https:/www.observa.minciencia.gob.cl/indicadores/investigacion-y-desarrollo-id(Accessed 21 February 2024).(In Spanish.)MinCiencia.No date b.Visual

299、izacin interactiva del sistema Interactive system visualisation.Observa.Santiago,Ministry of Science,Technology,Knowledge and Innovation.https:/www.observa.minciencia.gob.cl/sistema(Accessed 21 February 2024).(In Spanish.)MinCiencia.2021.Poltica Nacional de Inteligencia Artificial National Artificia

300、l Intelligence Policy.Santiago,Ministry of Science,Technology,Knowledge and Innovation.https:/www.minciencia.gob.cl/areas/inteligencia-artificial/politica-nacional-de-inteligencia-artificial/.(In Spanish.)OECD.No date.OECD.Stat.Paris,Organisation for Economic Co-operation and Development.https:/stat

301、s.oecd.org/(Accessed 21 February 2024).40CHINAPromoting the Artificial Intelligence for Science approachGong Ke Executive Director of the Chinese Institute for the New Generation Artificial Intelligence Development StrategiesLiu XuanResearch Fellow of the National Academic of Innovation Strategy,CAS

302、TKey takeaways:The government in China is supporting the integration of AI across different fields of science through programs and infrastructure.China is active on the international front regarding AI technologies and has achieved the development of platforms and software supporting AI.Artificial I

303、ntelligence for Science(AI4S)is an emerging mode that integrates AI and scientific research.It refers to the use of AI technologies and methods to learn about,simulate,predict and optimize various phenomena and laws in nature and human society.This case study focuses on the example of AI4S in China,

304、exploring the impact of machine learning and AI on the scientific system.The Chinese government attaches great importance to AI4S,promoting innovations in AI algorithms and models oriented towards major scientific problems.They have established open platforms in typical research areas of AI4S,encour

305、aged academic institutions to open their data resources and set norms for ethical conduct with AI4S.At the national and local government levels in China,policy initiatives in the field of AI4S are mainly as follows.Special research programs and infrastructureIn March 2023,the Ministry of Science and

306、 Technology,in collaboration with the National Natural Science Foundation of China,launched a special initiative called the Implementation Plan for Scientific Research Driven by Artificial Intelligence(20222025)to support the adoption of AI tools in basic sciences such as mathematics,physics,chemist

307、ry and astronomy.The intention is to address major challenges such as climate change,the energy transition,drug development,genetic research,biological breeding and new materials.The projects include cross integration of AI and materials science,cross integration of AI and basic mathematics,cross in

308、tegration of AI and information technology,cross integration of AI and life sciences,and cross integration of AI and ethical and social issues(Ministry of Science and Technology,2023a).41Meanwhile,the Ministry of Science and Technology is leveraging the national project Science and Technology Innova

309、tion 2030 Next Generation of Artificial Intelligence(Ministry of Science and Technology,2021)as a driver to build open intelligent computing power infrastructure,facilitate the active opening of data resources from various sectors,and generate policy synergy to advance AI4S.In April 2023,the Shangha

310、i government supported Shanghai Jiao Tong University in launching the Open Platform of AI4S with Open-Sourced Models and Scientific Data(Jiefang Daily,2023).Ethics governance and regulationsIn 2017,the Chinese national plan for developing AI was released(State Council,2017),in which it is pointed ou

311、t that AI has both technical and social features.Two committees were established by the Chinese government to implement the plan:a technical committee and a governance committee.The governance committee is composed of relevant experts from universities,research institutes and enterprises.It has rele

312、ased documents such as Governance Principles of the Next Generation of AI Developing Responsible AI(National Next Generation AI Governance Professional Committee,2019)and Next Generation Artificial Intelligence Ethics Standards(National Next Generation AI Governance Professional Committee,2021).In 2

313、021,the Chinese government also established the National Science and Technology Ethics Committee,which has released a list of high-risk AI research and development areas(Ministry of Science and Technology,2023b).This ethics committee has a subcommittee dedicated to AI,consisting of experts from rele

314、vant sectors and providing professional consultations to the State Council for the formulation of Chinas technology ethics policies.Finally,in 2023,after a month-long online open consultation,the State Cyberspace Administration of China along with multiple departments jointly issued Interim Measures

315、 for the Management of Generative Artificial Intelligence Services,marking the first regulatory policy for Chinas AI-generated content industry(The Cyberspace Administration of China,2023a).The international perspective China has an open and proactive attitude towards international cooperation in AI

316、.It supports the United Nations irreplaceable role in international AI governance,and actively participates in activities organized by bodies such as the United Nations Educational,Scientific and Cultural Organization(UNESCO),International Telecommunication Union(ITU),World Health Organization(WHO),

317、United Nations Industrial Development Organization(UNIDO)and United Nations Development Programme(UNDP).China has invited United Nations bodies representatives to join relevant AI conferences and forums in the country.In 2023,after a month-long online open consultation,the State Cyberspace Administr

318、ation of China along with multiple departments jointly issued Interim Measures for the Management of Generative Artificial Intelligence Services,marking the first regulatory policy for Chinas AI-generated content industry42In November 2023,the Chinese government launched the Global Initiative on AI

319、Governance,outlining eleven proposals that prioritize a people-centric approach and respect for the sovereignty of other countries.It was emphasized that China is willing to engage in communication,exchange and cooperation with all parties on global AI governance,promote the benefits of AI technolog

320、y to all mankind,and propose constructive solutions to the development and governance issues of AI that are of wide concern to all parties in the new era(The Cyberspace Administration of China,2023b).Chinas promotion of non-governmental exchanges and cooperation is exemplified in the August 2023 Int

321、ernational Young Scientist Salon AI for Science Taking Place in the Current Scientific and Technological Revolution,organized by the China Association for Science and Technology in Shanghai.Young scientists from eight countries including the United Kingdom,Greece and Germany participated in the disc

322、ussion and exchange(CAST,2023).Shanghai also hosted the January 2024 World Digital Education Conference,jointly organized by the Chinese Ministry of Education,the National Committee of UNESCO and the Shanghai Government.This conference focused on the theme of Digital Education:Application,Sharing,an

323、d Innovation,with subthemes of enhancing teacher digital literacy and competence;digitizing education and building a learning society;evaluating global trends and indices in digital education development;AI and digital ethics;challenges and opportunities of digital transformation for basic education

324、;and digital governance in education(Ministry of Education,2024).Overall development trendBased on relevant research reports and literature review(AI for Science Institute of Beijing,2023),the overall trend in the field of AI4S in China can be summarized as follows.Chinese academic institutions,univ

325、ersities and leading AI enterprises are proactive in the AI4S field,with internationally influential achievements such as MEGA-Protein,Pengcheng Shen Nong,Shanghai AI Labs FengWu and PanGu Weather(Fang,X.,et al.,2022)(K.Bi,et al.,2023).Abundant open scientific research data resources have accumulate

326、d for AI4S,with open-source data found in meteorology,astronomy and high-energy physics(Tan,S.et al.,2023).A considerable number of AI4S algorithms and foundational software have also emerged,including Huaweis MindSpore Science,Baidus PaddleScience,DP Technologys DeePMD and Zhipuais GLM,providing ri

327、ch datasets,foundational models and specialized tools for AI4S research(Huawei,2017).AI4S applications are being explored in various fields including life science,material science,energy science,electronic engineering and computer science,earth and environmental science,and industrial simulation.In

328、particular,institutions represented by the likes of Baidu and Huawei are actively promoting the development of AI4S industrial practice.43Foundational Artificial Intelligence for Science softwareBaidus PaddlePaddle began planning technical forms and product routes in the AI4S field as early as 2019.

329、It has since released the biological computing platform PaddleHelix,the quantum computing platform PaddleQuantum,and the scientific computing platform PaddleScience.Baidu has collaborated on exemplary projects with multiple universities and research institutions and launched the PaddlePaddle AI4S Co

330、Creation Program to build an ecological business opportunity.In May 2023,Baidu published a paper in the journal Nature opening up numerous possibilities for the integration of AI into fields such as biology and healthcare(Fang,X.,et al.,2022).Huawei,meanwhile,has launched PanGu large models for drug

331、 molecules,meteorology and ocean waves.Among them,the PanGu drug molecule large model can improve the screening speed of small molecule compounds,greatly improve research and development efficiency,and explore more possible combinations of molecular elements at lower costs.In July 2023,the research

332、results of the PanGu meteorological large model of Huawei Cloud were published in the journal Nature,and it is the first AI model to surpass traditional numerical forecasting methods in accuracy(K.Bi,et al.,2023).References Ministry of Science and Technology.2023a.科學技術部，“人工智能驅動的科學研究”專項實施方案（20222025年

333、）：https:/ Chinese).Ministry of Science and Technology.2021.科學技術部，科技創新2030“新一代人工智能”重大項目：https:/ Chinese).Jiefang Daily.2023.解放日報，上海建開源平臺推動“第五范式”：https:/ Chinese).State Council.2017.國務院，新一代人工智能發展規劃：https:/ National Next Generation AI Governance Professional Committee.2019.國家新一代人工智能治理專業委員會，新一代人工智能治理原則發展負責任的人工智能：https:/ Chinese).National Next Generation AI Governance Professional Committee.2021.國家新一代人