Ressources
Explorez une large gamme de ressources sur le GCED afin d’approfondir votre compréhension et de renforcer vos activités de recherche, de plaidoyer, d’enseignement et d’apprentissage.
12 résultats trouvés
Challenging Systematic Prejudices: An Investigation into Bias Against Women and Girls in Large Language Models Année de publication: 2024 Auteur: Daniel Van Niekerk | Maria Peréz Ortiz | John Shaw-Taylor | Davor Orlic | Ivana Drobnjak | Jackie Kay | Noah Siegel | Katherine Evans | Nyalleng Moorosi | Tina Eliassi-Rad | Leone Maria Tanczer | Wayne Holmes | Marc Peter Deisenroth | Isabel Straw | Maria Fasli | Rachel Adams | Nuria Oliver | Dunja Mladenić | Urvashi Aneja | Madeleine Janicky Auteur institutionnel: UNESCO | International Research Centre on Artificial Intelligence (IRCAI) This study explores biases in three significant large language models (LLMs): OpenAI’s GPT-2 and ChatGPT, along with Meta’s Llama 2, highlighting their role in both advanced decision-making systems and as user-facing conversational agents. Across multiple studies, the brief reveals how biases emerge in the text generated by LLMs, through gendered word associations, positive or negative regard for gendered subjects, or diversity in text generated by gender and culture. The research uncovers persistent social biases within these state-of-the-art language models, despite ongoing efforts to mitigate such issues. The findings underscore the critical need for continuous research and policy intervention to address the biases that exacerbate as these technologies are integrated across diverse societal and cultural landscapes. The emphasis on GPT-2 and Llama 2 being open-source foundational models is particularly noteworthy, as their widespread adoption underlines the urgent need for scalable, objective methods to assess and correct biases, ensuring fairness in AI systems globally. 
International Forum on AI and Education: Steering AI To Empower Teachers and Transform Teaching, 5–6 December 2022; Analytical Report Année de publication: 2023 Auteur: Fengchun Miao | Kelly Shiohira | Zaahedah Vally | Wayne Holmes Auteur institutionnel: UNESCO | JET Education Services The International Forum on AI and Education has contributed ‘to peace and security by promoting collaboration among the nations through education, science and culture’ (UNESCO, 1945). The Forum has now become the world’s leading event promoting knowledge-sharing, the understanding of peoples, and the achievement of international agreements, in the fast-developing and increasingly impactful field of AI and education. The fourth edition of the International Forum on Artificial Intelligence and Education aimed to foster knowledge sharing specifically on how to steer the design and use of AI to empower teachers and to transform teaching methodologies within the broad framework of digital transformation of education. The Forum sought to bring together a range of expertise and experiences across the globe, and, in keeping with UNESCO priorities, a special focus was placed on Africa. A total of 16 national strategies were presented at the Forum by the various national ministers and representatives invited to attend. The national strategies shared during the Forum also unveiled the varying levels of preparedness and policy responses toward AI across different regions, a reminder that the pre-existing digital divide still underpins the system-wide uptake and integration of AI in education. Therefore, not all of the strategies were directly related to or containing AI because some areas that were represented at the conference are still at the nascent stages of AI Strategy development. The presentations of national initiatives and strategies related to AI from these countries shed light on the general awareness of policy-makers on the impact of AI in education and the commitment of national governments to fostering AI competencies among students and teachers. This report focus on the following key themes: national strategies on AI and education; critical reviews of roles of AI in the digital transformation of education; ethical principles and their implementation with a specific focus on gender equality; AI competencies for teachers, and notable algorithms or AI platforms and AI-informed pedagogies. The report concludes with considerations for the future based on the authors’ own analysis of the key role of human teachers, steering the human-centered approach, mainstreaming gender equity, designing education-specific AI models and innovative pedagogy, and ensuring human agency in defining problems and designing solutions.a 
Руководство по использованию генеративного искусственного интеллекта в образовании и научных исследованиях Année de publication: 2023 Auteur: Fengchun Miao | Wayne Holmes Auteur institutionnel: UNESCO Данное руководство направлено на поддержку разработки соответствующих правил, политик и программ развития человеческого потенциала с целью обеспечения того, чтобы генеративный искусственный интеллект (ИИ) стал инструментом, действительно приносящим пользу и расширяющим возможности учителей, учащихся и исследова телей. В нем объясняются методы искусственного интеллекта, используемые в генеративном ИИ, а также приводится список моделей СРТ, которые становятся общедоступными, особенно модели с открытым исходным кодом. Руководство также открывает обсуждение о появлении ЕсІСРТ - моделей генеративного ИИ, обученных на конкретных данных для образовательных целей. Кроме того, в нем суммируются некоторые ключевые противоречия вокруг генеративного ИИ - от ухудшения цифрового неравенства до гомогенизации мнений, от более сложных дипфейков до вопросов авторского права. Исходя из гуманистического подхода, руководство предлагает ключевые шаги по регулированию инструментов генеративного ИИ, включая обязательную защиту конфиденциальности данных и установку возрастных ограничений для самостоятельного взаимодействия с платформами генеративного ИИ. Для обеспечения правильного использования инструментов в сфере образования и исследований, в руководстве затронуты вопросы этики и проектирования педагогического процесса и заложен подход, ориентированный на человека и учитывающий возрастные особенности. 
International Forum on AI and the Futures of Education: Developing Competencies for the AI Era Synthesis Report; 7-8 December 2020 Année de publication: 2021 Auteur: Fengchun Miao | Wayne Holmes Auteur institutionnel: UNESCO This synthesis report has been developed by the UNESCO Unit for Technology and Artificial Intelligence in Education within the Future of Learning and Innovation Team, drawing on the International Forum on AI and the Futures of Education held in Beijing and simultaneously online from 7 to 8 December 2020.The Forum underlined the importance of reviewing the very purpose of education, together with the opportunity for real transformation, and the role that AI might play. It is widely acknowledged that current educational practices and educational environment are very rigid, and that the future of education should be more flexible and responsive to changing circumstances and innovation. National education authorities should identify what skills young people need to enable them to live and thrive in the new realities of a constantly changing world, and what digital transformation makes possible in the national and international context. In short, new education models are needed to put students at the centre, to move away from a focus on memorizing content, to integrate the digital and the analogue, and to foster human cognitive, socioemotional and critical skills, all of which might – with foresight and careful attention – be enabled by AI and other digital technologies. 
AI and Education: Guidance for Policy-Makers Année de publication: 2021 Auteur: Fengchun Miao | Wayne Holmes | Ronghuai Huang | Hui Zhang Auteur institutionnel: UNESCO Artificial intelligence (AI) is envisioned as a new tool to accelerate the progress towards the achievement of SDG 4. Policies and strategies for using AI in education are central to maximizing AI’s benefits and mitigating its potential risks. Fostering AI-ready policy-makers is the starting point of the policy development process.This publication offers guidance to policy-makers in understanding AI and responding to the challenges and opportunities in education presented by AI. Specifically, it introduces the essentials of AI such as its definition, techniques, technologies, capacities and limitations. It also delineates the emerging practices and benefit-risk assessment on leveraging AI to enhance education and learning, and to ensure inclusion and equity, as well as the reciprocal role of education in preparing humans to live and work with AI.The publication summarizes three approaches to the policy responses from existing practices: independent approach, integrated approach and thematic approach. In a further step, it proposes more detailed recommendations and examples for planning AI and education policies, aligned with the recommendations made in the 2019 Beijing Consensus on AI and Education. 人工智能与教育: 政策制定者指南 Année de publication: 2021 Auteur: Fengchun Miao | Wayne Holmes | Ronghuai Huang | Hui Zhang Auteur institutionnel: UNESCO 该文件为政策制定者提供关于人工智能在教育领域内应用于发展的相关知识，包括其定义与技术一级对教学和教育的影响，并为政策制定者提供相关政策建议。 ارشادات استخدام الذكاء الاصطناعي التوليدي في التعليم والبحث Année de publication: 2023 Auteur: Fengchun Miao | Wayne Holmes Auteur institutionnel: UNESCO تهدف هذه الإرشادات إلى دعم تخطيط اللوائح والسياسات وبرامج تنمية القدرات البشرية المناسبة ً لضمان أن يصبح الذكاء الاصطناعي التوليدي أداة تفيد المعلمين والمتعلمين والباحثين وتمكنهم حقا. وهو يشرح تقنيات الذكاء الاصطناعي التي يستخدمها الذكاء الاصطناعي التوليدي ويرسم قائمة بنماذج ً المحولات التوليدية المدربة مسبقا المتاحة للجمهور، خاصة تلك التي تخضع لترخيص مفتوح المصدر. كما ً ً حول ظهور نماذج المحولات التوليدية التعليمية المدربة مسبقا - التي يتم تدريبها باستخدام يفتح نقاشا بيانات محددة لخدمة الأغراض التعليمية. علاوة على ذلك، فهو يلخص بعض الخلافات الرئيسية حول ً الذكاء الاصطناعي التوليدي، بدءا من تفاقم الفقر الرقمي إلى تجانس ً الآراء، ومن التزييف العميق إلى قضايا حقوق الطبع والنشر. واستنادا إلى رؤية إنسانية، تقترح الإرشادات خطوات أساسية لتنظيم أدوات الذكاء الاصطناعي التوليدي، بما في ذلك فرض حماية خصوصية البيانات ووضع حد عمري للمحادثات المستقلة مع منصات الذكاء الاصطناعي التوليدي. ً لتوجيه الاستخدام السليم للأدوات في التعليم والبحث، تقترح هذه الإرشادات نهج ً ا مناسبا للعامل البشري والعمر للتحقق من الصحة الأخلاقية وعمليات التصميم التربوي. Guidance for Generative AI in Education and Research Année de publication: 2023 Auteur: Fengchun Miao | Wayne Holmes Auteur institutionnel: UNESCO This Guidance aims to support the planning of appropriate regulations, policies and human capacity development programmes to ensure that generative artificial intelligence (GenAI) becomes a tool that genuinely benefits and empowers teachers, learners and researchers. It explains the Al techniques used by GenAI and maps out a list of GPT models that are made publicly available, especially those under open-source licences. It also opens a discussion on the emergence of EdGPT - GenAI models that are trained with specific data to serve educational purposes. Furthermore, it summarizes some of the key controversies around GenAI, from worsening digital poverty to the homogenization of opinions, and from deeper deepfakes to issues of copyright. Based on a humanistic vision, the Guidance proposes key steps for the regulation of GenAI tools, including mandating the protection of data privacy and setting an age limit for independent conversations with GenAI platforms. To guide the proper use of the tools in education and research, this Guidance proposes a human-agent and age-appropriate approach to the ethical validation and pedagogical design processes. Guía para el uso de IA generativa en educación e investigación Année de publication: 2023 Auteur: Fengchun Miao | Wayne Holmes Auteur institutionnel: UNESCO Esta guía apunta a apoyar la planificación de regulaciones, políticas y programas de desarrollo de capacidades humanas apropiados para garantizar que la lAGen se convierta en una herramienta que genuinamente beneficie y empodere a docentes, estudiantes e investigadores. Explica las técnicas de IA utilizadas por la lAGen y elabora una lista de modelos de GPT puestos a disposición del público, especialmente aquellos con licencias de código abierto. También abre un debate sobre el surgimiento de EdGPT: los modelos de lAGen entrenados con datos específicos para servir a fines educativos. Además, resume algunas de las controversias clave en torno a la lAGen, desde el empeoramiento de la pobreza digital a la homogeneización de opiniones, y desde los más profundos deepfakes a cuestiones de derechos de autor. Con base en una visión humanista, la guía sugiere pasos clave para la regulación de las herramientas de lAGen, que incluyen la obligación de proteger la privacidad de los datos y el establecimiento de un límite de edad para las conversaciones independientes con las plataformas de lAGen. Para guiar el uso adecuado de las herramientas en la educación y la investigación, esta guía propone un enfoque humano-agente y adecuado a la edad para los procesos de validación ética y diseño pedagógico. Orientations pour l’intelligence artificielle générative dans l’éducation et la recherche Année de publication: 2023 Auteur: Fengchun Miao | Wayne Holmes Auteur institutionnel: UNESCO Ces recommandations ont pour objectif de soutenir l'élaboration de réglementations, de politiques et de programmes visant à développer les capacités humaines, dans le but de garantir que l'intelligence artificielle générative (GenAI) devienne un outil qui profite réellement aux enseignants, aux apprenants et aux chercheurs et leur donne les moyens d'agir. Il explique les techniques d'IA utilisées par la GenAI et dresse une liste des modèles de GPT qui sont mis à la disposition auprès du public, en particulier ceux qui font l'objet d'une licence à source ouverte. Il ouvre également une discussion sur l'émergence de l'EdGPT - les modèles GenAI qui sont entraînés avec des données spécifiques pour se mettre au service des objectifs dans le domaine de l'éducation. En outre, il résume certaines des principales controverses formées autour de l'IA générative, de l'aggravation de la pauvreté numérique à l'homogénéisation des opinions, en passant par les trucages (deepfakes) et les questions liées aux droits d'auteur. Fondé sur une vision humaniste, le présent document propose des étapes-clefs pour la réglementation des outils de la GenAI, notamment en rendant obligatoire la protection de la confidentialité des données et en fixant une limite d'âge pour les conversations autonomes avec les plateformes de la GenAI. Pour accompagner l'utilisation pertinente des outils dans l'éducation et la recherche, ce guide propose une approche holistique et adaptée à tout âge pour les processus de validation éthique et de conception pédagogique.