A Framework for Building Trust in Generative AI Systems

在生成人工智能系统中建立信任的框架

• 批准号: 10074327
• 金额: $ 6.37万
• 依托单位: SENTIENT SPORTS LIMITED
• 依托单位国家: 英国
• 项目类别: Grant for R&D
• 财政年份: 2023
• 资助国家: 英国
• 起止时间: 2023 至 无数据
• 项目状态: 已结题
• 来源: https://gtr.ukri.org/projects?ref=10074327
• 关键词: Framework; Building; Trust; Generative; AI

At SentientSports, we are dedicated to revolutionizing the sports industry using cutting-edge generative AI technology. Our custom large language model is trained on real-time sports data and content to create a user-friendly interface for analyzing data-heavy sports domains. Our platform will provide unique insights and unprecedented access to real-time data for different stakeholders in sports, including fans, teams, agents, and financiers.We recognize that as we develop our technology, it's important to ensure the information we provide is trustworthy and explainable. This is especially challenging in the current era of generative AI, where ensuring accuracy and truthfulness is critical. We believe that the rise of these systems will create technical challenges in utilizing their outputs in our daily lives, and our first hurdle will be to ensure that humans trust the information provided by our AI.Our project involves constructing a trust framework that acts as an intermediary between the user and the LLM, elevating the value proposition by augmenting the reliability of the LLM's raw responses. The trust framework will enable regulated access to external tools such as specialist APIs and data sources, validate responses using a domain knowledge graph, and enhance response content to ensure its credibility, for example, through entity linking and referencing.Our team is uniquely positioned to solve this problem. Our founding team both hold AI PhDs from a top UK university and are world-leading researchers in both AI and sports. Additionally, our chairman is a pioneering professor in the AI space and heads up the UK's largest research hub for trustworthy AI.Our objective is to instil confidence in our users to enable them to make well-informed decisions based on the information we furnish. We acknowledge that the usefulness of our system's outputs will hinge on their credibility, and we are steadfast in our commitment to providing our users with accurate information to facilitate sound decision-making. We are convinced that by creating a trustworthy and comprehensible system, we can establish the benchmark for data analysis in the sports industry and contribute to the larger discourse concerning responsible AI applications.

在SentientSports，我们致力于使用尖端的生成式人工智能技术彻底改变体育行业。我们的定制大型语言模型基于实时体育数据和内容进行训练，以创建用户友好的界面，用于分析数据密集的体育领域。我们的平台将为体育运动中的不同利益相关者提供独特的见解和前所未有的实时数据访问权限，包括球迷、球队、经纪人和金融家。我们认识到，在我们开发技术的过程中，确保我们提供的信息值得信赖和可解释非常重要。这在当前的人工智能时代尤其具有挑战性，在这个时代，确保准确性和真实性至关重要。我们认为，这些系统的兴起将在我们的日常生活中利用它们的输出方面带来技术挑战，我们的第一个障碍将是确保人类信任我们的AI提供的信息。我们的项目涉及构建一个信任框架，作为用户和LLM之间的中介，通过增强LLM原始响应的可靠性来提升价值主张。信任框架将允许对专业API和数据源等外部工具进行受监管的访问，使用领域知识图验证响应，并增强响应内容以确保其可信度，例如通过实体链接和引用。我们的团队在解决这一问题方面具有独特的优势。我们的创始团队都拥有英国顶尖大学的AI博士学位，并且是AI和体育领域的世界领先研究人员。此外，我们的董事长是人工智能领域的先驱教授，并领导着英国最大的可信人工智能研究中心。我们的目标是为我们的用户注入信心，使他们能够根据我们提供的信息做出明智的决定。我们认识到，我们系统产出的有用性将取决于其可信度，我们坚定地致力于向用户提供准确信息，以促进合理决策。我们相信，通过创建一个值得信赖和可理解的系统，我们可以建立体育行业数据分析的基准，并为有关负责任的人工智能应用的更广泛的讨论做出贡献。