A trustworthy generative AI for producing personalised L&D

今天，教育是以一对多的方式提供的：一门课程给许多学生，这导致课程只能部分满足学生的需求。另一方面，针对每个学生的需求制作课程（即一对一的替代方案）既缓慢又昂贵。这并不是课程规划所独有的；学习和发展（L&D）系统的其他组成部分（例如，辅导、内容生成、评估和编码）也可以从这种个性化中受益。我们相信人工智能——包括机器学习（ML）——可以帮助解决这个问题。特别是，考虑到生成式人工智能的最新进展：ML模型可以接受简单的指令，并生成代码、文本、图像和视频等人工制品，直到最近，这些都只能由人类生产。然而，为了使这样的ML模型/产品成功，它们需要得到用户/涉众的信任。这对于生成人工智能模型尤其重要，因为生成人工智能模型通常非常大，因此很难向利益相关者解释。因此，本项目旨在组成一个联合体，利用其集体的专业知识提出解决方案：1。在“ML基础”中为教育/L&D构建生成式AI模型，2。开发一个框架来评估和获得利益相关者对我们模型的信任。验证1和2（例如，通过展示我们的L&D产品的成功采用）。在第一阶段，我们的重点将是建立适当的联盟（由值得信赖的人工智能、机器学习研究/工程、L&D业务和机器学习产品/用户体验方面的专业知识组成），并制作一份技术报告，概述我们对第二阶段项目的建议。根据普华永道（PWC）最近的一项调查，74%的首席执行官担心他们所需的关键技能在市场上的可用性。虽然一家典型的大公司每年在L&D上的花费超过100万美元，但一名典型的员工每月参与L&D的时间还不到一次。我们对人才市场的调查显示，符合目标的内容、改进的用户体验、个性化推荐和专家指导可以解决当前的不足。这些类似于人工智能在其他领域提供的变化。因此，我们的使命是为L&D空间带来这样的变化。改进的L&D服务（例如，更加个性化和负担得起）将具有深远的社会和商业价值；它将在提高国家和国际生产力方面发挥关键作用。它与政府的战略举措，如“提升水平”和“终身技能”保持一致。

Today, education is offered in a one-to-many fashion: One course for many students, which results in courses that are only partially fit for students' needs. On the other hand, producing courses specific to each student's needs (i.e., a one-to-one alternative) is slow and expensive. This is not unique to course planning; other components of learning and development (L&D) systems (e.g., tutoring, content generation, assessment, and coding) can also benefit from such personalisation. We believe that AI -- including machine learning (ML) -- can help solve this. Particularly, given the recent advances in generative AI: ML models that can take simple instructions and generate artefacts such as code, text, images and videos that until recently, could only be produced by humans.In order for such ML models/products to succeed, however, they need to be trusted by their users/stakeholders. This is particularly important for generative AI models that are usually very large and hence difficult to explain to stakeholders. Therefore, this project aims to form a consortium and use its collective expertise to propose a solution for:1. Building generative AI models for education/L&D in "ML Fundamentals",2. Developing a framework to evaluate and gain stakeholders' trust in our model(s)3. Validating 1 & 2 (e.g., by showing successful adoption of our L&D product).During phase-1, our focus will be on building the appropriate consortium (consisting of expertise in trustworthy AI, ML research/engineering, L&D business, and ML product/UX), and producing a technical report that outlines our proposal for the phase-2 project.According to a recent survey by PWC, 74% of CEOs were concerned about the availability of key skills they need, in the market. While a typical big company spends more than $1m/year on L&D, a typical employee engages less than once a month with them. Our surveys of the talent market shows that fit-for-purpose content, improved UX, personalised recommendation, and expert mentoring can address the current shortcomings. These are similar to the changes that AI offered in other domains. Therefore, we are on a mission to bring such a change to the L&D space.An improved L&D offering (e.g., more personalised and affordable) will be of profound societal and commercial value; it will play a critical role in improving national and international productivity. It is aligned with strategic government initiatives such as "Levelling up" and "Skills for Life".