Shared Post-Human Imagination: Human-AI Collaboration in Media Creation

共享的后人类想象力：媒体创作中的人机协作

• 批准号: AH/Z50564X/1
• 负责人: Szilvia Ruszev
• 金额: $ 24.54万
• 依托单位: Bournemouth University
• 依托单位国家: 英国
• 项目类别: Research Grant
• 财政年份: 2024
• 资助国家: 英国
• 起止时间: 2024 至 无数据
• 项目状态: 未结题
• 来源: https://gtr.ukri.org/projects?ref=AH%2FZ50564X%2F1
• 关键词: Shared; Post; Human; Imagination; AI

The rapid advancements in AI and ML have revolutionized the media landscape, opening new possibilities for content creation. Generative AI tools (GAIT) for creation and manipulation of text, image and sound such as ChatGPT or Midjourney became accessible for everyone and hold the promise of changing the ways in which media production can be effective and creative. However, these technological breakthroughs have also come with moral, ethical and legal challenges and risks that need to be addressed. Concerns regarding copyright and job security have been amplifying, as the recent strikes in Hollywood show. A rising number of publications and position papers in the UK and internationally strive to address the challenges and risks offered by GAIT. While the public discourse around AI fluctuates between the extremes of enthusiasm and dystopia, questions about bias, transparency, accountability, safety, justice and trustworthiness need to be addressed in their complexity.The proposed research aims to investigate responsible AI(RAI) in the context of media production, focusing on collaboration, creativity and representation. The research outputs seek to foster feedback loops among developers, users and policymakers in order to develop RAI practices. In detail:While AI is not yet on-par with human capabilities, we use AI algorithms as tools. Yet, it is productive to reflect on the implications of a human-AI collaboration in the context of RAI. In other words, we will ask questions such as: Who has access to and who has control over GAIT? Also, how can we create feedback loops to progress not only human media creation processes but the underlying dynamic AI system?Similarly, the notion of creativity has to be re-evaluated in the context of RAI. Questions of extractive database practices, copyright, intellectual property and the displacement of human labour are in the foreground. Yet, technologically aided serendipity offers new creative possibilities.And finally, based on existing studies, we are increasingly aware of the ingrained bias in large databases used to train GAIT. Our research aims to understand how these biases materialise through the use of GAIT. Furthermore, by applying the knowledge and methodologies of media studies research, we will propose ways in which socio-cultural complexities can be productively addressed avoiding the (re)production of stereotypical representations of social categories such as race, class, gender, etc.In order to address these research aims, we will apply a hybrid, practice-theory methodology. In four workshops, we will simulate the process of media production (script-writing, image generation, audio/music generation and post-production), experiment with significant GAIT and subsequently reflect on our experience within the context of our research questions.Research outputs will be:A report with information, analysis and recommendations for best practices of the use of GAIT in media creation in the context of RAI.A draft sci-fi short film created using GAIT reflecting on the research questions. (full version to be completed outside the research time).The outputs will be presented in a closing event to a group of media creators, policy makers and developers to foster dialogue and concrete actions.

人工智能和ML的快速发展彻底改变了媒体格局，为内容创作打开了新的可能性。用于创建和处理文本、图像和声音的生成性人工智能工具(步态)，如ChatGPT或MidTrip，变得每个人都可以使用，并有望改变媒体制作可以有效和创造性的方式。然而，这些技术突破也伴随着需要解决的道德、伦理和法律挑战和风险。正如最近好莱坞的罢工事件所表明的那样，人们对版权和工作安全的担忧一直在放大。英国和国际上越来越多的出版物和立场文件努力应对GATIT带来的挑战和风险。虽然公众对人工智能的讨论在热情和反乌托邦的极端之间波动，但关于偏见、透明度、问责、安全、正义和可信度的问题需要在它们的复杂性中得到解决。拟议的研究旨在调查媒体生产背景下的负责任人工智能，重点关注协作、创造力和代表性。研究成果旨在促进开发人员、用户和政策制定者之间的反馈循环，以便开发RAI实践。具体来说：虽然人工智能还不能与人类的能力相提并论，但我们将AI算法作为工具。然而，在RAI的背景下反思人类与人工智能合作的影响是富有成效的。换句话说，我们会问这样的问题：谁可以访问步态，谁可以控制步态？此外，我们如何创建反馈循环，以不仅推进人类媒体创造过程，而且推进潜在的动态人工智能系统？类似地，创造力的概念必须在RAI的背景下重新评估。摘录数据库做法、著作权、知识产权和人力转移等问题都在前台。然而，技术辅助的机缘巧合提供了新的创造性可能性。最后，基于现有的研究，我们越来越意识到用于训练步态的大型数据库中根深蒂固的偏见。我们的研究旨在了解这些偏见是如何通过步态的使用来实现的。此外，通过应用传媒学研究的知识和方法，我们将提出能够有效地解决社会文化复杂性的方法，以避免(重新)再现诸如种族、阶级、性别等社会类别的刻板印象。为了解决这些研究目标，我们将采用实践-理论混合的方法。在四个研讨会中，我们将模拟媒体制作的过程(剧本编写、图像生成、音频/音乐生成和后期制作)，试验重要的步态，然后在我们的研究问题的背景下反思我们的经验。研究产出将是：一份关于在RAI背景下使用步态在媒体创作中使用最佳做法的信息、分析和建议的报告。一份使用步态创作的科幻短片草稿，反映研究问题。(完整版将在研究时间以外完成)。成果将在闭幕活动中向一群媒体创作者、政策制定者和开发人员介绍，以促进对话和具体行动。