February 28, 2023
The Workshop on Legal and Ethical Governance Challenges Faced by Big Models
Location: Virtual
Date: February 28, 2023
Timetable:
| China Standard Time (UTC+8) | Central European Time (UTC+1) | Eastern Standard Time (UTC-5) | Item | Speaker |
|---|---|---|---|---|
| 21:00 – 21:05 | 14:00 – 14:05 | 08:00 – 08:05 | Opening | Rui Guo, Renmin University of China |
| Session 1: Big Models and Intellectual Property | ||||
| 21:05 – 21:25 | 14:05 – 14:25 | 08:05 – 08:25 | Keynote: AI/algorithmic transparency: a new role in the copyright ecosystem?
[slides (opens in new tab), video (opens in new tab)] Abstract: The presentation will discuss selected copyright implications of AI-generated content (such as that by ChatGPT) and focus on the issue of transparency. Currently AI-generated content generally cannot qualify for copyright protection, but it may be difficult to identify the human and/or AI source of digital content. An obligation of disclosure could help create transparency, which is what will be discussed in the presentation. |
Jacques de Werra, Digital Law Center, Université de Genève |
| 21:25 – 22:35 | 14:25 – 15:35 | 08:25 – 09:35 | Panel Discussion: Big Models and Intellectual Property
Abstract: This panel focuses on the complex relationship between big models and intellectual property. It aims to explore the legal and ethical issues that arise from the use of AI-generated content for commercial and academic purposes. The panel will also discuss the copyright implications of training large models and how it affects the rights of the content authors. |
Host: Rui Guo, Renmin University of China Panelist: Xia “Ben” Hu, Rice University Haochen Sun, The University of Hong Kong Maosong Sun, Tsinghua University Jacques de Werra, Digital Law Center, Université de Genève Jiyu Zhang, Renmin University of China |
| Session 2: Big Models and Privacy | ||||
| 22:35 – 22:55 | 15:35 – 15:55 | 09:35 – 09:55 | Keynote: Generative models have the memory of an elephant
[slides (opens in new tab), video (opens in new tab)] Abstract: Modern generative models seem to generalize extremely well! At the same time, we find that they are capable of *memorizing* large amounts of training data. I will describe some of the challenges in defining and measuring memorization in machine learning, and showcase substantial data memorization in state-of-the-art models for language generation, code completion, and even image generation! We’ll discuss some of the implications that such large memorization capabilities have for data privacy, copyright protection, and more. |
Florian Tramèr, ETH Zürich |
| 22:55 – 00:05+1 | 15:55 – 17:05 | 09:55 – 11:05 | Panel Discussion: Big Models and Privacy
Abstract: This panel delves into the crucial issue of protecting user privacy and enterprise content when using big models. It will examine the challenges of safeguarding sensitive information and original training data when using AI models. Additionally, the panel will discuss the significant role of privacy protection in the development of big models and its impact on the availability of high-quality training data in the long term. |
Host: Xing Xie, Microsoft Research Asia Panelist: Vivian Ding, Microsoft Lixin Fan, WeBank Yan Luo, Covington & Burling LLP Florian Tramèr, ETH Zürich Huishuai Zhang, Microsoft Research Asia |
| Session 3: Big Models and Technology Misuse | ||||
| 00:05+1 – 01:35+1 | 17:05 – 18:35 | 11:05 – 12:35 | Panel Discussion: Big Models and Technology Misuse
Abstract: The panel will explore the potential misuse of big models by humans and the pressing need to regulate its use. The discussion will focus on preventing the spread of misinformation, social polarization, and social injustice. The panelists will address the ethical and moral considerations in ensuring the responsible use of big models and minimizing their negative impact on society. |
Host: Rui Guo, Renmin University of China Panelist: Cong Cai, OPO Disability Group Fengming Cui, Harvard University Nathan Sanders, Harvard University Xing Xie, Microsoft Research Asia |
| 01:35+1 – 01:40+1 | 18:35 – 18:40 | 12:35 – 12:40 | Closing | Xing Xie, Microsoft Research Asia |