本課程從不同哲學角度探討機器學習AI系統的對齊問題，其中包括知識論、形上學、認知道德哲學、倫理學、語言哲學，檢視AI開發者擬設計能理解並依循人類價值觀、道德和意向的AI系統可能面臨的挑戰和考量。This course explores the alignment of machine learning AI systems from different philosophical perspectives, including epistemology, metaphysics, cognitive moral philosophy, ethics, and language philosophy, and examines AI developers' plans to design systems that can understand and follow human values, morals, and intentions. Challenges and considerations that AI systems may face.
1. Russell, S., & Norvig, P. (2021). Chap. 1 Introduction. In Artificial intelligence: A modern approach (4th ed.). University of California, Berkeley.
2. Christian, B. R. (2021). Introduction. In The alignment problem: Machine learning and human values. Norton, W. W. & Company, Inc.
3. Gabriel, I. (2020). Artificial intelligence, values, and alignment. Minds & Machines, 30, 411–437. https://doi.org/10.1007/s11023-020-09539-2
4. Christian, B. R. (2021). Chap. 1 Representation. In The alignment problem: Machine learning and human values. Norton, W. W. & Company, Inc.
5. Ratoff, W. (2021). Can the predictive processing model of the mind ameliorate the value-alignment problem? Ethics and Information Technology, 23, 739–750. https://doi.org/10.1007/s10676-021-09611-0
6. Russell, S. (2019). Chap. 7 AI: A different approach & Chapter 8 Provably beneficial AI & Chapter 10 Problem solved? In Human compatible: Artificial intelligence and the problem of control. Penguin.
7. Cruz, J. M. (2019). Shared moral foundations of embodied artificial intelligence. https://sites.williams.edu/jcruz/files/2019/04/AIEthics.pdf
8. Aligned with Whom? Direct and social goals for AI systems" by Anton Korinek, Avital Balwit (2022)
9. Kasirzadeh, A., & Gabriel, I. (2023). In conversation with artificial intelligence: Aligning language models with human values. Philosophy & Technology, 36(1), 27.https://doi.org/10.1007/s13347-023-00606-x
1. Russell, S., & Norvig, P. (2021). Chap. 1 Introduction. In Artificial intelligence: A modern approach (4th ed.). University of California, Berkeley.
2. Christian, B. R. (2021). Introduction. In The alignment problem: Machine learning and human values. Norton, W. W. & Company, Inc.
3. Gabriel, I. (2020). Artificial intelligence, values, and alignment. Minds & Machines, 30, 411–437. https://doi.org/10.1007/s11023-020-09539-2
4. Christian, B. R. (2021). Chap. 1 Representation. In The alignment problem: Machine learning and human values. Norton, W. W. & Company, Inc.
5. Ratoff, W. (2021). Can the predictive processing model of the mind ameliorate the value-alignment problem? Ethics and Information Technology, 23, 739–750. https://doi.org/10.1007/s10676-021- 09611-0
6. Russell, S. (2019). Chap. 7 AI: A different approach & Chapter 8 Provably beneficial AI & Chapter 10 Problem solved? In Human compatible: Artificial intelligence and the problem of control. Penguin.
7. Cruz, J. M. (2019). Shared moral foundations of embodied artificial intelligence. https://sites.williams.edu/jcruz/files/2019/04/AIEthics.pdf
8. Aligned with Whom? Direct and social goals for AI systems" by Anton Korinek, Avital Balwit (2022)
9. Kasirzadeh, A., & Gabriel, I. (2023). In conversation with artificial intelligence: Aligning language models with human values. Philosophy & Technology, 36(1), 27. https://doi.org/10.1007/ s13347-023-00606-x
| 評分項目 Grading Method | 配分比例 Grading percentage | 說明 Description |
|---|---|---|
| 閱讀筆記與提問單閱讀筆記與提問單 Reading notes and question sheets |
60 | 在每次上課前必須繳交閱讀文章重點與問題500字 |
| 個人/分組期末口頭報告個人/分組期末口頭報告 Individual/group final oral presentation |
10 | 同儕互評與老師評分加總 |
| 參與和討論參與和討論 Participate and discuss |
10 | 含出席及參與課程討論 |
