Edouard Harris - New Research: Advanced AI may tend to seek power *by default*
What does power seeking really mean? And does all this imply for the safety of future, general-purpose reasoning systems? Edouard Harris, an AI alignment researcher and one of Jeremie's co-founders of the AI safety company (Gladstone AI) comes back on the TDS Podcast to discuss AI's potential ability to seek power.
Intro music:
➞ Artist: Ron Gelinas
➞ Track Title: Daybreak Chill Blend (original mix)
➞ Link to Track: https://youtu.be/d8Y2sKIgFWc
0:00 Intro
4:00 Alex Turner’s research
7:45 What technology wants
11:30 Universal goals
17:30 Connecting observations
24:00 Micro power seeking behaviour
28:15 Ed’s research
38:00 The human as the environment
42:30 What leads to power seeking
48:00 Competition as a default outcome
52:45 General concern
57:30 Wrap-up
Видео Edouard Harris - New Research: Advanced AI may tend to seek power *by default* канала Towards Data Science
Intro music:
➞ Artist: Ron Gelinas
➞ Track Title: Daybreak Chill Blend (original mix)
➞ Link to Track: https://youtu.be/d8Y2sKIgFWc
0:00 Intro
4:00 Alex Turner’s research
7:45 What technology wants
11:30 Universal goals
17:30 Connecting observations
24:00 Micro power seeking behaviour
28:15 Ed’s research
38:00 The human as the environment
42:30 What leads to power seeking
48:00 Competition as a default outcome
52:45 General concern
57:30 Wrap-up
Видео Edouard Harris - New Research: Advanced AI may tend to seek power *by default* канала Towards Data Science
Показать
Комментарии отсутствуют
Информация о видео
Другие видео канала
![Nicolas Miailhe - AI risk is a global problem](https://i.ytimg.com/vi/cKclc-KThIE/default.jpg)
![Yan Li - The surprising challenges of global AI philanthropy](https://i.ytimg.com/vi/YiEhL28UCj8/default.jpg)
![Kevin Hu - Data observability and why it matters](https://i.ytimg.com/vi/j6FLnS5pgG4/default.jpg)
![Tan Vachiramon - Choosing the right algorithm for your real-world problem (TDS Podcast - Clip)](https://i.ytimg.com/vi/1Kjnf8miIFs/default.jpg)
![Danijar Hafner - Gaming our way to AGI](https://i.ytimg.com/vi/Bgz9eMcE5Do/default.jpg)
![Beena Ammanath - Defining trustworthy AI](https://i.ytimg.com/vi/4-0CFNaFNEc/default.jpg)
![Scaling Machine Learning | Razvan Peteanu](https://i.ytimg.com/vi/aPP4ynaJT9U/default.jpg)
![Panel: Creative Ways to Collect & Use Data for AI | H. Ngo, S. Sun, H. Kontozopoulos, and R. Tabrizi](https://i.ytimg.com/vi/GLcoTaSqHsQ/default.jpg)
![George Hayward: comedian, lawyer and data scientist (TDS Podcast - Clip)](https://i.ytimg.com/vi/x2EaniZseu8/default.jpg)
![Explaining With Impact | Sheldon Fernandez and Michael St. Jules](https://i.ytimg.com/vi/yKMq_1m486I/default.jpg)
![Ben Garfinkel - Superhuman AI and the future of democracy and government](https://i.ytimg.com/vi/QHEEQqh7wd4/default.jpg)
![Last Week In AI — 2021: The (full) year in review](https://i.ytimg.com/vi/9Wl5_t2qNV4/default.jpg)
![Eliano Marques - The (evolving) world of AI privacy and data security](https://i.ytimg.com/vi/9yRl4BDTxSg/default.jpg)
![Joaquin Quinioñero-Candela - Responsible AI at Facebook](https://i.ytimg.com/vi/UKd8hpOl1tQ/default.jpg)
![Irina Rish - Out-of-distribution generalization](https://i.ytimg.com/vi/QjXFN4UWZCg/default.jpg)
![Real Talk with the Director of Data Science at Columbia University](https://i.ytimg.com/vi/RYa2Kcub4NU/default.jpg)
![New to Data Visualization? Start with New York City | Thomas Hikaru Clark](https://i.ytimg.com/vi/Xyn8-2YEFOQ/default.jpg)
![Max Jaderberg - Open-ended learning at DeepMind](https://i.ytimg.com/vi/1yVoNAMLIy0/default.jpg)
![Katya Sedova - AI-powered disinformation, present and future](https://i.ytimg.com/vi/AAe_ZfW0pP8/default.jpg)
![Stuart Armstrong - AI: Humanity's Endgame?](https://i.ytimg.com/vi/Zr97Cxso4W4/default.jpg)
![Fail to Scale | Ian Scott](https://i.ytimg.com/vi/Cod4v_enYb4/default.jpg)