月度归档： 2024 年 3 月

永生与数字死亡的局限性
人们常常认为，如果我们治愈衰老或找到如何上传人类思想，人类就会永生。 今天我们将研究这个概念，看看它与天文时间线的关系如何。

请访问我们的网站：http://www.isaacarthur.net。
加入星云：https://go.nebula.tv/isaacarthur。
在 Patreon 上支持我们：/isaacarthur。
在 Subscribestar 上支持我们：https://www.subscribestar.com/isaac-a…
脸书群组： / 1583992725237264
Reddit：/艾萨阿瑟。
Twitter：/isaac_a_arthur 在 Twitter 上并转发我们未来的内容。
SFIA Discord 服务器：/discord 或从 Soundcloud 下载本集的音频：/digital-death。
Jakub Grygier 的封面艺术：https://www.artstation.com/artist/jak…

图形团队：
爱德华·纳德拉.
贾里德·伊格利.
贾斯汀·迪克森。
凯蒂·伯恩。
Mafic Stufios 的 Kris Holland：www.maficstudios.com。
米绍·约尔达诺夫。
皮埃尔·德梅特.
塞尔吉奥·博特罗：https://www.artstation.com/sboterod?f…
斯特凡·布兰丁。

脚本编辑：
安迪·波佩斯库。
康纳·霍根。
爱德华·纳德拉.
尤斯特拉修斯·格雷厄姆.
格雷戈里·莱尔.
杰斐逊·伊格利.
卢卡·德·罗莎.
马克·沃伯顿。
迈克尔·古谢夫斯基。
米奇·阿姆斯特朗。
莫尔博组织。
娜奥米·科恩.
菲利普·鲍尔多克.
西格蒙德·科佩鲁德.
史蒂夫·卡登。
蒂芙尼·彭纳。

音乐：
马库斯·朱尼卡拉，\

来源和详细信息：
https://www.youtube.com/watch?v=6icg0tu_5z4&si=9aWP6IlzbxD42XDZ

Geoffrey Hinton | Will digital intelligence replace biological intelligence?
Geoffrey Hinton will be presented at the University of Toronto on October 27, 2023 by the Schwartz Reisman Institute for Technology and Society and Department of Computer Science in collaboration with Vector Institute for Artificial intelligence and Cosmic Future at the Faculty of Arts & Science.

Opening remarks and introduction.
0:07:21 — 0:08:43 Overview.
0:08:44 — 0:20:08 Two different ways to do computation.
0:20.09 — 02:30.11 Can large language models understand what they say?
The first neural network language model.
Will we be in a position to control the super-intelligence when it surpasses human intelligence?
0:57:25 — 1:03:18 Does digital intelligence have subjective experience?
1:03:19 — 1:55:36 Q&A
1:55.37 — 1:58.37 Closing remarks

Talk title: \”Will digital intelligence replace biological intelligence?\”

Abstract: Digital computers are designed to let a person tell them what to do. The models can be used on different hardware pieces, even though they require high energy. We could use analog computations that are very low-power and take advantage of the unique properties of hardware to mimic biology in order to create computers that can learn. It is necessary to have a learning algorithm which can use analog properties even without a model. The computation becomes mortal when it uses the analog properties that are idiosynchratic to the hardware. The learned knowledge is lost when the hardware dies. It is possible to transfer the knowledge to a newer analog computer if the latter can mimic the older computer’s outputs. However, this process of education takes a long time. Digital computation allows for many copies of the exact same model to be run on different hardware. By averaging the weight changes of thousands of digital agents, they can share their learnings very efficiently. Chatbots such as GPT-4 or Gemini are able to learn thousands of time more than a single person. Digital computations can also use backpropagation, which is a much more scalable learning method than analog hardware. I believe this leads me to think that digital computation on a large scale is likely better at acquiring information than biological computation, and could soon be more intelligent than we are. Digital intelligences should be less vulnerable to war and religion because they are immortal and have not evolved. However, if one of these digital super-intelligences ever wants to take over, it’s unlikely we can stop them. Therefore, the most pressing question for AI research is to find a way to prevent this from happening.

Geoffrey Hinton

Geoffrey Hinton earned his PhD in artificial Intelligence from Edinburgh University in 1978. He moved from Carnegie Mellon to the Department of Computer Science of the University of Toronto where he now is an Emeritus Professor. Google acquired Hinton’s neural networks startup DNN research in 2013, which was developed from his research at U of T. Hinton then served as a Vice-President and Engineering Fellow for Google until 2023. He is the founder of Vector Institute for Artificial intelligence where he serves as Chief Scientific Advisor.

来源和详细信息：
https://www.youtube.com/watch?v=iHCeAotHZa4&si=8yjSkHNwWEfsd57F