308. AI Safety and Explainability 人工智能安全,访谈图灵奖得主Prof. SifakisBlow Your Mind (BYM S02)

308. AI Safety and Explainability 人工智能安全,访谈图灵奖得主Prof. Sifakis

53分钟 ·
播放数6226
·
评论数60

本期嘉宾是 Joseph Sifakis 教授。Sifakis 教授是知名的计算机科学家,在2012年获得达芬奇奖,并在2007年获得图灵奖。他的书作 Understanding and Changing the World 中译版《理解和改变世界》已经出版。


在这期播客中,我们讨论了多个话题,涉及计算机科学、哲学与AI及 autonomous system (自治系统) 对社会的影响。主要观点包括:

-写书的缘起,来自 Prof. Sifakis 对哲学问题的兴趣和他觉得现有哲学文献的不足,特别是关于知识和信息概念有所不足。

-AI与可解释性:知识的本质。强调AI生成的知识缺乏可解释性,而通过科学方法产生的知识具备可解释性。AI与科学方法相比,无法提供强有力的保证。

-自治系统(autonomous systems)的测试与安全:AI系统的安全性是个大挑战,如自动驾驶,现有的系统缺乏可解释模型,而实践测试又有局限性。

-AI的伦理与社会影响:将AI投入实际应用的伦理考量,特别是对人类责任和决策的影响。

-AI在未来技术中的角色:AI在未来可应用于系统监测,以及与传统控制系统集成以提高可靠性和安全性。混合系统 (hybrid system)的优势和挑战。

-Prof. Sifakis 的职业生涯和研究兴趣:职业轨迹覆盖模型检查的理论研究到系统设计的实际应用,以及他是如何选择研究课题。

-AI在教育中的应用:AI工具(如聊天机器人)对儿童发展和学习的影响,倡导在教育中节制地使用技术。

-对大众的建议:负责任地使用AI,辩证评估AI产出的内容,维护人类的价值观和责任。


Blow Your Mind (BYM) 是一档中英双语播客,涉猎广泛的主题,包括科技、经济、讲故事、历史、创业、创造力和数学。过往的嘉宾包括电影制作人、作家、表演艺术家和AI科学家。

======

Our guest is esteemed computer scientist Professor Joseph Sifakis. He is winner of the Leonardo da Vinci Medal (2012), and the Turing Award (2007). His book "Understanding and Changing the World" is available in Chinese with the title 理解和改变世界.

In the podcast, we talked about a number of topics, focusing primarily on the intersection of computer science, philosophy, and the impact of AI and autonomous systems on society. Key points include:

-Genesis of Sifakis' Book: Motivated by his interest in philosophical issues and the shortcomings he found in philosophical texts, particularly regarding concepts like knowledge and information.

-AI and Explainability: The nature of knowledge. Emphasizes the difference between AI-generated knowledge, which lacks explainability, and scientific knowledge, which provides it. Highlights the limitations of AI in providing guarantees compared to scientific methods.

-Testing and Safety of Autonomous Systems: Explores the challenges in ensuring the safety of AI systems, especially self-driving cars, due to the lack of explainable models and the limitations of empirical testing.

-Ethical and Societal Implications of AI: Addresses the ethical considerations of AI deployment in society, particularly the impact on human responsibility and decision-making.

-Role of AI in Future Technology: Discusses the future role of AI, especially in system monitoring and integrating AI with traditional control systems for better reliability and safety. The advantages and challenges of hybrid systems.

-Sifakis' Career and Research Interests: Reflects on his career trajectory, from theoretical research in model checking to practical applications in system design, and his personal approach to choosing research topics.

-AI in Education: Raises concerns about the impact of AI tools like chatbot on children's development and learning, advocating for a balanced approach to technology use in education.

-Advice for General Public on AI: Advice on using AI responsibly, emphasizing the importance of understanding and critically assessing AI outputs, and maintaining human values and responsibilities.


Blow Your Mind (BYM) is a Chinese and English bilingual podcast that covers a broad range of topics. Some of the recurring themes are: technology, economics, storytelling, history, entrepreneurship, creativity and mathematics. Past show guests include filmmakers, authors, performing artists and AI researchers. 



展开Show Notes
bymbrofeng
bymbrofeng
2023.12.07
置顶
感谢中信出版社给咱们 BYM 听众提供了福利:在小宇宙 APP 本节目下留言,我们会选出三位听众,赠送 Prof. Sifakis 的书作《理解和改变世界》。
ywzzz
ywzzz
2023.12.07
好喜欢brodeng这样不管听者的英文水平就来全英文播客的刺激啊
bymbrofeng
:
啊啊啊啊啊啊啊
HD92881919:哈哈
4条回复
Kim_zvaM
Kim_zvaM
2023.12.07
差点以为教授的声音是AI生产的哈哈哈
Lililili:可能是信号太差了🥹
41:10 brofeng: "The way you choose your research topics is guided by taste.Could you elaborate on that?

What do you mean by taste and how can a young researcher develop her own Taste?"

Joseph Sifakis: "What I tried in my life is I got interested in many different topics. I do not want to stay or on one topic. So I started to work hardware and then I did not like so much reception hardware. I moved to software.

And then I did many different things. I worked on verification. Then I got interested in a real time systems, embedded systems. Now I'm interested in autonomous systems. So I like to move on but I'm not sure. Very curious.

Here are some visions. Perhaps sometimes I'm wrong, but I don't like to follow other people. I want to cultivate my own vision. Of course, if that fits with other people, that's fine. But I have strong opinion as you can still talk about AI. Okay, I have strong opinion. It's very different from others.

So I'm going to have a global vision about what I'm doing and its relationship the way that what happens in the world. And this has motivated many of my research."
小端午dd:It’s almost a luxury nowadays not having to focus on one research topic only.
bymbrofeng
:
祝贺!你获得Sifakis 教授书作的中译版(中信出版)。请在本周五(12月15号)结束前发送收件地址。可以发送微博私信给 @bymbrofeng, 或email bymclub@outlook.com 注意收件地址限中国大陆地区。书将由中信出版寄出。
3条回复
土豆_Dtn9
土豆_Dtn9
2023.12.07
我是真想听啊,这英语我也听不懂啊,就听清了个 and
阿紫Zoe:我就听懂了个片头曲
范思其
范思其
2023.12.07
人工智能的发展一定程度上让人类的主体意识变得薄弱(hh或许真的是硅基生物的引路人呢)
在设计行业对人工智能的依赖 让不少设计师沉迷于让自己的工具变得越来越强大 这也是我们作为“人”这个主体发展所需要考虑到的潜在威胁。即便如此 我仍然对人工智能持有理性偏乐观的态度 人工智能本质上是赋能人类的便捷工具 为实现人类的需要 只不过是对人的大脑、智慧的一种模拟 人类的想象力 创造力 社会责任感 人文情怀才是最真切最真实的。
bymbrofeng
:
祝贺!你获得Sifakis 教授书作的中译版(中信出版)。请在本周五(12月15号)结束前发送收件地址。可以发送微博私信给 @bymbrofeng, 或email bymclub@outlook.com 注意收件地址限中国大陆地区。书将由中信出版寄出。
石小鱼
石小鱼
2023.12.13
听锋哥开头介绍内心os:哇,好棒棒,能听懂诶
听教授讲话:采样率全损限制了我们向教授学习😅
bymbrofeng
:
我听着还行啊!
石小鱼:收到,继续练习英语能力😄 BTW,是「峰」哥
HD510235m
HD510235m
2023.12.10
晚上听很合适 助眠效果不错(doge)
bymbrofeng
:
嗷嗷嗷!
Shelly_mVYg
Shelly_mVYg
2023.12.20
教授把AI生成的知识也叫做知识,和人的“empirical knowledge”区分开,这种分类学到了。这本书的精髓想必就是帮助人了解计算机科学家看世界的方式,很有意思,准备买了!
小端午dd
小端午dd
2023.12.10
41:42 I like that he used her! Her taste.
笑死 这期总算体现了我之前东一榔头西一棒子听的各期人工智能播客节目+半吊子英语专业带来的一丝丝价值所在 以及:gpt3.5直接语音输入识别还蛮好用的
Bro的英语,口齿清晰,用词精准,听着不费劲还特舒服,能感觉为了照顾听众已经特意在enunciate each word
HD80272x
HD80272x
2023.12.12
Prof. 是意大利口音吗
bymbrofeng
:
希腊 + 法国
Absurdity
Absurdity
2023.12.09
听到了1000个OK
bymbrofeng
:
OK!
Absurdity:🫢😂😬
NinaZzzz
NinaZzzz
2023.12.08
教授的口音对听力能力有待改进的我来说很有挑战,适合安静的空间边听边记录!下班路上收藏了,今晚哄睡后慢慢听
00:02 在听啦!
桃奈小安
桃奈小安
2023.12.07
打卡
05:34 来了
芒小菓
芒小菓
2023.12.07
2号观景位get✔
教授说话时候的电流音 让他听起来很有AI感~