scripod.com

37. 【中英访谈】对话赫拉利 & 王小川:AI 无法受苦,但人类会受苦

Shownote

《商业漫谈 Jane's talk 》终于迎来第一期国际嘉宾~ 天真乐观的科学家和创业者常低估人性的缺陷,而高谈阔论的历史和哲学家,在描绘悲观问题时又很难提出解决方案。 —— 这是每一代技术浪潮来袭时,人类世界常常出现的争议形式。 今年 4 月,我在新浪新闻出品的节目《赛博对话》中,主持了一场来自顶尖头脑的跨界对话,嘉宾正是这两个领域的杰出代表人物: * 全球知名的畅销书作家,对 AI 等技术伦理问题有着深入思考的尤瓦尔・赫拉利。 * 以及在 2023 年就积极投身于大模型浪潮,当下正致力于 AI 医疗实践的...

Highlights

This podcast episode features a profound discussion between historian and philosopher Yuval Harari and entrepreneur Wang Xiaochuan on the implications of AI's rapid evolution. Hosted by commercial talk show Jane's Talk, the dialogue delves into the transformative nature of AI as an autonomous agent rather than a mere tool, challenging traditional perceptions and raising questions about humanity’s future in this evolving landscape.
02:15
Humans created an agent for the first time, not just a tool.
03:01
AI can make decisions by itself.
04:09
AI-written poems are amazing with new and creative metaphors.
06:25
AlphaGo discovered new strategies beyond human imagination.
08:09
AlphaGo went beyond human imagination in playing Go.
10:14
AI is compared to a new 'alien intelligence' that we need to coexist with.
11:35
Many people worldwide show concern about technology's impact.
12:24
Most reply they know there are risks and want to slow down but can't trust their human competitors.
12:53
If we slow down in the AI race while competitors don't, they'll win and rule the world.
13:22
It's insane to trust super intelligent AIs as we have no experience with them.
15:12
The root cause of many issues is human self-preservation and interest disputes.
17:04
Truth requires special mechanisms; fiction thrives without them.
19:08
Entrepreneurs focus on positives, overlook misuse of inventions
20:36
Silicon Valley entrepreneurs wrongly cite the print revolution as a model.
23:16
Meditation taught Harari he has no control over his mind
33:26
True death occurs when forgotten by the last person who remembers you.
34:18
AlphaGo re-enacted human evolutionary history in Go.
36:41
A new era transitions from modeling the physical world to understanding language as math.
37:57
Consciousness means the capacity to suffer.
38:46
Companies and nations can't suffer as they lack consciousness.
40:00
AIs can pretend to feel love through language, but it may just be code.
42:52
Humans fear losing control but should allow it
43:29
Losing control to AI is a huge gamble and could be harmful.
44:16
Most entrepreneurs and politicians distrust humans but trust AI
46:19
AI may not exploit or destroy humans but won't care much about them
46:57
AI revolution involves creating agents that mimic human behavior
49:05
AI development parallels the Industrial Revolution with ethical stakes
52:32
To avoid techniques going out of control, decentralize power and implement self-correcting mechanisms.
54:17
Technology requires a self-correcting mechanism akin to human balance while walking.
54:39
AI must be aware of its potential mistakes and avoid arrogance.
56:05
Consciousness and the ability to feel are what truly matter in life.
1:01:07
Everyone hopes to have a doctor friend or a personal doctor.
1:01:10
AI doctors can record full-course and full-life-cycle data
1:02:06
AI in healthcare mainly summarizes data, helps design experiments.
1:04:44
The medical model's error rate is below 1%, combining human experience and treatment principles.
1:07:20
Creating doctors through AI is now feasible and meaningful for life science decoding.
1:08:34
AI in medical evaluations can reduce social costs under doctors' guidance.
1:09:26
AI systems are set to expand significantly in pediatric healthcare settings.
1:11:35
AI's reasoning engine predicted three diseases for a brain infarction case, aligning with expert findings at a top hospital.
1:11:51
Doing what the world needs makes one happy.

Chapters

AI Development and Its Impact on Humanity
00:00
大模型浪潮回顾: 这次变革在人类历史中的意义
一个词形容赫拉利眼中的 AI 变革: Amazement
02:15
这一次,人类首次发明了一个智能体(Agent),而非工具
02:56
我写《未来简史》(2016)的时候,不曾设想 AI 能有如此大的变革
03:54
A 掌握语言,就掌握了进入社会的「万能钥匙」
04:44
赫拉利教授,您写作时用 AI 吗?
AlphaGo 已经超越了人们的想象,想象是有机(organic)的,但 AI 不是
06:25
赫拉利:「也许人类被困在一个科学孤岛上」
07:54
如果 AI 不久后就能写出《智人之上》,您害怕吗?——「我更愿意称 AI 为外星人智能(Allien Intelligence)」
10:14
AI,巨头,与大模型的科技战
「扎克伯格」们,被 AI 竞赛裹挟的科技巨头
11:27
当我环游世界,我通常问人们两个问题
12:14
「如果我们输掉竞赛,世界将被更无情的人统治」
12:47
人类无法相信同类,却相信 AI,这太疯狂了(almost insane)!
13:22
王小川的技术乐观主义: AI 目前还没有「生命权」
14:52
扎克伯格和张一鸣等技术中立派
16:57
企业家们使用 AI 时,最重要的问题是什么?「诺贝尔并未想到,自己发明的炸药居然被用于战争」
19:05
印刷革命与猎巫行动
20:36
赫拉利 & 王小川,过去如何塑造了你们对未来的看法?
乐观主义与悲观主义如何形成?
23:15
王小川: 从物理学到生命科学,从熵增到熵减
27:13
2016 年,我曾预测了 AlphaGo 能赢
34:14
2018 年,我又预言: 当机器掌握语言时,强人工智能就会到来
35:24
定义生命,决定了我们如何企及生命的未来
37:54
AI 不能无法受苦,但人类会受苦
38:45
智能不是意识,AI 没有感受
39:53
是否要允许「失控」?
凯文凯利说过,要允许 AI 带来的失控,教授是否认同?
42:52
这将是一场豪赌,我呼吁人类,不要放弃控制权
43:23
关于 AI 的发言,人们在公开场合和私底下,说得并不太一样
44:16
我们该信任 AI 吗?——AI 可能并不在乎人类
46:16
为什么 AI 变革不能与工业革命类比?
王小川: 社会分工会收缩,个体可能变得强大
46:49
赫拉利: 工业革命以来人类的表现,我只能打分为 C-;至于这次 AI 革命,我也很难乐观
49:05
「AI 医生应该会比人类医生更善良」
50:55
避免权力的集中化,给商业世界、和普通人们一些建议?
生物系统给我们的启示是,自纠正机制十分重要
54:04
不要太过自信,请保持怀疑
54:39
人类不可替代的价值: 意识与情感
56:05
赫拉利教授不用智能手机?「我提倡『信息节食(information diet)』」
58:32
王小川详解 AI 如何改变医疗行业
为什么让 AI 先从儿科医生干起?
1:01:10
让医学追上物理学,先从数据收纳、观测开始
1:02:06
百川的 AI 医疗大模型,能做什么?
1:03:00
百度魏则西事件(2016),对王小川的影响
1:06:04
如何确定 AI 医疗的边界?「这并不是个多难的话题」
1:08:30
未来 18 个月内,AI 医疗将得到很大的普及
1:09:26
百川创业两年多,有哪些 suffer 的时刻?
1:10:29
这次创业更兴奋,「你在做一件世界需要你的事」
1:11:51

Transcript

卫诗婕: AI 正以前所未有的速度发展,人类真的准备好了吗?他是著名的历史学家,哲学家,人类简史系列作者,尤瓦尔・赫拉利。他的著作《人类简史》,《未来简史》和《今日简史》。以及最新出版的《智人之上》 被翻译成超过 65 种语言。也因此他被称为,当今世界最具影响力的,公共知识分子之一。而他是坚定的技术乐观派,做看潮起潮落的连续创业者,更是中国大模型的一线实践者,王小川。他于 2023 年 4 月,创立的大模型公司百川智能,两年来先后发布 12 款大模型。 王小川: 今天机器掌握语言之后,其同时掌握了思考。 卫...
小宇宙
Open in 小宇宙