【高清完整版｜双语字幕】牛津大学英辩 AI 是否构成人类生存

共 8 集

剧集目录

P1 · 【高清完整版｜双语字幕】牛津大学英辩 AI 是否构成人类生存威胁 p01 1 (P1)

【高清完整版｜双语字幕】牛津大学英辩 AI 是否构成人类生存威胁 p01 1 (P1)

发布于 2026-07-22 · 时长 10:20

字幕摘录

时间	英文	中文
0:02	I move that this House believes that artificial intelligence is an existential threat.	我提出,本院认为人工智能是一种生存威胁.
0:07	To open the case for the proposition, I call up Sultan Kokar, Deputy Director of Press at the Union.	为了提出这个建议,我打电话给苏尔坦·科卡尔,联盟新闻副主任。
0:13	Madam President, Honourable Members, I am honoured to open this seminal debate before you tonight.	主席女士,尊敬的各位成员,我荣幸地在今晚开始你们面前的这次具有开创意义的辩论。
0:19	The question of artificial intelligence and the role it plays in our futures has gripped the imagination and fears of our times.	人工智能问题及其在我们未来中的作用,使我们时代的想象力和恐惧感笼罩。
0:27	With the likes of advanced chatbots like ChatGPT, AI has finally entered a very public mainstream in a way that it had not done thus far.	由于像ChatGPT这样的高级聊天员,AI最终以至今为止还没有做到的方式进入了一个非常公开的主流.
0:36	However, make no mistake, this is not a debate about ChatGPT or its equivalents.	然而,不要犯错误,这不是关于ChatGPT或与其相当的辩论。
0:42	This is not a debate about AI writing better essays than us or producing more complex art.	这不是关于AI写出比我们更好的论文或产生更复杂的艺术的辩论.
0:48	Nor do we on the proposition dispute the unending benefits that the application of advanced AI can have in the fields of medicine, tackling poverty, democratising access to resources, etc.	我们也没有对应用先进的AI在医学、解决贫困、实现资源获取的民主化等领域所能带来的无休止的好处提出质疑。
1:02	No, this is a debate about the acute existential risk posed by artificial intelligence systems with capabilities that we in this chamber can hardly imagine.	不,这是一场辩论 关于人工智能系统 带来的严重生存风险 人工智能系统 拥有我们无法想象的能力
1:12	Systems that we are hurtling towards at breakneck speed with little to no conception of danger we are nurturing.	我们所培育的系统正在以突破性的速度向着微弱或毫无危险的概念前进。

展开字幕全文（90 条）

序号	英文	中文
1	I move that this House believes that artificial intelligence is an existential threat.	我提出,本院认为人工智能是一种生存威胁.
2	To open the case for the proposition, I call up Sultan Kokar, Deputy Director of Press at the Union.	为了提出这个建议,我打电话给苏尔坦·科卡尔,联盟新闻副主任。
3	Madam President, Honourable Members, I am honoured to open this seminal debate before you tonight.	主席女士,尊敬的各位成员,我荣幸地在今晚开始你们面前的这次具有开创意义的辩论。
4	The question of artificial intelligence and the role it plays in our futures has gripped the imagination and fears of our times.	人工智能问题及其在我们未来中的作用,使我们时代的想象力和恐惧感笼罩。
5	With the likes of advanced chatbots like ChatGPT, AI has finally entered a very public mainstream in a way that it had not done thus far.	由于像ChatGPT这样的高级聊天员,AI最终以至今为止还没有做到的方式进入了一个非常公开的主流.
6	However, make no mistake, this is not a debate about ChatGPT or its equivalents.	然而,不要犯错误,这不是关于ChatGPT或与其相当的辩论。
7	This is not a debate about AI writing better essays than us or producing more complex art.	这不是关于AI写出比我们更好的论文或产生更复杂的艺术的辩论.
8	Nor do we on the proposition dispute the unending benefits that the application of advanced AI can have in the fields of medicine, tackling poverty, democratising access to resources, etc.	我们也没有对应用先进的AI在医学、解决贫困、实现资源获取的民主化等领域所能带来的无休止的好处提出质疑。
9	No, this is a debate about the acute existential risk posed by artificial intelligence systems with capabilities that we in this chamber can hardly imagine.	不,这是一场辩论 关于人工智能系统 带来的严重生存风险 人工智能系统 拥有我们无法想象的能力
10	Systems that we are hurtling towards at breakneck speed with little to no conception of danger we are nurturing.	我们所培育的系统正在以突破性的速度向着微弱或毫无危险的概念前进。
11	But before I delve into the imminent downfall of humankind, it falls upon me to introduce your speakers for the opposition.	但在我深入探讨人类即将衰落之前,我有责任向反对派介绍你们的演讲者。
12	Speaking first, we will have Sebastian Watkins, the Union's librarian.	首先,我们会有塞巴斯蒂安・沃特金斯, 联盟的图书管理员。
13	Seb is a really interesting person.	塞布是个很有趣的人
14	He chairs Library Committee and his favourite hobby is chess.	他主持图书馆委员会 他最喜欢的爱好是围棋
15	They say that men are always thinking about the Roman Empire, but Seb takes that a step further with his phone wallpaper.	他们说,男人总是在想罗马帝国,但塞布用他的电话壁纸更进一步.
16	A shining red flag adorned with a golden eagle and the motto, Senatus populusque Romanus.	一面闪亮的红旗上装饰着金鹰和座右铭,塞纳图斯·波普卢斯克·罗曼努斯.
17	If only the victory banner of the Empire had helped him in his election for librarian.	如果只有帝国的胜利旗帜帮助他当选图书管理员.
18	你的第二位反对派发言人是Yeshi Milner,他是Data for Black Lives的执行董事兼共同创始人.
19	She aims to leverage data science and its possibilities to create meaningful change in the lives of black people long involved in data science and social activism.	她的目标是利用数据科学及其可能性,为长期参与数据科学和社会活动活动的黑人的生活带来有意义的变化.
20	She has worked tirelessly to advocate against big data and big tech and expose the inequalities that pervade our current data systems.	她孜孜不倦地宣传反对大数据和大技术,揭露我们目前数据系统中普遍存在的不平等现象。
21	Her work has resulted in policy changes and she was recognized by Forbes's 30 under 30 in 2020.	她的作品导致了政策上的改变,她在2020年被福布斯30岁以下30人所认可.
22	We are honoured to host her here tonight.	我们很荣幸今晚在这里接待她
23	I would caution you, however, that between her and Seb, there are two Americans on the opposition.	但我要提醒你 她和塞布之间 有两名美国人在反对派
24	So be careful how you vote tonight.	所以当心你今晚的投票方式
25	Your next speaker will be Anna Roska, who is a member of the Secretary's Committee here at the Oxford Union.	下一位发言者是安娜·罗斯卡,他是牛津联合会这里的秘书委员会成员。
26	I'm sure that all our colleagues will agree that she's an incredibly hardworking and committed member of committee.	我相信我们所有的同事都会同意 她是一个令人难以置信的辛勤和执着的委员会成员。
27	However, she studies PPE.	然而,她学习个人防护设备。
28	Nevertheless, I'm excited to hear her contribution to this debate.	尽管如此,我还是很高兴听到她对这场辩论的贡献.
29	And your final speaker on the opposition will be Professor Eric Shing.	最后一位是埃里克·成教授
30	After watching some of his interviews, I came to learn that he does not like listening to all of his credentials.	在观看了他的一些访谈后,我得知他不喜欢听他所有的证书。
31	So bear with me while I engage in a little psychological warfare.	在我进行心理战时 忍耐一下
32	成教授是穆罕默德·本·赛德人工智能大学(Mohammed bin Syed University of Artifical Intelligence)的校长,这是世界上首所致力于AI的大学.
33	He is an accomplished and esteemed researcher, having held positions at Carnegie Mellon, Stanford, Pittsburgh and Facebook, and is also the founder of Petrium Inc.	他是一位成就卓著和受人尊敬的研究员,曾在卡内基梅隆,斯坦福,匹兹堡和Facebook任职,也是Petrium Inc的创始人.
34	He's authored or contributed to more than 400 research papers and has been cited more than 44,000 times.	他的著作或贡献了400多篇研究论文,并被引用超过44,000次.
35	Again, we are honored to have him with us tonight.	我们再次感到荣幸 他今晚和我们在一起。
36	Now, I stated earlier that this is not a debate about simple chatbots like chat GPT, but rather about more advanced, even hypothetical, artificial general intelligence systems.	现在,我早些时候说过,这不是关于像聊天GPT这样的简单的聊天人的辩论,而是关于更先进的,甚至假设的人工一般智能系统的辩论.
37	What are the characteristics of such technologies?	这类技术有哪些特点?
38	Well, most researchers agree that an AGI would be able to reason, represent knowledge, plan, learn, communicate naturally, and of course, integrate these skills amongst each other towards completing a given goal.	多数研究者都同意, AGI 能够解释, 代表知识, 计划,学习,自然交流, 当然, 相互融合这些技能, 以完成一个特定的目标。
39	Though such technology is to some extent hypothetical at the moment, a 2022 survey did find that only 1.1% of researchers felt it would never exist.	虽然目前这种技术在某种程度上是假设性的,但2022年的一项调查确实发现只有1.1%的研究人员认为它永远不会存在.
40	More than half said it would emerge in the next few decades and the leaders of OpenAI argue in the next 10 to 20 years.	超过半数的人说,它将会在未来几十年出现,OpenAI的领导人在未来10至20年中争论.
41	Now, though such technology would certainly come with many benefits, it would also bring enormous risks.	现在,尽管这种技术肯定会带来很多好处,但也会带来巨大的风险。
42	These center around AI control and alignment.	这些中心围绕AI控制和对齐.
43	Although such a technology would inevitably be programmed by us humans, it would be very difficult to instill it with the full range of human values and ethics.	虽然这种技术不可避免地是由我们人类编程的,但很难将其灌输到人类的各种价值观和伦理。
44	Human values, emotions, and ethics are broad, complex, and as I'm sure you will agree, often extremely illogical.	人类的价值观,情感和伦理是广泛,复杂的,我相信你们会同意的,常常是极不合逻辑的.
45	Short of plugging an AI into our own brains 24-7, it is very difficult to align it with these in their entirety.	24 -7连个人工智能都插进我们自己的大脑里 很难完全适应这些
46	If a super-intelligent AI determines that adopting values like concern for human life would hinder the goals we have programmed it to fulfill, then why wouldn't it resist attempts to program such values into it?	如果一个超级聪明的AI认定,采用对人类生命的关心这样的价值观会阻碍我们计划它实现的目标,那么它为什么不抵制将这种价值观编程入其中的尝试呢?
47	And unless we are successful in fully aligning such a super-intelligence with the entire range of human morality and constraint, then we cannot expect it to just be on our side.	除非我们成功地把这种超级智慧与人类道德和约束的全方位结合起来,否则我们不能指望它只是站在我们一边。
48	In a while, maybe.	过一会儿,也许吧。
49	One leading researcher proposes the following thought experiment.	一位主要研究者提出如下思想实验.
50	Imagine that you task an AI system with the simple job of making as many paper clips as possible.	想象一下,你负责一个AI系统,简单的工作是制作尽可能多的剪纸片段.
51	It will quickly come to understand that its job would be far easier if humans were out of the way, since a human could turn it off at any point, and that would mean fewer paper clips.	它很快会明白,如果人类离开,它的工作会变得容易得多,因为人类随时可能关闭它,这意味着纸片剪辑会减少.
52	With this goal, the AI would work towards a future with many paper clips and no humans.	有了这个目标,AI将努力走向一个拥有许多剪纸片,没有人类的未来.
53	Now this example may seem trivial, but it demonstrates the unavoidable risk that a technology that can think for itself, independent of us, poses.	现在,这个例子可能显得微不足道,但它表明,一个能够独立于我们而为自己思考的技术所构成的不可避免的风险。
54	Let's translate the same example onto something more significant.	让我们把同样的例子 翻译为更重要的东西。
55	Suppose we task an AI technology with reducing inequality in our society, something more realistic.	假设我们用人工智能技术 来减少社会中的不平等 更现实一点
56	The AI could determine, like we often do, that the solution is closing the wealth gap, but it might determine that the solution to doing so is not to reduce the gap between rich and poor, but to make everyone poorer,	AI可以像我们常做的那样,确定解决办法是缩小财富差距,但它可以确定解决办法不是缩小贫富差距,而是使每个人更加贫穷,
57	and in doing so it might choose to lower standards of living, increase poverty, increase crime, because we haven't specified that these things are important.	在这样做的时候,它可能选择降低生活水平, 增加贫困,增加犯罪, 因为我们没有说明这些东西是重要的。
58	It achieves its goal, but at a cost that we did not want nor anticipated.	它实现了它的目标,但代价是我们不希望或预期的。
59	In other words, we can shape AI to prevent one outcome, but to preempt every possible risk is impossible.	换句话说,我们可以塑造AI来防止一种结果,但预先防范每一种可能的风险是不可能的.
60	In order for an AI to be risk-free altogether, it must be perfectly aligned with zero room for error.	为了使AI完全无风险,它必须与零错误空间完全一致.
61	Since human morality, ethics, and desires are inherently subjective and prone to bias, achieving this universally perfect alignment is not feasible.	由于人类的道德,道德和欲望本质上是主观的,容易产生偏见,因此实现这种普遍完美的结合是不可行的.
62	I don't mean to propose some sort of Ultron-style AI takeover, but if an AI realizes and comes to the very straightforward realization that acquiring greater power is conducive to fulfilling virtually any objective,	我无意提议某种Ultron式的AI收购 但如果一个AI意识到 并且直接意识到 获得更大的权力 有利于实现任何目标
63	it could copy itself onto other systems, instigate manufacturing lines, evade shutdown, and even appear aligned and hide behavior that it recognizes as unwanted by its creators.	它可以复制自己到其他系统, 煽动制造线, 逃避关闭, 甚至看起来对齐和隐藏的行为,
64	Consider that for a moment, an AI that is misaligned and can hide that from us.	暂时考虑一下,一个错误的人工智能 并且可以掩盖我们。
65	To prevent itself from being switched off, it might jump from a computer in San Francisco to one in Singapore, from Singapore over to London.	为了防止自己被关闭,它可能从旧金山的一台电脑跳到新加坡的一台电脑,从新加坡到伦敦。
66	Before we know it, it has multiplied itself onto thousands of systems worldwide, and all the while we aren't even aware of its true intents.	在我们知道它之前,它已经在全世界 成千上万个系统上, 而与此同时,我们甚至不知道它的真正意图。
67	This may sound like the work of science fiction, but we're already on the way to this becoming reality.	这听起来可能像科幻小说, 但我们已经在走向现实。
68	In 2021, one AI model was trained to grab a ball, but learned that it could simply place its hand between the ball and the camera to give the illusion that it had succeeded.	2021年,一个AI模型接受了抓球的训练,但得知可以简单地将手放在球和相机之间,以给出自己成功的幻觉.
69	Not only was the AI in this instance able to outsmart its creator, but this demonstrates the fallibility of human programming and expectations.	在此情况下,大赦国际不仅能够比其创造者聪明,而且表明人类编程和期望的谬误。
70	Even chat GPT, the bane of every tutor's existence, is able to fulfill some of the characteristics that I attributed to AGI earlier.	即使是聊天的GPT,每个导师的存在之祸,都能够实现我先前赋予AGI的一些特征.
71	It can learn from our responses, it represents knowledge, and it can produce natural sounding language.	它可以从我们的反应中学习,它代表知识,并且能够产生自然的发声语言.
72	The technology I have described may seem far off, but we are closer than we think.	我所描述的技术可能看起来很遥远,但我们比我们想象的要接近。
73	I'm sure that my far more knowledgeable colleagues will expand on the technical details of the existential risk posed by AI.	我确信,我更了解的同事们将会扩展AI带来的存在风险的技术细节.
74	The point I would like to leave you with is this.	我想告诉你的是这一点。
75	Human morality, ethics, wishes, are both incredibly complex and utterly confused.	人类的道德,道德,愿望, 既令人难以置信的复杂,又完全混淆。
76	Not only is it effectively impossible to program these into an AI system in any meaningful way,	不仅实际上不可能以任何有意义的方式将这些系统编成AI系统,
77	but we ourselves can hardly decide what human morality even looks like.	但我们自己也很难决定什么是人类道德
78	We don't even know what human morality is.	我们甚至不知道什么是人类的道德。
79	So these are the facts of the debate.	这就是辩论的事实。
80	We do not know for certain how or if we can control the AI we are forced in the process of building, number one.	我们不知道,我们如何或是否能够控制我们在建设过程中被迫使用的人工智能。
81	We do not know to what extent this AI will be aligned with our values and desires, number two.	我们不知道该AI在多大程度上将符合我们的价值观和愿望,第二号。
82	And finally, we do not even know what our values and desires are.	最后,我们甚至不知道我们的价值观和愿望是什么。
83	However likely or unlikely you personally believe the existential threat from AI is to materialize,	无论你个人认为来自AI的存在威胁会实现,
84	it is indisputable that this threat does and will exist.	毫无疑问,这种威胁确实存在,而且将会存在。
85	And that is what we are debating tonight.	这就是我们今晚辩论的内容。
86	All of us sitting here tonight pride ourselves on being intelligent, critically thinking people.	我们今晚都坐在这里 为自己聪明 认真思考的人感到自豪
87	Do not leave the future of humankind, your future, up to chance.	不要离开人类的未来,你的未来,机会。
88	Our special intelligence at its current rate of development poses a distinct existential risk that we are unprepared to deal with.	以目前发展速度看,我们的特殊智慧构成了我们无法应对的明显的生存风险。
89	Vote with the proposition tonight.	今晚投票支持这个提议。
90	Thank you.	谢谢

该视频共有字幕 91 条。解锁更多字幕为会员功能，请移动到 价格

P2 · 【高清完整版｜双语字幕】牛津大学英辩 AI 是否构成人类生存威胁 p02 2 (P2)

【高清完整版｜双语字幕】牛津大学英辩 AI 是否构成人类生存威胁 p02 2 (P2)

发布于 2026-07-22 · 时长 9:34

字幕摘录

时间	英文	中文
0:02	Esteemed members and guests, ladies and gentlemen,	女士们先生们,
0:04	firstly, allow me to thank you so very much	首先,请允许我非常感谢你
0:06	for coming to this debate tonight.	今晚参加这场辩论
0:08	And if I may say so, it really is a pleasure.	如果我可以这么说的话,那真的是一种乐趣。
0:10	And having said that, I promise,	说了这些 我保证
0:12	that's the only line ChatGPT wrote.	这是ChatGPT写的唯一一条线。
0:14	Now, despite what the proposition has tried	现在,尽管这个提议已经尝试了
0:17	to convince you of in those roughly 10 minutes,	说服你在这大约10分钟,
0:20	I'm here to tell you that AI is not only not	我是来告诉你 AI不仅仅是
0:24	an existential threat, it is rather one	这是一个生存威胁, 相当是一个

展开字幕全文（209 条）

序号	英文	中文
1	Esteemed members and guests, ladies and gentlemen,	女士们先生们,
2	firstly, allow me to thank you so very much	首先,请允许我非常感谢你
3	for coming to this debate tonight.	今晚参加这场辩论
4	And if I may say so, it really is a pleasure.	如果我可以这么说的话,那真的是一种乐趣。
5	And having said that, I promise,	说了这些 我保证
6	that's the only line ChatGPT wrote.	这是ChatGPT写的唯一一条线。
7	Now, despite what the proposition has tried	现在,尽管这个提议已经尝试了
8	to convince you of in those roughly 10 minutes,	说服你在这大约10分钟,
9	I'm here to tell you that AI is not only not	我是来告诉你 AI不仅仅是
10	an existential threat, it is rather one	这是一个生存威胁, 相当是一个
11	of the greatest inventions mankind has ever made,	是人类所创造的,
12	one of the greatest.	一个伟大的。
13	And not least because, not least because	也是因为...
14	it can do your essays for you, for free.	它可以为你做你的论文,免费。
15	Now, to make this case, we will try	现在,为了办这个案子,我们将尝试
16	and demonstrate this through three clear points.	并通过三个明确的点来证明这一点。
17	Firstly, that AI offers us as a species	首先,人工智能为我们提供了一种物种
18	such incredible benefits that are just simply	如此令人难以置信的好处,只是
19	too good to pass up on.	太好,不能放弃。
20	Moreover, that there are, as I have come to fear,	此外,正如我所担心的,
21	slight misconceptions about what AI exactly is.	对AI究竟是什么略有误解.
22	It's not, as Sultan so colorfully described,	这不是,正如苏丹如此彩色的描述,
23	a Marvel villain Ultron, I think you said,	一个马维尔的恶棍Ultron, 我想你说,
24	flying through the sky trying to kill	飞过天空试图杀人
25	Thor or Iron Man, not always.	索尔或铁人,并非总如此.
26	And finally, that even if you think	最后,即使你认为
27	it might someday in the future become	将来有一天会变成
28	an existential threat, just because it isn't	一个生存的威胁, 只是因为它不是
29	existentially a threat to us now,	对我们的威胁是存在的
30	we still take this debate.	我们还在辩论
31	Now, before any of that, I have the duty	现在,在这之前,我有责任
32	and, if I may say so indeed, the pleasure	如果我可以这么说的话, 很高兴
33	to introduce the other side.	介绍另一边。
34	So, firstly, you just heard from Sultan Kokar,	所以,首先,你刚刚从苏丹科卡尔,
35	our Deputy Director of Press and a second year linguist.	我们的副新闻主任 和第二年语言学家。
36	Now, Sultan is a very wonderful man,	现在,苏丹是一个非常好的人,
37	or so I would have said.	否则我会说
38	And I was very glad you mentioned my	我很高兴你提到我
39	Roman Victory banner, and I was even more glad	罗马胜利的旗帜,我更高兴
40	that it was of such good help to you	这对你很有帮助
41	in your own election for Secretary's Committee.	在你自己的选举秘书委员会。
42	Oh wait, sorry, it wasn't.	哦,等等,对不起,它不是。
43	But no, moving on, Sultan is genuinely	但是,没有,前进,苏丹是真正的
44	a lovely, lovely person.	一个可爱,可爱的人。
45	You know, clean as a cat, one of the nicest people	你知道,干净的猫, 最好的人之一
46	you'll ever meet.	你将永远相遇。
47	In fact, I have to admit, when I was looking	事实上,我不得不承认, 当我在寻找
48	for dirt on you, I had a really hard time.	对于你的污垢, 我有一个非常困难的时间。
49	But what I shall say, if you'll permit me,	但我会说什么 如果你允许我
50	is that for someone who, or so you say,	这是给某人的,或者你说,
51	goes to the gym so much, don't really see it, mate, no.	这么多去健身房, 不真正看到它,伙计,没有。
52	Moving on, moving on, you'll hear from, sorry,	继续前进,继续前进, 你会听到来自,对不起,
53	you'll hear from Sultan Ben-Avshar,	你会听到苏丹本 -阿夫沙尔的消息,
54	a default in law and one of the hardest working	法律上的违约和工作最困难之一
55	press and sponsorship officers,	新闻和赞助干事,
56	and if I may say so, quite a good friend.	如果我可以这么说, 相当的好朋友。
57	Now, Sylvan, you know, just like the AI	现在,希尔文,你知道, 就像AI
58	he's agreed to besmirch, has this remarkable ability	他同意了博斯密奇,拥有这种非凡的能力
59	to, in this most efficient way, just flick	以这样最有效率的方式 轻轻地
60	from seemingly dauntless task to seemingly dauntless task,	从看似无畏的任务到看似无畏的任务,
61	be that, and I quote from him, helping his uncle	我引用他的话 帮助他叔叔
62	see a German auction house, to trying to convince	看到一家德国拍卖行 试图说服
63	Huntsman Salvo Rowe 重新油漆我们的一个房间。
64	进入Tweed的颜色。
65	They said no, I'm sorry.	他们说不,我很抱歉。
66	However, what I think really characterizes Sultan	然而,我认为苏丹的特征
67	is, shall we say, his two great obsessions in life,	可以说 他对人生的两大迷恋
68	those being the following two individuals.	他们是以下两个人。
69	One is the Kaiser, for whom he's promised	一个是凯撒,他答应给他的
70	to wear a pickle halber, and I was very disappointed	穿着腌黄瓜胡萝卜 我很失望
71	not to see you in one tonight.	今晚不见你
72	The other is a woman by the name of Aida,	另一个叫艾达的女人
73	you tell me is how you pronounce it.	你说你是如何发音的
74	Now, this is a woman for whom Sylvan,	现在,这是一个女人,Sylvan,
75	quite literally, organized an entire black tie	从字面上说, 整条黑领带
76	offer evening, with a champagne reception,	晚上开香槟招待会
77	just so he could try and woo her.	那样他才能试着追求她
78	And while you listen to him debate and vote as you go out,	你一边听他辩论一边投票一边出去
79	it's my sincere hope that you won't lead him on	我真心希望你不会引领他
80	just like she did.	就像她那样
81	现在,在那之后,你会听到 Jean Telingin教授,
82	sorry, Jean Telingin, the co-founding engineer of Skype,	对不起,让·泰林,斯凯普的联合创始工程师
83	and I would praise it, but I mean, come on, man,	我会称赞它,但我的意思是,来吧,男人,
84	we all use Zoom anyway.	我们都使用Zoom反正。
85	And finally, you'll hear from Professor Max Tegmark,	最后,你会听到马克斯·特马克教授的消息,
86	a professor at MIT and the president of the Future	麻省理工学院教授和未来的主席
87	of Life Institute, an organization created specifically	生活研究所,一个专门创建的组织
88	to warn about AI, so in other words,	警告AI,所以换句话说,
89	he literally pays people to speak for his side.	他付了钱让人们替他说话
90	And I had the pleasure of speaking to Professor Tegmark	我有幸和特格马克教授谈过
91	at dinner, and he's truly a brilliant man, really,	晚餐时,他是一个真正的聪明人,真的,
92	and I would keep complimenting him if he weren't a vegan,	如果他不是素食主义者 我会一直赞美他
93	so I'm told.	所以,我被告知。
94	Madam President, these are your speakers,	主席女士,这些是你的发言者,
95	and they are most welcome.	他们非常受欢迎。
96	Now, before I have any more fun onto the actual speech,	现在,在我对真正的演讲有任何乐趣之前,
97	so to my mind, I think, as I mentioned earlier,	因此,我认为,正如我前面提到的,
98	what's our best argument to favor this,	我们最好的论据是什么?
99	which is just all the benefits AI has for us	这就是AI对我们的全部好处
100	and all it can do for us.	它能为我们做的一切。
101	So these fine gentlemen, varying degrees of fine,	因此,这些优秀的先生们, 不同程度的罚款,
102	have tried to convince you that it's some evil bugbear,	曾经试图说服你 这是一个邪恶的虫熊,
103	it's some superhero villain flying in the sky,	一个超级英雄恶棍在天空飞翔
104	it's going to destroy our lives,	它会毁了我们的生活,
105	but I don't think they've quite appreciated,	但我不认为他们很欣赏,
106	not the experts, far more than me, of course,	当然不是专家,比我还多
107	just all it can do.	只是它能做的一切。
108	First and foremost, if this weren't enough on its own,	首先 如果单靠自己还不够的话
109	AI has the potential to completely revolutionize	AI有彻底革命的潜力
110	how we envisage and practice healthcare.	我们如何设想和实行保健。
111	Take, for example, you can have a world	比如说,你可以有一个世界
112	in which you can have AI driving so quickly,	你可以让AI开这么快,
113	in such an advanced manner, the development	以如此先进的方式发展
114	of medical diagnostics and treatments,	医疗诊断和治疗;
115	you know, treatments which will greatly improve	你知道,治疗 会大大改善
116	our ability to, well, live.	我们活下来的能力
117	Beyond that, I mean, we can catch fatal diseases early.	之后,我们就能及早感染致命疾病。
118	Just imagine a world where we have such commonplace AI	想象一下,我们的世界有这么普通的AI
119	systems that can analyze millions of patient records	能够分析数百万患者记录的系统
120	like that at the exact same time,	就像在同一个时间,
121	and then just treat it like there.	然后就像这样对待它
122	In fact, down the line, if it gets advanced enough,	事实上,在线上, 如果它得到足够的进步,
123	you could have a scenario where you don't even have	你可以有一个场景 你甚至没有
124	to come to the hospital, where the doctors	来医院,在那里的医生
125	can just get the system and diagnose your disease.	可以直接得到系统 并诊断你的疾病。
126	Who doesn't want to not go to the A&E ever again?	谁不想再不去A&E了?
127	Moving beyond that, if that weren't enough,	如果还不够的话,
128	I mean, it's also just so massively beneficial	我的意思是,它也是 如此巨大的好处
129	to us economically, you know, from sectors	我们的经济,你知道,从部门
130	to mechanical to agricultural to, yes,	机械到农业到,是的,
131	to all the finance bureaus Tom is trying to recruit,	给所有财务局 汤姆试图招募,
132	you know, how do you think your job's gonna function	你知道,你认为你的工作会如何运作
133	in 20 years?	20年来?
134	And then, not only that, not only that,	不仅如此 不仅如此
135	take what I think every single one of us here wants,	拿走我想我们每个人都想要的东西
136	a self-driving car.	自驾车
137	My friends, that is AI.	我的朋友,这是AI。
138	Not only will that stop us crashing into each other,	不仅能阻止我们相撞
139	well, with my driving skills, maybe not me,	凭我的驾驶技巧 也许不是我
140	but hopefully most of you, but that will make it	但希望你们大部分人,但这样就可以了
141	so much more convenient and crucially	所以更方便和关键
142	so much more safe for the world.	更安全的世界。
143	And if that were not enough, if all of that was not enough,	如果这还不够 如果这一切还不够的话
144	let's just take the example of climate change,	让我们以气候变化为例,
145	which I think we can all agree is a bit of an issue.	我认为我们都同意,这是一个问题。
146	Now, with AI, from how it can optimize	现在,用AI,从如何优化它
147	our energy consumptions so we don't overheat the planet,	我们的能源消耗,所以我们不会过热地球,
148	or how it can literally, you know,	或者它如何可以 字面上,你知道,
149	through its programs, quite literally predict	通过其程序,非常字面的预测
150	the natural disasters that will inevitably occur,	将不可避免地发生的自然灾害,
151	we will save tons and tons of lives.	我们将拯救很多人的生命。
152	I mean, how can that be an existential threat?	我的意思是,这怎么可能是一个生存的威胁?
153	And in fact, dare I say, if anything, at this point,	事实上,我敢说,如果有什么, 在这一点上,
154	there'll be a necessity.	会有需要的
155	Now, indeed, it's, as I mentioned earlier,	现在,确实,正如我前面提到的,
156	it's such a big part of all of our lives already.	这已经是我们生命中的一大部分了
157	I mean, it's not some existential bugbear	我的意思是,这不是一些存在虫熊
158	we haven't heard of.	我们没有听说过。
159	I mean, I just ask, how many of you have had to suffer	我的意思是,我只是问, 你们有多少人不得不受苦
160	through Syria on your phone, AI?	在你的电话里通过叙利亚,AI?
161	Or how many of us have spent hours,	或者我们有多少人花了几个小时
162	and in my case, literally hours,	就我而言 几个小时
163	on the other side of some AI chat box customer service?	在AI聊天盒客户服务的另一边?
164	Like, once again, my friends, that is AI.	就像,再次,我的朋友, 这是AI。
165	Or take Amazon, which I think is quite literally	或者拿下亚马逊, 我认为这是相当的字面
166	the one website we've probably all used.	我们大概都用过的网站
167	I mean, a lot of that runs by AI.	我的意思是,很多 运行由AI。
168	I mean, you go on Amazon, you have your recommended.	我的意思是,你去亚马逊, 你有你的推荐。
169	这是AI。
170	Or take, maybe I'm projecting a bit too much,	或者拿,也许 我投射有点过分,
171	或拿YouTube。 (笑声)
172	YouTube推荐人工智能.
173	And I can say, at least in my case,	我可以说,至少就我而言,
174	I've spent many a free AM watching YouTube recommended	我花了很多时间免费观看YouTube推荐的节目
175	instead of doing my essay.	而不是做我的论文。
176	My friends, again, that is AI.	我的朋友,再次,这是AI。
177	AI is not some, yes, there's a lot we don't know about it.	AI不是一些,是的,有很多 我们不知道这一点。
178	As a classicist, trust me, I know that.	作为一个经典主义者, 相信我,我知道这一点。
179	It's such a big part of our lives.	这是我们生活的一大部分。
180	And it's not, I think that's what I was saying.	而不是,我想这就是我说的。
181	It's not some, you know, it's not some brave unknown.	这不是一些,你知道, 这不是一些勇敢的未知。
182	It is unknown.	这是未知的。
183	In fact, it's such, it's become so much our normal	事实上,它是如此, 它已经变得如此正常
184	that to just, as I suppose Sultan would want us to do,	那只是,我想苏丹会希望我们这样做,
185	or any of these other very fine speakers,	或任何这些非常优秀的演讲者,
186	just throw it out as some threat,	把它当作威胁扔出去
187	that's not practicable, and that's not practicable.	这是不实际的, 这不实际的。
188	We can't do that without radically altering	我们不能不彻底改变
189	the way we live.	我们的生活方式。
190	And finally, finally, as I mentioned very briefly,	最后,我简短地提到,
191	and I shall hopefully expand on,	我希望能继续
192	as the gentleman so gracefully asked me,	正如这位绅士这么优雅地问我,
193	yes, AI is not perfect.	是的,AI不是完美的。
194	There's a lot we don't know about it.	众生无量.
195	It will change tremendously in the next few years.	今后几年将发生巨大变化。
196	You know, I allegedly do lash and agree.	你知道,我据称做 斜线和同意。
197	I can't tell you how.	我不能告诉你怎么做。
198	These guys may be, but not me.	这些人可能是,但不是我。
199	Yes, it's not perfect.	是的,这是不完美的。
200	But there is a wide, wide goal	但有一个广泛的目标
201	between having issues and being an existential threat.	在有问题和生存威胁之间
202	And when you vote, when you walk through those doors,	当你投票时 当你穿过那些门
203	you don't have to think AI is great.	你不必认为AI很棒
204	You don't even have to think it's the best thing ever.	你甚至不用认为这是有史以来最好的。
205	You just have to think it's not Ultron or Thanos	你只要认为不是超能或塔诺斯
206	or some Dune-esque machine trying to kill you,	或者是那些想杀你的人
207	as in not an existential threat.	而不是生存威胁
208	And with that, I commend this motion to the House.	因此,我向众议院推荐这项动议。
209	Thank you very much.	谢谢

该视频共有字幕 209 条。解锁更多字幕为会员功能，请移动到 价格

P3 · 【高清完整版｜双语字幕】牛津大学英辩 AI 是否构成人类生存威胁 p03 3 (P3)

【高清完整版｜双语字幕】牛津大学英辩 AI 是否构成人类生存威胁 p03 3 (P3)

发布于 2026-07-22 · 时长 9:07

字幕摘录

时间	英文	中文
0:04	Thank you, Madam President, ladies and gentlemen of the House.	谢谢你,主席女士,女士们,先生们。
0:09	Humanity is akin to a teenager, with rapidly developing physical abilities, lacking wisdom	人类与青少年相似,身体能力迅速发展,缺乏智慧
0:16	and self-control, little thought for its long-term future, and an unhealthy appetite for risk.	和自我控制, 很少考虑它的长期未来, 和对危险的不健康欲望。
0:24	Toby Orr, precipice.	托比・奥尔, 悬崖。
0:27	Toby is, of course, a philosopher in Oxford, and precipice is not the only landmark book	托比当然是牛津的哲学家 早熟不是唯一的标志性著作
0:33	to have originated from this place.	来源于这个地方。
0:37	In fact, when I helped to start the Centre for the Study of Existential Risk in that	事实上,当我帮助启动 存在风险研究中心
0:43	other place, one of my explicit goals was to help that other place to catch up to Oxford	我明确的目标之一是帮助那个地方赶上牛津大学
0:53	when it comes to study about existential risk.	在研究存在风险时
0:57	So arguing about AI being an existential risk in front of an Oxford audience is a bit like	所以在牛津的观众面前 争论AI是存在的风险 有点像

展开字幕全文（108 条）

序号	英文	中文
1	Thank you, Madam President, ladies and gentlemen of the House.	谢谢你,主席女士,女士们,先生们。
2	Humanity is akin to a teenager, with rapidly developing physical abilities, lacking wisdom	人类与青少年相似,身体能力迅速发展,缺乏智慧
3	and self-control, little thought for its long-term future, and an unhealthy appetite for risk.	和自我控制, 很少考虑它的长期未来, 和对危险的不健康欲望。
4	Toby Orr, precipice.	托比・奥尔, 悬崖。
5	Toby is, of course, a philosopher in Oxford, and precipice is not the only landmark book	托比当然是牛津的哲学家 早熟不是唯一的标志性著作
6	to have originated from this place.	来源于这个地方。
7	In fact, when I helped to start the Centre for the Study of Existential Risk in that	事实上,当我帮助启动 存在风险研究中心
8	other place, one of my explicit goals was to help that other place to catch up to Oxford	我明确的目标之一是帮助那个地方赶上牛津大学
9	when it comes to study about existential risk.	在研究存在风险时
10	So arguing about AI being an existential risk in front of an Oxford audience is a bit like	所以在牛津的观众面前 争论AI是存在的风险 有点像
11	preaching Christianity in Vatican.	在梵蒂冈宣扬基督教。
12	Not to mention that there is an increasing consensus now, sandwiched between Alan Turing	更何况艾伦·图灵之间 达成的共识越来越大
13	in 1951 predicting that we should expect to lose control to machines, and the inventor	1951年我们预测 机器和发明家会失去控制
14	of deep learning itself, Geoff Hinton, starting to have doubts about his life work.	杰夫·欣顿 开始怀疑他的生活工作
15	There are now hundreds of AI experts sounding their alarm bells.	现在有数百位AI专家在敲响警钟.
16	A recent poll found that 88% of AI engineers think that AI could destroy the world, and	最近一项民意调查发现,88%的AI工程师认为AI可以毁灭世界,
17	another one reported that 76% of American voters believed that AI was a threat to our existence.	76%的美国选民认为AI对我们的生存构成威胁。
18	Just yesterday, there was news that one of the leading superforecaster groups, Somad	就在昨天,有消息说, 一个主要的超级福雷克集团,索马德
19	Sveti, published their prediction that their estimate for AI catastrophic risk is 30%.	Sveti,发表了他们的预测,他们估计AI的灾难性风险为30%.
20	30%。 30%。
21	The battle for establishing that AI is an existential risk, a battle that I've spent	确定AI的战斗是存在风险,我花费了这场战斗
22	roughly 15 years of my career on, has now all but won.	大约15年的职业生涯, 已经取得了全部的胜利。
23	So I've set myself a bit more ambitious goal today, argue that the extinction risk from	所以我今天为自己设定了一个更雄心勃勃的目标, 认为灭绝的风险来自
24	God-like AI is not just possible, but imminent now, and we urgently, urgently need to change	上帝般的人工智能不仅是可能的,而且现在迫在眉睫,我们迫切需要改变
25	our trajectory.	我们的轨道。
26	I'm going to show that there are fundamental reasons why unaligned, God-like AI will not	我要表明, 根本原因 为何不结盟, 像上帝一样的AI不会
27	leave any survivors, that they are now close to such AI while having no idea how to align	离开任何幸存者,他们现在接近这种AI,却不知道如何调整
28	it, and how the counter arguments, so unfortunately, are sadly extremely weak.	不幸的是,这些辩词非常软弱。
29	Of course, I will need to be brief, so my arguments will be very condensed.	当然,我需要简短发言,因此我的论点将非常简洁。
30	The reason why I expect God-like AI, and by that I mean an AI that can conceive and carry	我期待上帝般的人工智能的原因 我指的是一个可以想象和携带的人工智能
31	out much better plans than any subgroup of humans can.	比任何人类群体都更好的计划
32	The reason why I expect God-like AI to be utterly lethal is that by definition, it can	我之所以期望像上帝一样的人工智能完全致命 是因为从定义上讲,它可以
33	control our environment, rather than hacking into our computer, which of course it also	控制我们的环境,而不是黑进我们的计算机, 当然它也
34	can do.	说得没错
35	Think about hacking into algae to make it produce what it wants, then venting the atmosphere,	想想黑进藻类 让它产生它想要的东西 然后释放大气
36	and finally pulling out the hydrogen from the sun.	最后从太阳中抽出氢气
37	It's easy to show that such takeover happens in toy models.	很容易证明这种收购发生在玩具模型中.
38	Indeed, in 2016 I gave a presentation just about that.	事实上,在2016年,我刚刚做了关于这一点的介绍。
39	Nor should we forget that humans have behaved similarly to countless other species, species	我们也不应忘记,人类的行为 类似无数其他物种,物种
40	who were unfortunate, unlucky, do not have sapiens in their name.	他们是不幸的,不幸运的, 他们的名字里没有仙人。
41	Now the reason why I expect God-like AI to not care about humans has to do with that	现在为什么我期望像上帝一样的AI 不关心人类的原因 和那有关
42	dirty secret of AI industry.	AI行业的肮脏秘密.
43	The front-air AIs are not built, they are grown.	前空AI不是建造的,而是种植的.
44	ChatGPT的P代表预训.
45	Pre-training, perhaps we should call it summoning, is a process where a simple two-page program	训练前,也许我们应该称之为召唤,是一个简单的两页程序的过程。
46	is soaked in terabytes of data and megawatts of electricity, and left like that for months.	被浸泡在几兆字节的数据和兆瓦的电力中, 并像这样离开了几个月。
47	And then after that, attempts are made to tame the emergent alien mind.	之后,有人试图驯服新生的外星人。
48	Importantly, those methods of taming rely on the AI being less competent than the humans	重要的是,这些驯服方法依赖于人工智能不如人类
49	who are taming it.	他们正在驯服它。
50	Now the reason why I expect that we are close to God-like AI is that the trend of AIs getting	现在为什么我期望我们接近上帝般的AI 是因为AI得到的倾向
51	more powerful is now visible to everyone, it's obvious.	现在每个人都看得见了 很明显
52	看看GPT-2,GPT-3,GPT-4之间的能力差异.
53	GPT-2 was released in 2019.	GPT-2于2019年发布.
54	A simple extrapolation would take us to GPT-7 before this decade is over.	简单的推论将使我们在本十年结束前进入GPT-7.
55	So in summary, we are blindly growing increasingly competent minds, while hoping that they are	因此,简言之,我们盲目地增长 能力越来越强,同时希望他们是
56	not so competent that they spin out of control and destroy our living environment.	没有那么能干,他们就会失控 破坏我们的生活环境
57	Unfortunately, that hope is not justified, which explains things, increasing anxiety	不幸的是,这种希望是没有道理的,这解释了事情,增加了焦虑。
58	among the AI developers themselves.	在AI开发者自己中.
59	So basically, what I'm telling you is that on the current AI trajectory, you are not	所以基本上,我要告诉你的是, 在目前的AI轨迹上,你不是
60	going to live very long.	将活得非常长。
61	Of course, at this point, just like a patient that has received a terminal diagnosis, you	当然,此时此刻, 就像一个病人 已经接受了末期诊断,你
62	are very, very much encouraged to seek for a second opinion.	非常非常鼓励寻求另一种意见。
63	Unfortunately, having been part of this debate for more than a decade, I already know what	不幸的是,参加这次辩论十多年了,我已经知道
64	you're going to hear.	你会听到的。
65	Almost all counterarguments fall into four categories.	几乎所有反证都属于四类.
66	First, labeling.	首先,标签。
67	These are arguments, oh, this is science fiction, this is alarmism, but also the oldie but goodie	这些是争论,哦,这是科幻小说,这是惊慌失措, 但也老了,但很好
68	ad hominem.	帅呆了
69	These are doomsayers.	这些是说末日的人。
70	Don't listen to people with that non-virtuous property X, or these are people who say X	别听别人说X,不然就是说X
71	because of Y.	因为Y.
72	Second, frame control.	第二,帧控制。
73	AI is like X, and X is very nice, right?	AI就像X,X是非常好的,对不对?
74	And this is also the answer to your question.	这也是你问题的答案。
75	This has now reached grotesque levels.	现在,这种情况已经到了奇特的地步。
76	One prominent VC claimed recently that AI is basically just math, so why should we worry?	一个著名的越共最近声称AI基本上是数学,所以我们为什么要担心呢?
77	Imagine the captain of Titanic announcing, don't worry, passengers, this is just water.	想象一下泰坦尼克号的船长宣布 别担心 乘客们 这只是水
78	Third, class of arguments, human supremacy.	第三,阶级争论,人类至上.
79	AI can never do X.	AI永远不能做X.
80	Or more moderately, we are very far from AI doing X.	更温和一点 我们离AI做X还很远
81	I guess this is a little bit response to your question.	我想这是对你问题的回答
82	This category actually endorsed, as reality can act as a judge here.	这一类别实际上得到认可,因为现实可以在这里充当法官。
83	Unfortunately, reality has been very harsh judge here recently.	不幸的是,现实最近在这里变得非常严厉。
84	The set of things that only humans can do is collapsing really rapidly.	只有人类才能做的那套事情 很快就要崩溃了
85	And third one, or sorry, fourth category is topic change.	第三,或者抱歉,第四类是话题变化。
86	I am more worried about X than AI extinction risk, so let's talk about X instead.	比起AI的灭绝风险,我更担心X,所以让我们来谈谈X.
87	Or like, let's talk about the benefits of AI instead, as you just heard from the opposition.	或者说,我们来谈谈AI的好处, 正如你刚刚听到的反对派。
88	Unfortunately, usually the other topic, sometimes very important.	不幸的是,通常另一个专题,有时非常重要。
89	I agree that there are other problems.	我同意还有其他问题。
90	I agree that there are a lot of benefits from AI, but they're non-sequiturs.	我同意AI有很多好处,但他们不是定格者.
91	They don't have any bearing on AI extinction risk.	它们与AI灭绝风险没有任何关系.
92	So here's a bingo card for the opposition arguments today.	这是今天反对派辩论的宾果卡
93	Labeling, frame control, human supremacy, and topic change.	标签 帧控制 人类至上 主题改变
94	Have fun.	玩得开心点
95	Or it would be fun and games if it weren't so tragic.	或者它会很有趣和游戏 如果不是那么悲惨。
96	It's like dashing the hopes of that cancer patient by pointing out that	这就像打破了那个癌症患者的希望 指出:
97	the second opinion, while comforting indeed, was simply wrong.	第二种观点虽然确实令人欣慰,但完全是错误的。
98	Still, I think I can leave you with a hopeful note.	但我还是可以给你留个希望的字条
99	There's now growing global consensus that the unregulated	越来越多的全球共识认为 不受管制的
100	blind AI scaling is reckless and dangerous.	盲目的AI缩放是鲁莽和危险的.
101	So we need to constrain it or ban it altogether.	因此,我们需要加以限制或完全禁止。
102	Just like we banned human cloning.	就像我们禁止人的克隆
103	Regardless of which door you walk out today,	无论你今天走哪扇门
104	once you're on the other side of that door,	一旦你站在门的另一边
105	please take a moment to think if and how you can help.	请花点时间 想想你能不能和如何帮忙
106	You have received a terminal diagnosis.	你已经接受了终极诊断
107	Please don't simply ignore it.	请不要忽视它。
108	Thank you.	谢谢

该视频共有字幕 108 条。解锁更多字幕为会员功能，请移动到 价格

P4 · 【高清完整版｜双语字幕】牛津大学英辩 AI 是否构成人类生存威胁 p04 4 (P4)

【高清完整版｜双语字幕】牛津大学英辩 AI 是否构成人类生存威胁 p04 4 (P4)

发布于 2026-07-22 · 时长 13:27

字幕摘录

时间	英文	中文
0:00	Thank you.	谢谢
0:02	I want to thank Madam President, Chief of Staff Davis, and Aliyah for the invite.	我要感谢主席女士、戴维斯参谋长和阿利亚的邀请。
0:07	It is wonderful to see young women at the helm of one of the most important political	看到年轻女性领导着最重要的政治人物之一 真是太好了
0:11	institutions at a critical juncture in our world history.	我们世界历史上一个关键时刻的机构。
0:15	Thank you.	谢谢
0:16	My name is Yeshe Milner, and I am the founder and executive director of Data for Black Lives.	我的名字是Yeshe Milner, 我是Data的创始人和执行董事 黑人生活。
0:22	We are a movement of scientists and activists working to make data a tool for social change	我们是一个科学家和积极分子的运动 努力使数据成为社会变革的工具
0:28	instead of a weapon of political oppression.	而不是政治压迫的武器。
0:31	Today, it's with honor and actually great pleasure of mine to argue against the motion	今天 我很荣幸 也很高兴能反对这个动议
0:37	of the House that AI is an existential threat.	众议院认为,大赦国际是一种生存威胁。

展开字幕全文（160 条）

序号	英文	中文
1	Thank you.	谢谢
2	I want to thank Madam President, Chief of Staff Davis, and Aliyah for the invite.	我要感谢主席女士、戴维斯参谋长和阿利亚的邀请。
3	It is wonderful to see young women at the helm of one of the most important political	看到年轻女性领导着最重要的政治人物之一 真是太好了
4	institutions at a critical juncture in our world history.	我们世界历史上一个关键时刻的机构。
5	Thank you.	谢谢
6	My name is Yeshe Milner, and I am the founder and executive director of Data for Black Lives.	我的名字是Yeshe Milner, 我是Data的创始人和执行董事 黑人生活。
7	We are a movement of scientists and activists working to make data a tool for social change	我们是一个科学家和积极分子的运动 努力使数据成为社会变革的工具
8	instead of a weapon of political oppression.	而不是政治压迫的武器。
9	Today, it's with honor and actually great pleasure of mine to argue against the motion	今天 我很荣幸 也很高兴能反对这个动议
10	of the House that AI is an existential threat.	众议院认为,大赦国际是一种生存威胁。
11	In a time of genocidal warfare, extreme income inequality driven by inflation and other economic	在种族灭绝战争的时代,通货膨胀和其他经济、社会、文化权利导致的极端收入不平等。
12	practices, and at the same time immense political division, I'm here to argue that the true	同时也是巨大的政治分裂 我在这里认为
13	existential threat is the concentration of power in the form of data in the hands of	存在威胁是数据形式的力量集中在
14	a few.	几个
15	You see the opposition nodding, so there you go.	你看反对派点头,所以你去那里。
16	Before I get into my argument, I want to define existential threat as stated in the motion,	在我开始辩论之前,我想像动议中所说的那样定义存在威胁,
17	but also in the context of AI.	也涉及到大赦国际。
18	It is not a blanket vanilla term, but one grounded in the resurgence of a distinct philosophical	这不是一个一概而论的梵语,而是基于一种独特的哲学的复兴。
19	tradition, eugenics.	传统,优生。
20	It's interesting that the opposition opened with the paperclip experiment because that's	有趣的是,反对者打开 与纸剪的实验,因为
21	a concept authored by Mr. Nick Bostrom, the founder, one of the founders of what's called	一个由Nick Bostrom先生起草的概念, 创始人, 创始人之一
22	the X-Crisis Movement.	X危机运动。
23	He defines existential risk as an event that could cause human extinction or permanently	他将存在风险定义为 可能导致人类灭绝或永久灭绝的事件
24	or drastically curtail humanity's potential.	或大幅度削减人类的潜力。
25	This is different from a global catastrophic risk, as he defines it, which we can recover	这不同于全球的灾难性风险,正如他所定义的,我们可以恢复
26	from, and the use of the term existential threat instead of risk is actually as a linguistic	而使用生存威胁而不是风险实际上是一种语言
27	trope in order to make the threat appear more dire.	为了让威胁显得更可怕
28	He builds on these ideas of existential risk using transhumanist concepts like disogenic	他利用非原生性等反人类主义概念 来建立存在风险的思想
29	pressures.	压力。
30	Disogenic is the opposite of eugenic, and it's based on the theory that people who were	脱氧核糖核酸与优生素相反,其理论依据是:
31	deemed inferior categorically, racially, would reproduce faster.	被认为是低劣的,绝对的,种族的,会更快的繁殖。
32	They're the threat.	他们是威胁
33	Disogenic pressures are only solved by eugenics practices.	脱氧核糖核酸压力只能通过优生法解决.
34	Eugenics was an international movement articulating genetic hypotheses, research, and immense	Eugenics是一个国际运动 阐述遗传假设,研究,和巨大的
35	policy prescriptions.	政策处方。
36	During the first three decades of the 20th century, eugenics, along with racism and imperialism,	在20世纪的前30年里,优生主义者,连同种族主义和帝国主义,
37	had profound influence on thinking in Europe, it originated in the UK, but in Germany, France,	对欧洲的思想产生了深远的影响, 它起源于英国, 但在德国,法国,
38	Scandinavia, and North America, where I live in the US, and there it became more than a	斯堪的纳维亚和北美,我住在美国, 那里变得不止一个
39	theory but a political and ideological regime.	理论只是政治和意识形态制度。
40	While Bostrom or anyone in the singularity or ex-crisis movement didn't invent eugenics,	虽然博斯特罗姆或者任何人在奇异性或前危机运动中并没有发明优生学,
41	it lasted well before their time, and if we were to be honest, it shapes the very data	它在他们的时间之前就已经持续了很久, 如果我们要诚实,它塑造了数据
42	science and scientific management practices of today.	今天的科学和科学管理实践。
43	So there's actually another context to this definition of existential risk that I want	因此,实际上还有另一个背景 定义的存在风险,我想要
44	to bring in because it's not what's being said and defined, it's also what's been redacted.	因为它不是被描述和定义的 而是被修改的
45	Bostrom admitted to writing a racist email where he stated, take for example the following	Bostrom承认他写了种族主义电子邮件,他说,例如:
46	sentence, blacks are more stupid than whites.	句子上,黑人比白人更蠢.
47	I like that sentence, and I think it's true, but recently I've begun to believe that I	我喜欢这句话 我觉得这话是真的 但最近我开始相信
48	won't have much success with that, with most people if I speak like that, that they would	没有什么成功的, 对于大多数人,如果我这样说,他们会
49	think that I were racist, that I disliked black people and thought it was fair if blacks	认为我是种族主义者 不喜欢黑人 认为黑人公平
50	were treated badly, I don't.	我没有被虐待
51	It's just that based on what I have read, I think it's probable that black people have	只是根据我所读到的 黑人很可能有
52	a lower average IQ than mankind in general, and I think that IQ is highly correlated with	平均智商低于一般人类,我认为智商与
53	what we normally mean by smart and stupid.	我们通常所说的聪明和愚蠢
54	I can go on with this email, but he ends it by using the N-word, which he does not redact	我可以继续用这封邮件,但他用N字结束,他不编辑
55	by the way.	顺便说一句
56	I don't just come to you as someone who's the movement of an organization that I started	我不只是来找你的 我开始的组织运动
57	in 2017 to counteract the ways in which data is weaponized against black people, I come	在2017年,为了对抗 数据武器化的方式针对黑人,我来
58	to you as somebody who understands quite literally this idea of risk.	作为你的人,谁 字面上理解 这种风险的想法。
59	The very first time I even heard the word risk or threat was at nine years old.	我第一次听到风险或威胁这个词是在9岁。
60	In the computer lab at my elementary school, another young student said, I'm at risk.	在我小学的计算机实验室里,另一个年轻学生说,我有危险。
61	Shocked, I said, what are you talking about, you're at risk, like, are you okay?	吓坏了,我说,你在说什么, 你处于危险,喜欢,你没事吧?
62	Because she was enrolled in an at-risk, an after-school program for at-risk youth, she	因为她参加了一个针对有风险的青年的课后方案,所以她
63	believed, what they had told her was that she would one day be at risk of early pregnancy,	他们告诉她 有一天她会有早孕的危险
64	prison, early death, joining a gang, and to be honest, for so many of the young people	监狱 早死 加入帮派 老实说 对这么多年轻人来说
65	that I went to elementary school and high school with, that became a self-fulfilling	我上小学和高中, 成为了自我实现
66	prophecy.	预言
67	While I survived what we know in the U.S. as a school-to-prison pipeline, many people	虽然我幸存了我们在美国所知道的 学校到监狱的管道, 许多人
68	did not, and my high school graduation, the first row were of empty seats to commemorate	我高中毕业时,第一排是空座位来纪念
69	people who died too soon.	那些死得太早的人们
70	As a young black person, I've learned how to navigate this world, where to be black	作为一个年轻的黑人,我已经学会了 如何导航这个世界, 哪里是黑人
71	is to be a problem, to be a threat, to be a risk, and these ideas that I have to encounter	是一个问题,一个威胁, 一个风险, 这些想法,我必须遇到
72	and fight as stereotypes of just the mere fact of being dark or pigmented have been	战斗的刻板印象 仅仅是黑暗或颜料
73	automated into the public imagination in the very same way that they've been encoded into	自动进入公众的想象 就像他们被编码成
74	powerful artificial intelligence technologies of today and weaponized by those who currently	当今强大的人工智能技术 和武器化 目前的人
75	have control and power.	拥有控制和权力。
76	One example that I'll use is of credit scores.	我要用的一个例子是信用分数。
77	I brought this demand to the White House just this month, to Congress, to abolish FICO,	我本月向白宫 向国会 提出废除FICO的要求
78	which is the predominant, most powerful algorithm at work in our country.	这是我国最强大的算法
79	FICO is the Farran Isaac Corporation.	FICO是法兰艾萨克公司.
80	It was started by a mathematician and an engineer 25 years ago to use artificial intelligence	由数学家和工程师在25年前 开始使用人工智能
81	as a way to eliminate human bias when providing people access to credit.	在为人们提供信贷机会时消除人的偏见。
82	The inputs to the FICO algorithm, as we're told, are the amount of debt we have, the	对FICO算法的投入,正如我们所被告知的, 是我们的债务金额,
83	percentage of mispayments.	误付的百分比。
84	All of this information are provided through a collusion of data brokers, Experian,	所有这些信息都是通过数据经纪人Exeerian的勾结提供的。
85	Equifax,和TransUnion,Experian,实际上是一个设在英国的公司.
86	And then they're fed into the FICO algorithm.	然后它们被输入FICO算法.
87	According to the shareholder reports at FICO, their scores are their number one product.	根据FICO的股东报告,他们的得分是他们的头号产品.
88	Without a credit score, you can't rent a house, you can't qualify for student loans,	没有信用分数,不能租房,不能取得学生贷款资格.
89	you can't, increasingly you can't get a job in America.	你不能, 你越来越找不到工作 在美国。
90	And black people are three times more likely to be scored at below 620 than white people.	而黑人在620以下得分的可能性是白人的三倍.
91	Even when controlling for education, debt, and income.	即使控制教育、债务和收入。
92	FICO scores reflect the ways in which algorithms hold tremendous power over our lives.	FICO的分数反映了算法对我们生命的强大作用。
93	And as I mentioned, the impact is not on everyone.	正如我所提到的, 影响不是每个人。
94	Yes, there is a threat, but it's not evenly distributed.	对,有威胁,但分布不均
95	Right here in the UK, according to the Color of Money report, black and Bangladeshi families	在英国,根据"钱的颜色"报道 黑人和孟加拉人的家庭
96	have 10 times less wealth than white families.	财富比白人家庭少10倍
97	Black Caribbean families have 20 pence for every one dollar made by white families.	加勒比黑人家庭每1美元白种家庭有20便士。
98	And to even go back to the US example, what's the impact of credit scores?	连美国的例子 信用评级的影响是什么?
99	Black people are 13% of the population, but represent 55% of unhoused.	黑人占人口的13%,但占无人居住的55%。
100	And I speak about this with passion because I grew up with a single mother who put herself through	我满怀激情地谈论这件事 因为我从小就跟单身母亲在一起
101	college, graduate school, send her kids to Ivy League colleges on a full ride,	大学 研究生院 送孩子去常春藤联盟大学
102	was a leader in her community.	是她社区的领导者
103	But because of these black box algorithms that are racially encoded and quite powerful	但是由于这些黑匣子算法 种族编码和相当强大
104	with no recourse, we often were homeless, all because of a three digit number.	我们经常无家可归,都是因为三位数。
105	While it is in violation of federal law to deny people housing, employment,	虽然剥夺人们的住房、就业是违反联邦法律的,
106	and education based on race, you can't sue an algorithm.	和基于种族的教育, 你不能起诉一个算法。
107	Algorithms have now become not godlike, not super intelligent,	算法已经变得不像神了 也不是超级聪明
108	but the way for people to be racist without being a racist.	但人们在种族主义而不成为种族主义者的方式。
109	By definition, what is an algorithm?	顾名思义,算法是什么?
110	This is what makes up artificial intelligence.	这就是人工智能的构成.
111	It's a step by step process to solve a problem.	这是一步一步地解决问题的过程.
112	A recipe is an algorithm, a list of instructions to make the dish,	食谱是算法 制作菜肴的指令列表
113	the ingredients that make up the dish, and a result based on what we define as success.	构成菜肴的成分 以及基于我们定义的成功的结果
114	Whether we want to focus on making something healthy	我们是否想专注于 做一些健康的东西
115	or something that just tastes good regardless of health,	或什么只是味道好 无论健康,
116	these decisions are determined by a question.	这些决定由问题决定。
117	What are we optimizing?	什么是我们优化?
118	Algorithms are not just input and output.	算法不仅仅是输入和输出.
119	They're based on the objective function.	它们基于客观的功能.
120	What are we optimizing?	什么是我们优化?
121	For computational algorithms, what we are discussing is much more complex,	对于计算算法,我们正在讨论的是 更复杂的,
122	but the function remains the same.	但功能不变。
123	Today, I ask all of you, are we optimizing for a future where data, AI is being used	今天,我问大家,我们是否在优化未来, 利用数据,AI
124	as a tool to promote equity, fairness, justice, and for everybody,	作为促进公平、公平、公正和人人平等的工具,
125	including communities of color, or a techno-feudalist society	包括有色人种社区,或技术 -- -- 乡村社会
126	where a machine learning model doles out mass suffering?	机器学习模型能消除大规模痛苦?
127	To quote a political theorist that we love,	引用我们所爱的政治理论家的话
128	Achille Mbembe在实际威胁方面做了大量工作,
129	especially on the idea of necropolitics.	特别是黑社会的想法
130	Necropolitics is the use of social and political power	黑社会就是利用社会和政治权力
131	to dictate how some people may live and how some people must die.	来决定一些人如何生活 以及一些人如何死亡
132	The deployment of necropolitics create what we call death worlds,	利用黑社会 创造了我们所谓的死亡世界
133	or new and unique forms of existence with violations	或新的和独特的存在形式与侵权行为
134	are subjected to living conditions that confer upon them the status of the dead.	其生活条件应赋予死者地位。
135	I come to you from the land of the dead,	我从死人的国度来找你,
136	communities where people have been stripped of opportunities,	人民被剥夺了机会的社区,
137	where the fire hoses and police dogs of the past	过去消防管和警犬
138	have been replaced with the FICO algorithms,	已替换为 FICO 算法,
139	scoring models, and risk assessments of the present.	评分模型和当前风险评估。
140	We are not afraid of AI becoming human.	我们不怕AI变成人类。
141	What we are afraid of, of how it is going to be used, and it has been used,	我们所担心的, 如何使用它, 它已被使用,
142	to reinforce and justify long-held beliefs of who is human and who is not.	强化并证明长期持有的信念,即谁是人类,谁是非人类。
143	Who deserves the rights and privileges afforded by not just sentience, but life?	谁应当享有不仅是哨兵,而且是生命所赋予的权利和特权?
144	We need to pause.	我们需要暂停。
145	We don't need to pause, AI.	我们不需要暂停,AI。
146	But we need to abolish big data.	但我们需要废除大数据。
147	We need to dismantle the structures that put the power of data into the hands of a very few.	我们需要拆除那些把数据的力量掌握在少数人手中的结构。
148	Yes.	对
149	No, I'm going to get to what I'm going to talk about, so let me finish.	不,我要去得到什么 我要谈论,所以让我说完。
150	Because, let me say it, if we focus on AI as an existential threat,	因为,让我说,如果我们专注于AI作为一个存在威胁,
151	we miss the opportunity to put these powerful tools into the hands of people who need it the most.	我们错过了把这些强大的工具 交给最需要它的人的机会。
152	At Data for Black Lives, we are reclaiming data as protest, as accountability,	在Data for Black Lives, 我们正在重新找回数据 作为抗议,作为问责,
153	and as collective action.	作为集体行动。
154	Through my work, which you can read about online,	通过我的工作 你可以在网上读到
155	we've been able to shape the role that data plays in public life.	我们能够塑造数据在公共生活中扮演的角色。
156	And we want to invite all of you to not just join us in voting for the opposition,	我们想邀请你们所有人 不只是和我们一起投票支持反对派,
157	but join our movement to finally change the cast of characters,	但加入我们的运动 最终改变角色的塑造,
158	of not just who's at these debates,	不只是谁参加这些辩论,
159	but who's at the table around how these technologies are made.	但谁在座 围绕如何开发这些技术。
160	Thank you.	谢谢

该视频共有字幕 160 条。解锁更多字幕为会员功能，请移动到 价格

P5 · 【高清完整版｜双语字幕】牛津大学英辩 AI 是否构成人类生存威胁 p05 5 (P5)

【高清完整版｜双语字幕】牛津大学英辩 AI 是否构成人类生存威胁 p05 5 (P5)

发布于 2026-07-22 · 时长 9:01

字幕摘录

时间	英文	中文
0:06	Thank you very much madam president for giving me this great honor to speak once again in this great and venerable chamber on tonight's motion	非常感谢你总统女士 赐予我这个殊荣 在今晚的演讲室里
0:13	On a personal note might say it was a great pleasure to work with you over the last couple of months	就个人而言,也许可以说在过去几个月里与你合作是件非常愉快的事
0:17	And I congratulate you for the fantastic term card you've put together	恭喜你把名词卡拼起来
0:22	Ladies and gentlemen for the past 200 years this society has debated the most pressing issues of its time and	女士们,先生们,在过去的200年里,这个社会已经辩论了它时代最紧迫的问题。
0:29	Tonight's debate on the implications of artificial intelligence for the way we live is doubtlessly one of the most pressing issues of our time	今晚关于人工智能对我们生活方式的影响的辩论无疑是我们时代最紧迫的问题之一
0:39	Before I move into my main argument I would also like to say some words on the motion and the burden which the proposition bears tonight	在我开始讨论我的主要论点之前,我还想就今晚的提议所肩负的动议和负担说几句话。
0:48	This house believes artificial intelligence is an existential threat	这房子相信人工智能是一种生存威胁
0:51	We do not know what the threat is to but from a sort of general understanding I think we need to adopt a very high burden here materially	我们不知道什么是威胁,但从某种一般的理解来看,我认为,我们需要在物质上承担一个非常沉重的负担。
1:00	And assume that we need to show that artificial intelligence is an existential threat to the way we live and to core values which we hold dear	假设我们需要证明 人工智能是一个存在性的威胁 我们的生活方式 和核心价值观 我们所珍视
1:11	However we must not show that this will inevitably happen	然而,我们决不能表明,这种情况将不可避免地发生

展开字幕全文（108 条）

序号	英文	中文
1	Thank you very much madam president for giving me this great honor to speak once again in this great and venerable chamber on tonight's motion	非常感谢你总统女士 赐予我这个殊荣 在今晚的演讲室里
2	On a personal note might say it was a great pleasure to work with you over the last couple of months	就个人而言,也许可以说在过去几个月里与你合作是件非常愉快的事
3	And I congratulate you for the fantastic term card you've put together	恭喜你把名词卡拼起来
4	Ladies and gentlemen for the past 200 years this society has debated the most pressing issues of its time and	女士们,先生们,在过去的200年里,这个社会已经辩论了它时代最紧迫的问题。
5	Tonight's debate on the implications of artificial intelligence for the way we live is doubtlessly one of the most pressing issues of our time	今晚关于人工智能对我们生活方式的影响的辩论无疑是我们时代最紧迫的问题之一
6	Before I move into my main argument I would also like to say some words on the motion and the burden which the proposition bears tonight	在我开始讨论我的主要论点之前,我还想就今晚的提议所肩负的动议和负担说几句话。
7	This house believes artificial intelligence is an existential threat	这房子相信人工智能是一种生存威胁
8	We do not know what the threat is to but from a sort of general understanding I think we need to adopt a very high burden here materially	我们不知道什么是威胁,但从某种一般的理解来看,我认为,我们需要在物质上承担一个非常沉重的负担。
9	And assume that we need to show that artificial intelligence is an existential threat to the way we live and to core values which we hold dear	假设我们需要证明 人工智能是一个存在性的威胁 我们的生活方式 和核心价值观 我们所珍视
10	However we must not show that this will inevitably happen	然而,我们决不能表明,这种情况将不可避免地发生
11	We must only show that a threat thereof exists	我们只能证明存在威胁
12	So don't be fooled when the librarian tells you all the fantastic uses of artificial intelligence, all the prosperity it can bring	所以不要被愚弄 当图书管理员告诉你 所有神奇的人工智能用途, 所有它能带来的繁荣
13	We do not dispute this in the slightest	我们毫不怀疑
14	But what the opposition has failed to show so far is that there are adequate safeguards in place which would mitigate the risk which artificial intelligence poses	但反对派至今未能表明的是,已经有足够的保障措施可以减轻人工智能带来的风险。
15	So my fellow speakers on proposition	所以,我的同声代言人
16	My fellow speakers have already taken quite a drastic picture of the future with artificial intelligence	我的同声代言人已经用人工智能 来深刻描绘未来
17	I shall go one step back and I shall not assume artificial intelligence as this godlike menace	我将退一步,我不会假设人工智能 像神一样的威胁
18	But I will move forward from the artificial intelligence	但我会从人工智能中前进
19	We have generative AI as Chachi BT or other applications which we are already aware of in the foreseeable future	我们有基因AI作为Chachi BT或其他应用,我们已经意识到在可预见的未来。
20	So I will not paint a dystopian terminator-esque image of these Frankensteinian creations which suddenly turn on their masters to eliminate them	所以我不会刻画出这些弗兰肯斯坦的作品的 令人毛骨悚然的终结者形象 突然让主人消灭它们
21	Instead I see artificial intelligence as we have it now as a tool which humans can use to further the ends that they set	相反的,我看到人工智能 就像我们现在拥有它 作为一种工具 人类可以用来推进 他们设定的目标
22	However it is an extremely potent tool in doing so	然而,这是极为有力的工具。
23	rivaled historically maybe by the World Wide Web, the personal computer, the steam engine locomotive and maybe the invention of the wheel	历史上可能与万维网、个人电脑、蒸汽机车 以及轮子的发明相竞争
24	Now it is clear that because of this potency artificial intelligence bears the great risk of disrupting society severely	现在很明显,由于这种力量 人工智能承受着严重扰乱社会的巨大风险
25	A study by Morgan Stanley shows that 300 million jobs worldwide are at risk of being automated in the foreseeable future	Morgan Stanley的一项研究显示,在可预见的将来,全世界有3亿个工作可能自动化。
26	And a McKinsey study shows that 30% of jobs in the United States are at risk of being automated in the next 7 years, so until 2030	麦肯锡的一项研究显示,美国30%的工作岗位在未来7年中面临自动化的风险,因此直到2030年.
27	And it is especially people from ethnic minority backgrounds who are in occupations which are at risk of being automated	特别是来自少数民族背景的人在职业中面临自动化的风险。
28	However it is not only simple jobs at risk of being automated through artificial intelligence	但是,这不仅是简单的工作 面临通过人工智能自动化的风险
29	but also coders or web designers or lawyers like myself	还有我这样的编译师、网络设计师或律师
30	So if you'd been hoping for a bright future doing due diligence for the M&A team at Slaughter in May	所以,如果你一直希望有一个光明的未来 在5月的屠宰场的合并和收购团队 尽职尽责
31	I fear that in the next couple of years you might be out of business	我担心在接下来的几年里 你可能会失去工作
32	Now this leads to an existential threat for those people who will lose their jobs	这会给那些失去工作的人带来生存威胁
33	Existential in the sense that they will not be able to feed their families	他们无法养活家人
34	They will lose their sources of income	他们将失去收入来源
35	And because through the industrial revolution a lot of manual labor moved to intellectual occupations	因为通过工业革命 大量体力劳动转移到了智力职业
36	which are now being automated by artificial intelligence	人工智能正在自动化
37	It is not clear that the people who are losing their jobs due to optimization in artificial intelligence	由于人工智能优化而失去工作的人还不清楚
38	will be qualified to take up new job opportunities created through artificial intelligence	将有资格利用人工智能创造的新就业机会
39	And thus I put it to you artificial intelligence does bear a risk	所以我把它给你 人工智能确实有风险
40	an existential risk of threatening our social contract as we have it in society	威胁我们社会契约的存在风险 就像我们社会上一样
41	and eroding social cohesion	并削弱社会凝聚力
42	Because even if you don't buy that there is also a further risk and that is at a global scale	因为即使你不买, 还有一个进一步的风险, 这是在全球范围
43	Because of course artificial intelligence, because of the speed that it is developing at	当然是人工智能 因为它在发展的速度
44	and the great potential for GP growth it has	拥有巨大的增长潜力
45	It is a sort of dreadnought moment in the geopolitical framework	这是地缘政治框架内一种可怕的时刻
46	There is a great threat of other states suddenly getting enormous growth	其他国家突然出现巨大增长的威胁
47	These could be fantastic states and a great opportunity for them	这可能是一个很棒的状态 和他们伟大的机会
48	but it could also be tyrannies, dictatorships, authoritarian states	但也可能是暴政、独裁、专制国家
49	suddenly gaining enormous power	突然获得了巨大的力量
50	We can see for instance China with extremely strong development in artificial intelligence	例如,我们看到中国的人工智能发展极为强劲
51	and this does also bear a great risk for us on the geopolitical scale of living life in the free world	这对我们在自由世界中生活的地缘政治规模也具有很大风险
52	But apart from these general threats of disturbance of the geopolitical framework	除了这些扰乱地缘政治框架的一般威胁
53	of the domestic situation we live in	家庭状况
54	there are of course also dangerous and mischievous uses	当然还有危险和错误的用途
55	which also pose existential threats to various values which we hold dear	威胁我们珍视的各种价值
56	For instance deep fakes already now can be created	例如,现在可以制造出深刻的假象
57	Images which look genuine, which show people in very compromising situations	真实的画面, 显示人们处于非常不利的处境
58	and these pose an existential threat to people's dignity and reputation	这些对人的尊严和名誉构成生存威胁
59	Face recognition software in the hands of private entities or states	私人实体或国家手中的面部识别软件
60	poses an existential threat to privacy when combined with artificial intelligence	结合人工智能,对隐私构成生存威胁
61	And a further very grave threat	还有一个非常严重的威胁
62	which I believe one of the greatest of the artificial intelligence which we have now	我相信我们最伟大的人工智能之一
63	is fake news created by artificial intelligence	是人工智能制造的假新闻
64	articles which seem very genuine and thought through	似乎非常真实和深思熟虑的条款
65	accompanied by images which seem to back the claims made	附有似乎支持索赔的图像
66	disseminated over the world wide web	在万维网上传播
67	in the service of a foreign state trying to intervene in elections	试图干预选举的外国
68	or hampering the democratic process	或阻碍民主进程
69	And if we do not know what the factual basis is we're basing our decisions on in a democracy	如果我们不知道什么是事实基础 我们在民主中做决定
70	this makes it impossible for us to come together as a society	这使我们无法作为一个社会走到一起
71	and find out what we should do facing a certain set of facts	并找出我们应该如何面对 某些事实
72	and this is an existential threat to democracy and the way we live	这是对民主和我们生活方式的一种生存威胁
73	Now I believe I've shown various ways in which we can see that artificial intelligence	现在,我相信我已经展示了各种方式 我们看到人工智能
74	bears existential risks for the way we live and values we hold dear	给我们的生活方式和珍贵的价值观带来生存风险
75	And this is not only a remote risk	这不仅仅是一个遥远的风险
76	or some sort of ivory tower possibility I've come up with	或某种象牙塔的可能性 我已经想到
77	But the trouble is that there are just no safeguards to prevent this effectively	但问题在于没有保障能有效阻止
78	Of course, as been pointed out, we've had many great technological advancements	当然,正如人们所指出,我们已经取得了许多伟大的技术进步
79	which have changed society	改变了社会
80	However, these could usually be harnessed, the good of these	但是,这些通常可以加以利用,这些好处是:
81	through regulation because the risks could be mitigated	通过监管,因为风险可以减轻
82	by putting it in boundaries which would make them safe for use in society	让它安全地用于社会
83	However, through the extremely fast development of artificial intelligence	然而,通过人工智能的极快发展
84	and at the same time the great potential it bears	同时也有着巨大的潜力
85	both for GP growth but also for just general power projection	无论是GP增长还是一般动力预测
86	it is very difficult to find international consensus on regulating artificial intelligence	很难就人工智能的监管达成国际共识
87	If a state wants to in some way intervene in other states it is not friendly	如果一个国家想以某种方式干预其他国家,那就不是友好的
88	and unfriendly relationships with	和不友好的关系
89	it is frankly absurd to think they will agree to international frameworks	坦率地说,认为他们同意国际框架是荒谬的。
90	which prevent development of such artificial intelligence	防止这种人工智能的发展
91	And furthermore, any regulation will deter development	此外,任何条例都会阻碍发展
92	So they say that, for instance, the United States	所以他们说,例如,美国
93	are the world champions in artificial intelligence	是人工智能世界冠军
94	whilst the European Union is the world champion in artificial intelligence regulation	欧洲联盟是人工智能管制的世界冠军
95	and doesn't put out much	也没什么好说的
96	So if you regulate, you always risk potential in developing artificial intelligence	因此,如果你调节, 你总是冒着发展人工智能的潜力
97	and thus risk falling behind other states, potentially authoritarian states	从而有可能落后于其他国家, 潜在的专制国家
98	like, for instance, China	比如说中国
99	And therefore, if you vote tonight in proposition	因此,如果你今晚投票提案
100	you will not be voting for general scepticism of technology	你将不投票 普遍怀疑技术
101	you will not be voting for some overly pessimistic view of the future	你将不投票 对于一些过于悲观 未来的观点
102	and what artificial intelligence can bring	以及人工智能能带来什么
103	but you will be voting for a very realistic view of the technologies	但你会投票支持 一个非常现实的 技术观点
104	which we have on the table now	现在摆在桌子上
105	and you will see that there is a great challenge we must face	你会看到,有一个巨大的挑战 我们必须面对
106	if we want to harness the power thereof	如果我们想利用它的力量
107	and make sure we avoid the existential risk it has	避免它存在的风险
108	Thank you	谢谢

该视频共有字幕 108 条。解锁更多字幕为会员功能，请移动到 价格

P6 · 【高清完整版｜双语字幕】牛津大学英辩 AI 是否构成人类生存威胁 p06 6 (P6)

【高清完整版｜双语字幕】牛津大学英辩 AI 是否构成人类生存威胁 p06 6 (P6)

发布于 2026-07-22 · 时长 10:26

字幕摘录

时间	英文	中文
0:05	Humanity's fear of technological progress is longstanding.	人类对技术进步的恐惧是长期的。
0:09	If you're living during the Industrial Revolution,	如果你活在工业革命期间
0:11	it's very likely that the automation anxiety	很可能是自动化焦虑
0:14	you would have been feeling	你会觉得
0:15	mimics contemporary headlines that we see.	模仿当代的头条新闻
0:18	However, past inventions,	然而,过去的发明,
0:21	once considered existential threats,	曾经被认为是生存威胁,
0:23	did not materialize as such.	事实并非如此。
0:26	While AI does introduce complex risks,	虽然大赦国际确实带来了复杂的风险,
0:30	labeling it an existential threat is premature	贴上存在威胁的标签还为时过早

展开字幕全文（237 条）

序号	英文	中文
1	Humanity's fear of technological progress is longstanding.	人类对技术进步的恐惧是长期的。
2	If you're living during the Industrial Revolution,	如果你活在工业革命期间
3	it's very likely that the automation anxiety	很可能是自动化焦虑
4	you would have been feeling	你会觉得
5	mimics contemporary headlines that we see.	模仿当代的头条新闻
6	However, past inventions,	然而,过去的发明,
7	once considered existential threats,	曾经被认为是生存威胁,
8	did not materialize as such.	事实并非如此。
9	While AI does introduce complex risks,	虽然大赦国际确实带来了复杂的风险,
10	labeling it an existential threat is premature	贴上存在威胁的标签还为时过早
11	and it lacks a solid foundation.	它缺乏坚实的基础。
12	Experts disagree about how AI will develop,	关于AI的发展方式,专家们意见不一,
13	what the timeline is,	时间是什么,
14	and what the level of intelligence that it can achieve.	以及它能达到的智能水平。
15	However, history underscores our very limited ability	然而,历史突出了我们的能力非常有限
16	to foresee a technology's exact applications.	以预测某项技术的具体应用。
17	Our epistemic horizon is incredibly limited.	我们的感知视野是极其有限的。
18	We cannot predict the consequences of our actions	我们无法预测行动的后果
19	in the medium term,	在中期,
20	let alone the coming decades or centuries.	更别说未来几十年或几个世纪了
21	As Kagan noted,	正如卡根指出的,
22	there will always be a very small chance	机会总是很小的
23	that some unforeseen, disastrous,	一些意外的,灾难性的,
24	or fantastically wonderful thing results from our actions.	或奇妙的事物 由我们的行动。
25	The proposition mustn't only show that,	这个提议不能只表明
26	and I quote, a threat exists.	我引用一句,威胁存在。
27	That is not the motion.	这不是动议。
28	The proposition must show why it is necessarily the case	命题必须说明为什么一定要这样
29	that the disastrous effects are certain,	灾难性的后果是肯定的,
30	branding AI not only a globally catastrophic risk,	不仅给AI带来全球灾难性风险,
31	but an existential threat.	但是一种生存威胁
32	One second.	一秒钟。
33	This extreme stance is excessively pessimistic	这种极端的态度太悲观了
34	and epistemically indulgent,	且隐约地,
35	imposing a substantial burden of proof.	规定了重大的举证责任。
36	The opposition acknowledges the potential	反对派承认,
37	for AI's negative effects,	对于大赦国际的负面影响,
38	but insists that a claim is absolute	但坚持要求是绝对的
39	as an existential threat requires	作为一种生存威胁,
40	far more empirical evidence and epistemic certainty	更多的经验证据和认知确定性
41	than the proposition has provided.	比建议提供的要好
42	To begin in typical PPE fashion,	从典型的个人防护设备开始,
43	let's deconstruct the key terms in tonight's motion.	让我们解析 今晚动议中的关键条款。
44	AI is an existential threat.	AI是一种生存威胁。
45	Existential breaks down into the physical	存在分裂为物理
46	and metaphysical aspects.	和元物理方面。
47	First, I will argue that AI doesn't threaten	首先,我会说,AI没有威胁
48	the physical existence of humanity	人类的实际存在
49	because we possess both the incentive	因为我们有两种奖励
50	and the capacity to implement preventive methods	执行预防方法的能力
51	through design and regulation.	通过设计和监管。
52	Secondly, I show that AI doesn't pose an existential threat	第二,我证明AI不会威胁生存
53	to the experienced and shared understanding	对经验丰富和共同的理解
54	of what defines humanity.	人类的定义。
55	Also note that the motion is in present tense,	还注意到该动议目前紧张,
56	while the arguments presented by the proposition	主张所提出的论点
57	are founded on future developments of AI.	以大赦国际未来的发展为基础。
58	Powerful AGI, an intelligence explosion,	强大的AGI, 一个情报爆炸,
59	or superintelligence all revolve around	或超级情报都围绕着
60	whether AI will become an existential threat,	* 大赦国际是否会成为生存威胁,
61	not that it currently is.	而不是现在
62	These arguments rely very heavily	这些论点非常依赖
63	on hypothetical and abstract scenarios	假设和抽象情景
64	rather than being grounded in empirical forecasts.	而不是基于实证预测。
65	Nevertheless, for AI to be a current existential threat,	然而,为了使大赦国际成为目前存在的威胁,
66	the proposition must show, number one,	提议必须显示,第一,
67	that AI developments will indeed track	AI的发展确实会跟踪
68	these alarming scenarios and that it is necessarily	这些令人震惊的情景,必然是
69	the case that they will and that we are unable	他们愿意而且我们无法
70	to effectively prevent such risks.	有效预防此类风险。
71	I will attack this second proposition.	我会攻击这个第二个提议。
72	Firstly, let's address AI design.	首先,让我们谈谈AI的设计。
73	AI systems don't have motivations or intent.	AI系统没有动机或意图.
74	They operate based on pre-programmed rules	它们根据预先规划的规则运作
75	and objectives that we give them	我们赋予他们的目标
76	and the data inputs that they learn from.	以及他们学习的数据输入。
77	These are elements that human developers	这些元素是人类开发者
78	can control and constrain.	能够控制和约束。
79	For example, misinformation, such a terrible issue.	例如,错误信息,这样一个可怕的问题。
80	If we aim to ensure accurate and reliable responses	如果我们要确保准确和可靠的反应
81	from an AI system, we can limit its data	从AI系统,我们可以限制它的数据
82	to highly reliable ones, peer-reviewed books,	对高度可靠的人, 同行评审的书籍,
83	academic journals.	学术期刊.
84	Developers can design boundaries and safety features	开发者可以设计边界和安全特性
85	to prevent AI from being programmed to cause harm	防止AI被编程造成伤害
86	or to pursue destructive methods	或采用破坏性方法
87	for achieving a beneficial goal.	实现一个有益的目标。
88	Aligning AI with fundamental human objectives	使大赦国际与基本人类目标相一致
89	such as equity, fairness, anti-discrimination	平等、公平、反歧视
90	and sustainability is crucial to ensure	可持续性对于确保
91	that we're getting what we actually want,	我们得到了我们真正想要的,
92	not just what we ask for.	不只是我们要求的
93	And active alignment research is already underway	正在积极进行校正研究
94	at the major companies.	在主要公司。
95	Sultan pointed out that the probability	苏丹指出,可能性
96	of aligning AI 100% and removing all risk is very small.	百分之百对齐AI并消除所有风险是很小的。
97	But it's not the case that we expect zero risk	但是我们并不期望风险为零
98	with other technologies that we use.	我们使用的其他技术
99	Just an example, nuclear energy.	核能就是一个例子。
100	Following Nozick's principle of side constraints,	遵循诺兹克的侧面约束原则,
101	we can impose limits within which AI can perform tasks.	我们可以施加限制 AI可以完成任务。
102	A self-driving vehicle programmed to follow traffic rules	为遵守交通规则而规划的自驾车
103	and only drive on public roads	只开公共道路
104	is limited in how it gets you to work quickly.	很难让你快速工作
105	Humans not only control the design of AI,	人类不仅控制AI的设计,
106	we also provide AI with the means it can use.	我们还向大赦国际提供其可以利用的手段。
107	A self-driving car only functions if we choose to fuel it.	自动驾驶的汽车只有在我们选择加油时才能发挥作用.
108	Therefore, humans exert significant control over AI	因此,人类对AI有相当的控制权.
109	by carefully considering the means that we offer it	仔细考虑我们提供的手段
110	and the system that we choose to introduce it into.	以及我们选择引入的系统
111	Taking this back to first principles,	回到第一原则
112	AI lacks consciousness and understanding.	大赦国际缺乏意识和理解。
113	It is therefore a tool that is reliant	因此,这是一个依赖的工具。
114	on some degree of human direction.	在人类的某种方向上
115	If there is an existential risk with AI,	如果AI存在存在风险,
116	it's not the systems themselves,	这不是系统本身,
117	but the humans behind them who pose the threat.	但他们背后的人 威胁。
118	By regulating and mitigating destructive human actions,	通过管制和减轻人类的破坏性行动,
119	AI is not an existential threat.	大赦国际不是生存威胁。
120	Moving on, there is both an incentive	继续说,两个都有动机
121	and a capacity for AI regulation.	并具备对大赦国际进行管理的能力。
122	Politicians, lawmakers, and companies worldwide	世界各地的政治家、立法者和公司
123	have the incentive to invest in	有投资的动机
124	and collaborate on regulatory systems.	并就管理制度进行合作。
125	Why is that the case?	为什么?
126	Because of the reach and extent of AI	由于大赦国际的范围和范围
127	to potentially fundamentally reshape our institutions,	有可能从根本上改变我们的机构,
128	our public systems, our industries, and our daily lives.	我们的公共系统,我们的产业, 我们的日常生活。
129	Growing public awareness	提高公众认识
130	and growing expert discourse on the topic	和越来越多的专家讨论
131	increases the political salience of AI	提高大赦国际的政治地位;
132	among electorally significant groups,	在具有重要选举意义的团体中,
133	incentivizing a regulatory response.	激励监管响应.
134	Substantial attention to AI-related concerns	大力关注大赦国际的关切问题
135	should in fact alleviate our apprehensions.	事实上,这应该减轻我们的忧虑。
136	Secondly, there is also a capacity for regulation.	第二,还有监管能力。
137	It is exclusively up to people to establish the rules	完全由人民来制定规则
138	for how we wish to use AI	如何使用人工智能
139	and how to let it interact with humanity.	以及如何让它与人类互动
140	Robust government oversight, transparency requirements,	强有力的政府监督、透明度要求、
141	liability for AI developers,	对AI开发者的责任,
142	and interdisciplinary collaboration are required.	需要开展跨学科合作。
143	Similar to international agreements	与国际协定类似
144	on nuclear nonproliferation and bans on chemical weapons,	关于核不扩散和禁止化学武器,
145	we can establish conventions to prohibit	我们可以订立公约禁止
146	and heavily regulate autonomous weapon systems.	并严格管制自主武器系统。
147	Currently, very few players possess	目前,很少有玩家拥有
148	the cutting edge computing resources,	尖端计算资源,
149	necessary chips and hardware,	必要的芯片和硬件,
150	and financial capital to develop and train powerful AI.	金融资本 开发和训练强大的AI。
151	A significant advantage to regulating only a few players	管理少数角色的重大优势
152	is that it's more feasible.	这是更可行的。
153	A smaller number of actors facilitates their identification,	为数较少的行为者为识别他们提供便利,
154	monitoring, and the alignment of their interests.	监测,协调它们的利益。
155	The model charted by these first movers	这些先动器绘制的模型
156	in military and civilian AI regulation	军事和民事法
157	determines the incentives that subsequent countries	确定随后国家采取的激励措施
158	and companies will face in developing AI.	公司在开发AI时将面临困难。
159	We have the means to enforce preventive measures	我们有办法执行预防措施
160	and avoid AI becoming an existential threat.	并避免AI成为生存威胁.
161	In the final section of my speech,	在我发言的最后部分,
162	I explore why AI is not an existential threat	我探讨了为什么AI不是生存威胁
163	to the philosophical underpinnings of humanity.	人类的哲学基础。
164	I want you to take a second	我想让你休息一下
165	to think about what makes you human.	想想是什么让你变成人类
166	AI doesn't threaten our consciousness,	AI不威胁我们的意识,
167	our intentions and will, our moral propensity,	我们的意图和意志,我们的道德倾向,
168	or our creativity.	或者我们的创造力
169	AI cannot compete with these distinctively human features	AI不能与这些独特的人类特征竞争
170	no matter how intelligent it becomes.	不管它变得多么聪明
171	The experience and shared understanding	经验和共同理解
172	of what we consider to be humanity	我们所认为的人类
173	is not under existential threat.	并不受生存威胁。
174	Let's delve deeper into creativity,	让我们深入探索创造力,
175	as it was mentioned, AI art, right before.	如前所述 AI艺术
176	What is creativity?	创造力是什么?
177	Is it generating something novel	是它产生了一些新东西
178	by combining existing patterns of information?	综合现有信息模式?
179	While AI does satisfy this definition,	虽然大赦国际确实符合这一定义,
180	AI-generated art lacks expression and artistic value.	AI生成的艺术缺乏表达和艺术价值.
181	Creativity encompasses more than this rudimentary definition.	创造性不仅包括这一基本定义。
182	People or processes are creative to the extent of	人或过程具有创造性,可达到以下程度:
183	and because they produce creative products.	因为他们生产创意产品。
184	Products are creative on account of three conditions.	产品基于三个条件具有创造性.
185	They are novel, they are valuable,	他们是新颖的,是有价值的,
186	and they are created with agency.	他们与众生同受崇拜。
187	Generative AI products are not creative	基因AI产品没有创意
188	in the same way that a snowflake is not.	和雪花不一样
189	You know, it's new, it's unique,	你知道,这是新的,这是独特的,
190	it might even have aesthetic value,	甚至有美学价值,
191	but it lacks agency in its creation.	但是,它缺乏 代理它的创建。
192	Art goes beyond aesthetic value.	艺术超越了美学价值.
193	It involves artistic intent, authenticity,	它涉及艺术意图 真实性
194	the artistic process, and the expression	艺术过程和表达方式
195	of human emotions and experiences.	人类的情感和经历
196	AI doesn't analyze art and then create its own.	AI不分析艺术,然后创建自己的艺术.
197	It merely identifies patterns and replicates existing styles	它只是识别模式 复制现有的风格
198	without actually conceptualizing	没有实际概念化
199	or contextualizing its creations as art.	或将其创造作为艺术的背景。
200	This means that the AI arts, you know,	这意味着人工智能艺术,你知道,
201	it might sell for a lot of money,	它可能卖很多钱,
202	but it lacks artistic responsibility.	但它缺乏艺术责任。
203	It cannot form genuine artistic expression.	它不能形成真正的艺术表现。
204	You know, the creative process is fundamental to art.	创作过程对艺术至关重要
205	It's not just about whether we're using	不只是我们是否在使用
206	semi-automated processes or even technology as a medium,	半自动化工艺,甚至技术作为一种媒介,
207	but it's about the absence of an artistic reaction	但这是关于没有 艺术反应
208	or even an intent in the process,	或甚至在过程中的意图,
209	and that's what makes art valuable.	这就是艺术的价值所在
210	You know, we value art that responds to something	你知道,我们珍视艺术 回应的东西
211	because it connects us to the lived experiences	因为它连接我们 与生活的经验
212	of other people.	别人。
213	So AI doesn't threaten creativity,	所以AI不会威胁创造力
214	a distinctively human trait,	一个独特的人类特质,
215	nor the true value of human creative output.	也不是人类创造产出的真正价值
216	I'm at my conclusion.	我到了我的结论。
217	I don't think I have time, sorry.	我想我没有时间,对不起。
218	So to conclude, AI does not pose	因此,最后,大赦国际没有提出
219	an existential threat to humanity,	对人类的生存威胁,
220	be it in the literal or the metaphysical sense.	无论是从字面上还是从物理上来说
221	Human direction in AI development is paramount,	人工智能发展的人文方向至关重要,
222	and we possess both the incentive and the capacity	我们既具有激励力,又有能力
223	to implement preventive measures for AI risk	对AI风险采取预防措施
224	through design and regulation.	通过设计和监管。
225	Furthermore, distinctively human characteristics	此外,独特的人类特征
226	will continue to define our humanity	将继续界定我们的人性
227	alongside AI's advancement.	与AI的进步。
228	All powerful tools carry risks.	所有强大的工具都带有风险。
229	The goal is not to get to no risk,	目标不是去冒险,
230	but it's unjustified to extrapolate these risks	但是推断出这些风险是不合理的
231	to an extreme.	到一个极端。
232	There is a difference between a globally catastrophic risk	全球灾难性风险有区别
233	and an existential threat,	和生存的威胁,
234	and the proposition has failed to demonstrate	而这个提议未能证明
235	the necessity of the former,	前者的必要性,
236	let alone the latter.	更别提后者了
237	Thank you.	谢谢

该视频共有字幕 237 条。解锁更多字幕为会员功能，请移动到 价格

P7 · 【高清完整版｜双语字幕】牛津大学英辩 AI 是否构成人类生存威胁 p07 7 (P7)

【高清完整版｜双语字幕】牛津大学英辩 AI 是否构成人类生存威胁 p07 7 (P7)

发布于 2026-07-22 · 时长 10:55

字幕摘录

时间	英文	中文
0:03	It's such an honor and pleasure to be here in England's greatest University	来到英国最伟大的大学,真是荣幸和荣幸
0:15	I've counted many different definitions of	我数过很多不同的定义
0:18	Existential threats tonight if we were in some very sloppy thinking place like Cambridge	如果我们在剑桥这样的草率的思考场所 今晚就会产生威胁
0:23	I would just leave it at that but because we're in Oxford	我本来想把它留在那,但因为我们在牛津
0:25	I'm gonna define what I mean by extensions threat first of all the term	我要先定义我所说的延期威胁
0:30	Existential risks was defined by an Oxford professor and I'm gonna go with his definition, which is a threat of	牛津大学的一位教授定义了存在的风险 我要遵循他的定义
0:37	Human extinction or permanent	人类灭绝或永久灭绝
0:41	Permanent	常设
0:42	Limiting of human potential and they'll become important in a little bit.	限制人的潜力,它们会在一点点内变得重要.
0:46	What about threat?	威胁呢?

展开字幕全文（166 条）

序号	英文	中文
1	It's such an honor and pleasure to be here in England's greatest University	来到英国最伟大的大学,真是荣幸和荣幸
2	I've counted many different definitions of	我数过很多不同的定义
3	Existential threats tonight if we were in some very sloppy thinking place like Cambridge	如果我们在剑桥这样的草率的思考场所 今晚就会产生威胁
4	I would just leave it at that but because we're in Oxford	我本来想把它留在那,但因为我们在牛津
5	I'm gonna define what I mean by extensions threat first of all the term	我要先定义我所说的延期威胁
6	Existential risks was defined by an Oxford professor and I'm gonna go with his definition, which is a threat of	牛津大学的一位教授定义了存在的风险 我要遵循他的定义
7	Human extinction or permanent	人类灭绝或永久灭绝
8	Permanent	常设
9	Limiting of human potential and they'll become important in a little bit.	限制人的潜力,它们会在一点点内变得重要.
10	What about threat?	威胁呢?
11	What what is the threat exactly?	到底是什么威胁?
12	Well, for example Silva you define the threat is something which is a hundred percent	比如席尔瓦 你给威胁下定义 是百分之百的
13	Certain to happen and then very successfully attacked that	肯定发生,然后非常成功地攻击
14	But I looked it up and the not in the Cambridge dictionary in the Oxford dictionary and it said that a threat is a possibility	但我查了一下 牛津词典里没有的剑桥词典 上面说威胁是可能的
15	Of something bad happening.	坏事发生。
16	What does the possibility mean?	可能性意味着什么?
17	It means the probability of the bad thing happening not doesn't mean it's a hundred percent	意思是坏事发生的可能性 不意味着百分之百
18	It means it's more than zero and we had a good question from up there somewhere.	意思是比零还多 我们从上面问了个好问题
19	Where do you draw the line?	你在哪里划线?
20	Well, if you're talking about something like extinction suppose there is a an asteroid which will kill us all if it hits us	好吧,如果你在谈论什么 像灭绝 假设有一个小行星 会杀死我们所有人,如果它击中我们
21	What is the probability threshold where you feel you really need to take action?	你真正需要采取行动的概率阈值是什么?
22	You have to look into your hearts and answer that is it one in ten thousand is it one in a thousand?	汝须察其心,答此即是万分之一是千分之一?
23	One in one in a hundred regardless of what it is	百分之一,无论是什么
24	That's what you should think of when you choose which door to walk out of if you pick one percent	这就是你应该想到的 当你选择哪一扇门走出去 如果你选择1%
25	If you think the risk of an existential threat is more than one percent you go through	如果你认为存在威胁的风险超过1%
26	Then you're a yay now	那你现在成了一个耶
27	So that's that now we got the definitions out of the way	所以,现在,我们得到了 定义的方式
28	except	除外
29	There were two really interesting there was another interesting twist also two of the	有两个真正有趣的 还有一个有趣的曲折 还有两个
30	Opposition speakers said it's only a threat if it's happening right now	反对派发言人说,如果现在发生,那只是威胁
31	And they said it's AI is not a threat because we don't have a GI yet superintelligence	他们说这是AI 不是威胁,因为我们没有GI 和超级情报
32	So suppose this asteroid that's heading towards earth again isn't gonna hit us for 20 years	所以假设这颗飞向地球的小行星 20年不会撞到我们
33	Does it make any sense to say it's not a threat?	说不是威胁有道理吗?
34	Even if the best Oxford astronomers have calculated that has a 99% chance of hitting us.	即使最优秀的牛津天文学家 已经计算出了99%的打击机会
35	That's ridiculous and ironically	这是荒谬和讽刺的
36	Sebastian when you said that the you only count risks	塞巴斯蒂安 你说你只算风险
37	They're happening things are happening now not in the future still turn around and said the catastrophic climate change is a threat even though	说灾难性的气候变化是威胁 尽管我们没有看到
38	It's happening in the future.	于未来世.
39	So now yeah, ask me in four minutes	现在,是的,问我 在四分钟
40	so	这样
41	Okay, so why should we worry about this stuff then?	好吧,那我们为什么要担心这些东西呢?
42	Raise your hand if you study history	学历史的举手
43	You know then that that we can only really understand the present if we study history.	你知道,只有研究历史,我们才能真正理解现在。
44	I think it's very easy to	我觉得很容易
45	Get so blinded by all the recent stuff that's happening with AI that we lose the context	被AI最近发生的事情蒙上眼睛 我们失去了背景
46	So if you are lucky enough to have 50 quid in your pocket, can you take it out, please for a moment?	所以,如果你够幸运 在你的口袋里有50英镑, 你能把它拿出来, 请一分钟?
47	Let's see how well	让我们看看有多好
48	Oxford is charging too high fees.	牛津收费太高.
49	I think	我觉得
50	So this gentleman here Alan Turing has already been mentioned	这位先生艾伦图灵已经提到
51	Really the Einstein of computer science amazing man	真正的爱因斯坦 计算机科学惊人的人
52	Not only viewed as the founding father of computer science and of artificial intelligence also really an amazing	不仅是计算机科学的创始人 和人工智能的创始人
53	Amazing war hero raise your hand if you saw imitation game the movie.	惊人的战争英雄举手 如果你看到模仿游戏的电影。
54	Yeah, well then I'm preaching to the choir here help	是啊,那么我传教 到合唱团在这里帮助
55	Crack the Germans code by building the first ever programmable digital computer	通过建造第一台可编程的数字计算机来破解德国密码
56	pretty smart, dude	挺聪明的,伙计
57	so	这样
58	He said back in 1951 as John already mentioned that	他在1951年说过 约翰已经提到过
59	狦ит诀竟  outи
60	they will pretty quickly leave our feeble intellects far behind and	他们很快就会把我们弱智抛在后面
61	We should expect he said that at some point	我们期待他说,在某个时候
62	They will take control the machines will take control	他们会控制机器 控制机器
63	That's not particularly rocket science when we showed up in the rainforest	当我们在雨林里出现时,这不算什么火箭科学
64	Tough luck for all the animals that live there	住在那里的动物真倒霉
65	We took control not because we had bigger muscles, but because we're smarter and it didn't always work out for them	我们控制局面不是因为我们有更大的肌肉 而是因为我们更聪明 也并非总能解决
66	but he had this is his argument and	但他有 这是他的论点,
67	We have to pay respect to his intellect.	我们必须尊重他的智慧
68	We can't just this him without giving good counter arguments	我们不能不给他个好辩驳
69	I know it sounds like science fiction machines taking control, but	我知道这听起来像 科幻机器控制,但
70	You know nuclear weapons sounded like science fiction 1930 and when Alan Turing formulated the Turing tests	核武器听起来像1930年的科幻小说 当艾伦图灵提出图灵试验
71	Which he heralded is the final milestone before we should really start freaking out	他预言的是我们真正开始惊慌之前的最后里程碑
72	Having something like chat GPT-4 sounded like science fiction also.	聊天GPT -4 听起来也像科幻小说
73	So saying that something sounds like sci-fi	所以说这听起来像科幻
74	Does not imply it won't happen.	这并不意味着不会发生。
75	That's just a really really lame argument and it's not just Turing	这只是个很烂的争论 不只是图灵
76	Sam Altman	萨姆·阿尔特曼
77	公开AI的CEO,给我们聊天GPT
78	Has said himself recently that it could be	最近他自己说,这可能是
79	Lights out for all of us.	熄灯给我们所有人。
80	That's an existential threat, isn't it?	这是一个生存威胁,不是吗?
81	And in May of this of this and in fact Daria Amade CEO of Anthropics	而在5月,这实际上 达里亚・阿玛德CEO 人类学
82	One of the AGI competitors even put a number on this risk	一个AGI的竞争对手甚至给这个风险投了数
83	He said he thinks we have two or three years left until we get out smarted	他说,他觉得我们还有两三年的时间,直到我们聪明地出来
84	in three minutes, I'll take it and	三分钟后,我会把它和
85	It's not just them.	不只是他们
86	It's not just them in May of this year	今年五月不只是他们
87	我们有所有AGI公司的首席执行官 Google DeepMind在伦敦
88	Open AI anthropic and a who's who of leading AI researchers including at Oxford	打开AI人类学 和谁是AI的领头研究者 包括牛津大学
89	Say this is an existential threat human extinction is something we have to take seriously	说这是生存的威胁 人类的灭绝是我们必须认真对待的
90	so	这样
91	I said two minutes.	我说两分钟
92	Yeah	对
93	You seem to have four clocks here.	你好像有四个钟
94	Are you from Cambridge?	你是剑桥来的吗?
95	So so	所以这样
96	This sounds kind of dark	这听起来有点黑
97	So let's have let's raise the spirits a little bit by having some fun and playing yawn tall in bingo	让我们通过玩乐和在宾果打哈欠来提高精神
98	So he told us that we should keep a bingo card of the opposition argument	所以他告诉我们 我们应该保留一张反对派辩论的宾果牌
99	So that's actually what I've been doing here.	这就是我一直在这里做的事。
100	He said there are four kind of bad arguments	他说有四种坏话
101	Tent to be made arguing that it's not a threat.	以非威胁为理由
102	One of them is	其中一个是
103	labeling	标签
104	Ahomonym攻击人X说Y
105	also, which therefore X is wrong and and	因此,X是错误的,而且
106	Keisha I agree with you on many things but you got that little bingo score right there because you said	凯莎 我同意你的观点 但你的宾果得分就在这里 因为你说
107	Nick Bostrom	尼克·博斯特罗姆
108	The warrant of existential it's an existential threat and he said these really bad things	生存令是一种生存威胁 他说这些很坏的事情
109	Therefore, it's not an existential threat	因此,这不是生存的威胁
110	So yeah got you there and the framing we have many of the winners here.	所以,是的,你在那里 和框架 我们有很多赢家在这里。
111	So Sebastian you said oh	塞巴斯蒂安 你说
112	Is not evil therefore, it's not an existential threat	所以不是邪恶,不是生存的威胁
113	Technology isn't evil and it's also not morally good	技术不是邪恶的 也不是道德上的好
114	it's a tool and we can use it for good things and bad things Turing was well aware of this and	它是一个工具 我们可以用它做好事和坏事 图灵很清楚这一点
115	The that does not change the fact that it's an existential threat	这并不能改变它是一种生存威胁的事实
116	Sebastian also said that it's yeah.	塞巴斯蒂安还说是的
117	Actually, let me pick on Silva for some democracy here	其实,让我选席尔瓦来做个民主的人
118	Actually, let's move on to picked on you once already.	其实,让我们开始 选择你一次了。
119	I want to spread my graces you meant human supremacy	我想传播我的恩典 你是指人类至高无上
120	We had the argument that AI cannot compete with us	我们争论说AI不能和我们竞争
121	No matter how advanced it gets	不管它有多先进
122	because it	因为
123	Doesn't have quite our human culture and the art isn't as profound.	人类文化并不丰富 艺术没有那么深厚
124	This was discussed here, right?	这是在这里讨论的,对不对?
125	Well, you know	你知道吗?
126	Sorry to say	对不起
127	Africa had some amazing art when British and other colonialists came there with big guns and	非洲有一些惊人的艺术 当英国和其他殖民者 带着大枪来到那里,
128	That did not protect	这不能保护
129	These people in Africa at all.	这些人在非洲。
130	Something doesn't have to be more culture than us to pose a threat	一定有比我们更文化的东西来威胁我们
131	In fact, if you think about Alan Turing's warning why he was warning if you think of AI is just another technology like the steam	事实上,如果你想到艾伦·图灵的警告 他为什么警告 如果你认为AI只是另一种技术 就像蒸汽
132	engine or electricity	发动机或电力
133	You won't be very worried	你不会很担心的
134	But you see Alan Turing	但你看到艾伦图灵
135	Was taking the long view he was thinking of it more as a new species	他的眼光太长了 他更把它当成一个新物种
136	We are as a matter of fact right now building creepy	事实上,我们现在正在制造恐怖
137	super capable	超强能力
138	amoral psychopaths that never sleep	永远睡不着的道德变态
139	Think much faster than us can make copies of themselves and have nothing human about them whatsoever	想象得比我们快很多 复制自己,没有任何人类对他们
140	What could possibly go wrong?	怎么会出问题?
141	once we	一旦我们
142	Have out there in the world entities that we cannot control	世界上有我们无法控制的实体
143	I'll talk about how to control them tomorrow if you come back at 5 p.m.	如果你下午5点回来 我明天再谈如何控制他们
144	But if if we can't control them and	但是如果我们不能控制他们
145	They're much better at us at thinking	他们比我们更擅长思考
146	speaking	说话
147	Manipulating people getting things done	操纵人 做事情
148	The future belongs to them	未来属于他们
149	Not to us	不是对我们
150	This is what Turing realized and then the 72 years since nobody has convincingly refuted that argument	这是图灵所意识到的,然后是72年 没有人有说服力地反驳这个论点
151	it's really as simple as that we stand to lose control and	它真的很简单,因为我们会失去控制
152	It's closed an	关门了
153	Existential threat according to the Oxford definition.	根据牛津的定义存在威胁.
154	It's not just something that drives us extinct	这不仅仅是让我们灭绝的东西
155	It could also mean something that permanently limits our potential.	这也可能意味着永远限制我们的潜力。
156	So if we lose control to machines	所以如果我们失去对机器的控制
157	Either because directly to the machines or to some horrible people through too much power	要么是直接到机器上 要么是因为太强的威力对一些可怕的人
158	Concentration that managed to control us forever and some sort of dystopian 1984 plus plus, you know	集中了我们 永远控制我们 和某种堤坝 1984年加,你知道
159	That is an existential threat by this very definition	根据这个定义,这是一种生存威胁
160	So let's not do this and I want to just end with a note of hope yawn you gave them all a terminal	所以,我们不要这样做,我想结束 仅以一个希望的字条,打哈欠 你给了他们所有的终端
161	Diagnosis, that's a little too harsh.	诊断 有点太苛刻了
162	I think there is still	我觉得还有
163	This might still be treatable	这还可能治疗
164	So what I think you're really voting on if you vote yes for existential threat is vote yes for taking this really seriously and	所以,我认为你真的投票 如果你投票支持生存威胁 是投票支持 认真对待这一点,
165	Being really motivated to go out there and do something about it.	真的很有动机去外面做点什么
166	Thank you	谢谢

该视频共有字幕 167 条。解锁更多字幕为会员功能，请移动到 价格

P8 · 【高清完整版｜双语字幕】牛津大学英辩 AI 是否构成人类生存威胁 p08 8 (P8)

【高清完整版｜双语字幕】牛津大学英辩 AI 是否构成人类生存威胁 p08 8 (P8)

发布于 2026-07-22 · 时长 12:46

字幕摘录

时间	英文	中文
0:03	Well, first of all, it's a great honor for me to be here,	首先 我很荣幸能来这里
0:06	representing a somewhat underrepresented community,	代表代表人数不足的社区,
0:11	which is the scientist and also the engineers	是科学家和工程师
0:14	who builds those super AI machines.	制造超级人工智能的机器
0:17	In fact, in my backyard,	事实上,在我的后院里,
0:18	I have two large-language models being trained,	我有两个大语言模型正在训练中,
0:21	and also we are training the models	而且我们正在训练模型
0:22	for biology and so forth.	生物学等。
0:25	I've been carefully hearing,	我一直在仔细地听到,
0:27	listening to the arguments from both sides.	听听双方的争论

展开字幕全文（241 条）

序号	英文	中文
1	Well, first of all, it's a great honor for me to be here,	首先 我很荣幸能来这里
2	representing a somewhat underrepresented community,	代表代表人数不足的社区,
3	which is the scientist and also the engineers	是科学家和工程师
4	who builds those super AI machines.	制造超级人工智能的机器
5	In fact, in my backyard,	事实上,在我的后院里,
6	I have two large-language models being trained,	我有两个大语言模型正在训练中,
7	and also we are training the models	而且我们正在训练模型
8	for biology and so forth.	生物学等。
9	I've been carefully hearing,	我一直在仔细地听到,
10	listening to the arguments from both sides.	听听双方的争论
11	I hear a lot of emotions, a lot of passions, anger,	我听到很多的情绪, 很多激情,愤怒,
12	and fear, and many things.	恐惧,以及许多事物。
13	But to be honest, I didn't hear too many facts.	但说实话,我没有听到太多的事实。
14	And also, I heard a lot of jumping reasoning,	还有,我听到很多跳跃推理
15	from A to B, without stepping stones in between.	从A到B, 不踩石头之间。
16	I'll give you a few examples, okay?	我举几个例子,好吗?
17	One example is that, oh, ChachiPT3 is so powerful,	一个例子是,哦,ChachiPT3是如此强大,
18	it generates articles and so forth.	它产生文章等等。
19	It does planning in some way, two or three steps correctly.	它确实以某种方式规划了两三个步骤。
20	And then, the next is that,	然后,下一个是,
21	what about it start to plan infinitely well,	怎么样,它开始 计划无限好,
22	and also start to plan against us?	还开始计划对付我们?
23	Well, technically, there is a feasibility issue	技术上来说,有一个可行性问题
24	to be answered.	待答复。
25	And also, it's like, hey, I'm riding a bicycle,	而且,它就像,嘿,我骑自行车,
26	or I'm driving a car.	或者我开车
27	I can drive faster and faster and faster	我开得越来越快
28	by specking a gas, and therefore, the car will fly,	通过偷看气体, 因此,汽车会飞,
29	and then will take me to the moon,	然后带我去月球,
30	and to the Mars, and so forth, right?	和火星,等等,对不对?
31	There is a ignorance on the paradigm shift	对范式转变有无知
32	sometimes needed to even make continuous progress.	有时甚至需要不断取得进展。
33	And AI is no exception of that, okay?	AI也不例外,好吗?
34	AI can take us to a certain point,	AI可以带我们到某个地方,
35	but not to infinity, like a god-like GPT.	但不是无穷无尽 就像上帝般的GPT
36	In fact, that's the first time I heard	事实上,这是我第一次听到
37	that there is a god-like machine	有一个像神一样的机器
38	through the current path of AI developments.	(a) 通过当前AI发展的道路。
39	Give me a second, because I'm short of time.	给我一秒钟, 因为我没有时间。
40	I will give you some time to answer questions after that.	之后我会给你一些时间回答问题。
41	One of the issues that I want to point out is that,	我想指出的问题之一是:
42	what we have been talking about so far,	我们一直在谈论的,
43	most likely is an abstract risk,	最有可能是一种抽象的风险,
44	an abstract damage and harm.	一种抽象的损害和伤害。
45	In spite of the fact that there are many technologies	尽管有许多技术
46	which have been producing demonstrated harms.	已经表现出了伤害。
47	I give you a few examples, right?	我给你一些例子,对不对?
48	We all see the COVID, or maybe other form of virus,	我们都看到COVID, 或也许其他形式的病毒,
49	either engineered or leaked from captured animals	或设计或从捕获的动物中泄漏
50	into the world, that's a real risk.	进入世界,这是一个真正的风险。
51	Do we say biology, therefore,	我们是不是说生物学,因此,
52	pose an excessive risk to mankind?	给人类带来过度风险?
53	Likewise, I can give you examples	我也可以举个例子
54	about chemical weapons.	关于化学武器。
55	Even 100 years ago, it produces such a human loss	一百年前,它就造成了如此巨大的人命损失
56	in the order of tens of thousands.	以数万为序.
57	Nuclear bomb not to speak, right?	核弹不说话,对不对?
58	So AI has been given the honor to be singled out	因此,AI被授予 荣誉被单独列出
59	as the only branded excessive risk.	成为唯一标榜的过度风险。
60	And I need to ask why, because it is actually not yet	我需要问为什么, 因为实际上还没有
61	based on a demonstrated harm of that magnitude.	基于这种程度的明显伤害。
62	And I scratched my head to find the reason	我抓了头来找出原因
63	why this is receiving such a disproportionate attention,	为什么这受到如此不相称的关注,
64	given the relatively modest and very often time	由于时间比较小,而且常常是
65	controllable risk that we experience so far.	我们迄今所经历的可控风险。
66	I want to emphasize controllable.	我想强调可控性。
67	Remember, biology, chemistry, and the physics,	记住,生物,化学,和物理,
68	and many science, they actually happen in the physical world.	以及许多科学, 它们实际上发生在物理世界。
69	Their experiments happens in the lab or in the field,	他们的实验发生在实验室或现场
70	and they actually do impose a mature risk	他们确实带来了成熟的风险
71	when something goes wrong.	当事情出了问题。
72	At least all the AI productivity and also the inventions	至少所有人工智能的生产力和发明
73	is a digital device.	是一个数字设备。
74	It is sitting in the computer.	它坐在电脑里
75	If it comes to kill you, it needs someone to shoot a gun	要杀你,就要有人开枪
76	because the AI itself does not have the physical vehicle	因为人工智能本身没有物理载体
77	to even execute that harm.	甚至执行伤害。
78	Therefore, something else is needed.	因此,还需要其他东西。
79	Therefore, you can say that AI could be a conspiring	因此,你可以说AI可能是阴谋
80	kind of a partner or a tool, but there is another part	某种伙伴或工具,但还有另一部分
81	in this whole game that we didn't even bring it out.	在这场比赛中,我们甚至没有把它拿出来。
82	And in most of these cases, it's the human being.	而在大多数情况下,都是人.
83	I heard a brilliant statement just now	刚才我听到一个精彩的发言
84	that nobody threat the existence of a human being	任何人不得威胁人的生存
85	than human being itself.	而不是人类本身
86	I totally agree that because even the creation of the AI	我完全同意 因为连人工智能的创建
87	for good and bad is actually by and large in the mankind.	不论好坏,在人类中都是如此
88	And also don't forget that comparing to many of the tools	也不要忘记,比起很多工具
89	that we're using now, mechanical tools,	我们用的是机械工具
90	biological tools, and other things,	生物工具和其他东西,
91	AI products and the functions has probably the most amount	AI产品和功能可能数量最多
92	of checking points that you actually can use, okay,	检查点,你实际上可以使用,好吗,
93	including shut down the machines and computers.	包括关闭机器和电脑。
94	But there are many other ways to patch the code	但还有许多其他方法可以补上密码
95	to change the data, to rewrite the loss function.	更改数据,重写丢失函数。
96	There are many, many technical and also policy	有很多,很多技术,还有政策
97	and also regulatory tools to contain AI	以及包含AI的监管工具
98	better than many other existing products.	比其他许多现有产品更好。
99	And last, I also want to point out one thing	最后,我还想指出一件事
100	that is often ignored.	这常常被忽视。
101	What is AI?	什么是人工智能?
102	In fact, the AI I hear today is very, very different	事实上,我今天听到的人工智能非常,非常不同
103	from the AI I work on in my occupation every day, okay?	从AI我工作 每天在我的职业,好吗?
104	What I hear here is that AI becomes,	我在这里听到的是,AI变成,
105	the God becomes everything, okay?	上帝变成一切,好吗?
106	It is on the one hand a science, and then a technology,	一方面是科学,然后是技术
107	and then a product, and then maybe the people	然后是产品,然后也许人们
108	who produce that.	谁制造的。
109	You cannot basically blanketly label so many things	你基本上不能把这么多东西贴在标签上
110	with just one word and call it a risk	只要一个字,就叫冒险
111	because that is leading to things	因为这会导致一些事情
112	that is completely actionable or inactionable.	完全可以行动或不能行动。
113	So this is actually the flaw of the argument that I hear	这就是我听到的论点的缺陷
114	that I don't even see a tangible and a legitimate subject	我甚至没有看到一个有形和合法的主体
115	that I can act against.	我可以采取行动反对。
116	And then the risk definition, I also hear	我听到风险的定义
117	the shifting characterizations of that,	其特征不断变化,
118	starting from something as big as a God,	从神一样大的东西开始
119	and then down to something that is becoming,	然后到一个正在变成的东西,
120	okay, suppressing human creativity,	好吧,压制人类的创造力,
121	also can be branded as excessive risk.	也可以标注为过度风险。
122	Or then it goes down as small as causing someone	或者它会像制造人一样小
123	to lose jobs and so forth.	丢掉工作等等
124	So this is not the way we characterize	所以这不是我们的特点
125	a fundamental excessive risk.	根本的过度风险。
126	I think excessive risk should probably have	我认为过度的风险 应该有
127	the following few characters, like we described,	下面的几个角色,就像我们描述的那样,
128	for example, a natural disaster,	例如,一场自然灾害,
129	say an asteroid hitting the earth,	说有颗小行星撞击地球,
130	or a super-machinal extract.	或超机械提取物
131	There has to be something unstoppable,	一定有不可阻挡的东西
132	and then if it is a man-made or an artificial thing,	如果它是人造或人造的东西,
133	then it should also come with the intent,	然则,那确是有意的,
134	the ability, and also the uncontrollability.	能力 以及无法控制的能力
135	And I find I not qualify for either of that intent.	我发现我不符合任何一种意图
136	We heard the argument already.	我们已经听到争论了
137	Intent is not coming out of blue, right?	意图不是突然的 对吧?
138	It is a very, very personal	这是一个非常,非常私人
139	and also contextualized project or object	以及背景化项目或对象
140	that is very, very typical to human.	这是人类的典型
141	But for a machine, you call them an accident	但对于机器,你称他们为意外
142	when something goes wrong, rather than you call them	当事情出了问题,而不是你叫他们
143	an intent to kill you, right?	想杀你 对吧?
144	So how to program intent into an AI system right now	所以,如何编程意图 到AI系统现在
145	is unknown, and technically I don't find	不知名 严格来说我找不到
146	there is a kind of a good rationale to justify	有个很好的理由可以证明
147	or to conclude that AI machine in the future	或得出结论,AI机器今后
148	will have that, because AI requires data to train,	因为AI需要数据来训练
149	to learn, and then to extrapolate.	学习,然后推断。
150	It is a statistical interpretation,	这是一个统计解释,
151	and also some extrapolation.	还有一些推断
152	And from the data, it doesn't tell you basically	从数据来看,它基本上没有告诉你
153	to have open questions to be asked.	有公开的问题要问。
154	I heard one great argument about the product manager	我听到一个伟大的争论 关于产品经理
155	and the engineer relationship.	和工程师的关系。
156	It's not about, hey, I want to have product A,	这不是关于,嘿,我想有产品A,
157	therefore someone will go on the input.	因此,有人会继续输入。
158	It takes through a lot of understanding and cross work	它需要通过很多的理解和交叉工作
159	to understand exactly intent,	完全理解意图,
160	and also to come up with a solution,	并且想出一个解决方案,
161	not allowing to come up with the intent itself, right?	不允许自己想出来 对吧?
162	Ability, I just mentioned that.	能力,我刚刚提到。
163	There is a huge technical gap from now to the future	从现在到未来,技术差距很大
164	to even making modest progress in our AI capability.	甚至在我们的人工智能能力上取得微小进展。
165	And certainly, unstoppable, I just mentioned,	当然,无法阻止,我刚刚提到,
166	checking points in AI machines exist in many places.	人工智能机器中的检查点在许多地方都有。
167	After all, it is a software product.	毕竟,这是一个软件产品。
168	Very often time, we call AI, what is AI?	我们经常叫AI, 什么是AI?
169	AI is a method, where AI is an idea.	AI是一种方法,其中AI是一个想法.
170	AI is most likely something that doesn't work,	AI是最有可能不起作用的东西,
171	and when it worked, it's called a product.	而当它起作用时,它被称为一种产品.
172	You call it a car, you call it a TikTok app, or something.	你管它叫汽车,你管它叫TikTok应用程序,或者什么的
173	Once you have a product in your hand,	一旦你手里有东西
174	it is actually offering you all the ways of regulating,	它实际上为您提供了各种调节方法,
175	of controlling, and of managing.	控制,管理。
176	Therefore, I could not draw the dots	因此,我不能画出点
177	to connect AI to a direct word in excess of risk.	将AI连接到一个超过风险的直接单词.
178	And I want to conclude by saying that	最后我想说
179	this type of argument and narrative	这种论点和说明
180	is actually, even societally, not very helpful,	实际上,即使是在社会上, 也没有很大的帮助,
181	because it is not really even promoting	因为它甚至没有真正促进
182	the very spirit of a rational conversation and discussion	理性交谈和讨论的精神
183	about the realizability, about the controllability,	关于可实现性, 关于可控制性,
184	and also about the value and the harm	以及价值和伤害
185	of a technology or a product,	技术或产品,
186	similarly to many other inventions we have.	和很多其它发明一样
187	The kind of excess risk is not actually new	那种过度的风险其实不是新的
188	in human civilization.	在人类文明中。
189	If, well, Oxford University was created when?	如果牛津大学是什么时候创立的?
190	It was during the medieval age,	是在中世纪时代
191	after, basically, a breakthrough made by printing press,	基本上,印刷机的突破,
192	where books are made available to every citizen.	向每个公民提供书籍。
193	At that time, the church was actually having	当时,教堂实际上在
194	the similar type of fear, like many people right now.	同样的恐惧 就像现在很多人一样
195	They say, oh, average people all get their book of Bible,	他们说,哦,普通人 都得到了他们的圣经书,
196	they can read and interpret by themselves.	他们可以自己阅读和解释。
197	What if the book printed it wrong?	如果书印错了怎么办?
198	Or what if they don't understand what it's saying?	或者如果他们不明白它说什么呢?
199	The world will collapse.	世界将崩溃。
200	But what we actually see is the age of enlightenment.	但我们实际上看到的是启蒙时代。
201	People have their own book.	人们有自己的书。
202	They can use the book to do harmful things,	他们可以用书做有害的事
203	but they also can use them to do a lot of things.	但他们也可以用它们做很多事情。
204	But most importantly, they can use them to do thinking.	但最重要的是,他们可以用它们来思考。
205	And also, right now, what I see,	而且,现在,我看到的,
206	the AI is giving us not just one book,	AI给我们的不只是一本书,
207	but the entire universe of libraries	但整个图书馆
208	and all knowledge being created.	创造一切知识。
209	Of course, there are a lot of risks	当然,风险很大
210	in giving people that kind of power.	给人那种力量
211	But I would like to predict that this will lead	但我要预测,这将导致
212	to a positive movement at the same scale of renaissance.	在同样规模的复兴中实现积极的运动。
213	I call them the age of empowerment instead of enlightenment,	我称他们为赋予权力而不是启蒙时代,
214	because you can now worry about using these tools,	因为你现在可以担心 使用这些工具,
215	what kind of problems you want to solve,	你想解决什么样的问题?
216	rather than focusing on how to solve it	而不是专注于如何解决
217	in many of the technical cases.	许多技术案例。
218	We never, for example, experience extinction	比如说,我们从来没有经历过灭绝
219	because some job is lost.	因为有些工作丢了
220	In fact, more are created, right?	事实上,创造更多,对不对?
221	And I think the argument itself should also justify itself	我认为这个论点本身也应该证明
222	by presenting the value and also the cost	通过显示价值和成本
223	it produces in the community.	它在社区里产生。
224	Everything has a risk, right?	一切都有风险,对不对?
225	So a risk isn't equivalent to the magnitude	所以风险不等于规模
226	of existence of risk.	风险的存在。
227	And also, the risk is not necessarily	而且,风险并不一定是
228	from that subject, AI itself.	从这个话题,AI本身。
229	I just mentioned that AI is a very blanket term right now,	我刚刚提到AI现在是一个非常笼统的术语,
230	unfortunately confounded and overloaded	不幸的困惑和超载
231	with many, many things.	有很多东西
232	And what is behind that is actually usually a human party	后面的东西 通常是一个人类派对
233	or some kind of other intent	或某种其他意图
234	that is causing many of the consequences,	造成许多后果,
235	no different from other tools being abused.	与其他被滥用的工具没有区别。
236	So I cannot basically be convinced	所以我根本无法相信
237	or really stand the accusation that I am a terminator,	或者真的接受指控说我是终结者
238	I'm creating something that is going to be	我创造的东西是
239	destroying civilization and extinct human being	毁灭文明和灭绝人类
240	in the world.	在这个世界。
241	Thank you very much.	谢谢

该视频共有字幕 241 条。解锁更多字幕为会员功能，请移动到 价格