【牛津辩论】AGI将是人类最后的发明？
字幕摘录
| 时间 | 英文 | 中文 |
|---|---|---|
| 0:06 | Without further ado, I call upon Chris Mantis Cravaris to open the house for side proposition | 我恳求克里斯・曼蒂斯・克拉瓦里斯 打开房子做副手 |
| 0:20 | Thank you, madam president in terms of the motion. | 谢谢你,总统女士 关于动议。 |
| 0:23 | We're debating tonight | 我们今晚要辩论 |
| 0:23 | Then all the way to 2014 the motto of Facebook was move fast and break things | 然后一直到2014年 Facebook的座右铭是快速移动 打破事物 |
| 0:30 | One imagines their lawyers saw this in prints and recognize the problem, which is that reads like an admission of liability | 人们想象他们的律师看到 这一点在指纹 并承认问题, 即读起来像承认责任 |
| 0:36 | They have since exchanged it for something far more palatable and far more insincere | 之后他们用它换来更美味更不诚实的东西 |
| 0:41 | Which is move fast with stable infrastructure | 以稳定的基础设施快速移动 |
| 0:45 | What I would say though is that the ethos of these corporations are not softened, but it has spread | 但我要说的是,这些公司的气质并没有被软化,但已经蔓延 |
| 0:50 | That is the acceleration that we're seeing in the tech world is the orthodoxy and it is underwriting | 这就是我们在科技界看到的加速 是正统的,它是承载的 |
| 0:56 | What is essentially an arms race and artificial intelligence? | 什么是军备竞赛和人工智能? |
展开字幕全文（1323 条）
| 序号 | 英文 | 中文 |
|---|---|---|
| 1 | Without further ado, I call upon Chris Mantis Cravaris to open the house for side proposition | 我恳求克里斯・曼蒂斯・克拉瓦里斯 打开房子做副手 |
| 2 | Thank you, madam president in terms of the motion. | 谢谢你,总统女士 关于动议。 |
| 3 | We're debating tonight | 我们今晚要辩论 |
| 4 | Then all the way to 2014 the motto of Facebook was move fast and break things | 然后一直到2014年 Facebook的座右铭是快速移动 打破事物 |
| 5 | One imagines their lawyers saw this in prints and recognize the problem, which is that reads like an admission of liability | 人们想象他们的律师看到 这一点在指纹 并承认问题, 即读起来像承认责任 |
| 6 | They have since exchanged it for something far more palatable and far more insincere | 之后他们用它换来更美味更不诚实的东西 |
| 7 | Which is move fast with stable infrastructure | 以稳定的基础设施快速移动 |
| 8 | What I would say though is that the ethos of these corporations are not softened, but it has spread | 但我要说的是,这些公司的气质并没有被软化,但已经蔓延 |
| 9 | That is the acceleration that we're seeing in the tech world is the orthodoxy and it is underwriting | 这就是我们在科技界看到的加速 是正统的,它是承载的 |
| 10 | What is essentially an arms race and artificial intelligence? | 什么是军备竞赛和人工智能? |
| 11 | That is a race being bankrolled by venture capital that is being justified by competitive pressure between market actors | 这是一场由风险资本出资的竞赛,市场参与者之间的竞争压力为这场竞赛提供了理由。 |
| 12 | And that is being sharpened by geopolitical rivalry between countries | 国与国之间的地缘政治竞争使情况更加严峻 |
| 13 | companies like open AI | 公司喜欢开放AI |
| 14 | Anthropic and deep-seek are racing toward a form of artificial general intelligence | 人类和深渊正在走向人工智能 |
| 15 | And what we're looking here is a form of agentic intelligence that is as capable as a human being at most tasks | 而我们所看到的是 一种代理智能形式 与人类一样有能力 最多的任务 |
| 16 | That is it is able to replace your average cognitive worker in the work | 这可以代替你一般的认知工作者 |
| 17 | It does it is not merely just good at a single skill like chess or like driving | 不仅擅长象棋或驾驶这样的技能 |
| 18 | But rather it can do a bit of everything all in one at a fraction of the price | 但是,它可以做一点 所有的东西在一个 价格的一小部分 |
| 19 | The problem though is that if this technology is indeed mishandled | 但问题是,如果这种技术确实处理不当 |
| 20 | It is not only capable of remaking the world as we know it but even unmaking it this evening | 它不仅能够像我们所知的那样重塑世界 甚至今晚无法造就它 |
| 21 | Then I will be arguing two things | 那我就争两件事 |
| 22 | First I will be explaining why a GI will be built and second | 首先,我会解释为什么 将建立一个GI和第二 |
| 23 | I will be explaining why it is a great invention in the classical sense of being vast and being terrible | 我将解释为什么它是一个伟大的发明 在古典意义上是巨大的和可怕的 |
| 24 | Which is precisely why it may well be your last invention | 这正是你最后一次发明的原因 |
| 25 | But before I continue it falls on me of course to introduce the speakers for side | 但在我继续之前,我当然要替大家介绍一下 |
| 26 | opposition | 反对 |
| 27 | Speaking first is air ventura Marty elected to the union secretary's committee in the 37th round of single transferable voting | 首先,在37轮单一可转移投票中,马蒂当选工会秘书委员会成员 |
| 28 | For those of you for those of you not in the know | 你们中那些不知道的人 |
| 29 | This is one of the few institutions where you must secure a democratic mandate before you're allowed to do any unpaid labor | 这是少数机构之一 在你被允许做任何无报酬的工作之前 你必须得到民主授权 |
| 30 | Or before you can clean up after other people's mess | 在别人的乱七八糟之后你才能收拾干净 |
| 31 | They are therefore is very fortunate and securing such a win at the election air | 因此,他们非常幸运,在选举空中赢得胜利。 |
| 32 | However has far more serious credentials when it comes to this debate | 然而,就本次辩论而言,有更为严肃的全权证书。 |
| 33 | She is both an AI enthusiast and also researches the economics of innovation | 她既是AI的爱好者 也是创新经济学的研究者 |
| 34 | I was very impressed however when she nonchalantly mentioned to me over dinner as one does | 但是,我印象非常深刻,当她不 狂喜地提到我 晚餐,像一个人做 |
| 35 | That she worked in the International Communications Office of President Macron in France | 她曾在法国马克龙总统的国际通讯办公室工作 |
| 36 | If anyone here then can convince us that everything will be fine while gently implying that none of this is anyone's fault | 如果这里有人能说服我们 一切都会好起来 同时轻轻地暗示这都不是任何人的错 |
| 37 | It's a and perhaps then a little French bureaucracy is exactly what we need in this debate | 这场辩论需要的是法国的官僚主义 |
| 38 | Either to mitigate AI risk or to trap the entire sector into a decade-long | 要么减轻人工智能风险 要么把整个部门困在十年里 |
| 39 | Consultation after which will announce excellent progress while nothing at all happens | 磋商结束后,将宣布取得良好进展,但一无所获 |
| 40 | Speaking second we have called Benedict fray | 说第二,我们叫本尼迪克特·弗雷 |
| 41 | Professor phrase the detail swatch professor of AI and work at the Oxford Internet Institute | 在牛津因特网研究所工作 |
| 42 | And I have to admit that I was a bit surprised to see him on side opposition on this motion | 我不得不承认,我有点惊讶 看到他站在一边反对这个动议 |
| 43 | And I think that that is you only need to take a look at the demands bibliography to see why and I have picked some | 我认为这是你只需要看看 要求书目,看看为什么,我选了一些 |
| 44 | Select works for the attention of the chamber to illustrate why that is first we have technology at work | 选择作品,供会议室注意,说明为什么我们首先有了技术 |
| 45 | 2.0 the future is not what it used to be then in 2019 we have the technology trap | 2.0 未来与2019年时不同 我们有技术陷阱 |
| 46 | capital labor and power in the age of automation and | 自动化和 |
| 47 | Released only in September we have how progress ends technology innovation and the fate of nations | 直到9月才公布,我们掌握了进步如何结束技术革新和各国的命运。 |
| 48 | Given that I'm very keen to hear why progress will not in fact ends despite the professor's recent academic work | 我非常想知道为什么尽管教授最近做了学术研究 进步实际上不会结束 |
| 49 | Speaking third we have professor Shafi Ahmed who in 2016 | 第三,我们有教授沙菲・艾哈迈德 谁在2016年 |
| 50 | I am very happy to report one a Guinness World Records for being the first to live stream a surgical procedure in | 我很高兴报告一个吉尼斯世界纪录 因为它是第一个 活流的外科手术 |
| 51 | Virtual reality I think this is a remarkable achievement although speaking personally | 虚拟现实 我认为这是一个了不起的成就 虽然我个人说 |
| 52 | I do think there are some parts of life that we perhaps should not be live streaming in | 我确实认为生活的某些部分 我们也许不应该生活流进 |
| 53 | 360 virtual reality and ultra-high definition on the internet | 360虚拟现实和互联网上的超高定义 |
| 54 | However, I'm very excited to hear about how it is that his own profession is likely to be wiped out by the advent of this | 然而,我非常激动地听到 他的职业是如何被毁灭的 |
| 55 | Sort of artificial intelligence and speaking forth is Alex Mitchell the CEO of Boomi | 某种人工智能和说话 是亚历克斯・米切尔的CEO布米 |
| 56 | Which has reportedly generated more than 14 million AI generated tracks as someone who is a bit musically challenged | 超过1400万个AI, |
| 57 | I will not pass any comment on the quality of AI generated tracks in general | 我对人工智能制作的音轨的质量不作任何评论 |
| 58 | But I do find it mildly alarming that AI is now coming even for the suffering artists of this world | 但我确实觉得这有点让人吃惊的是, AI现在甚至为这世界的痛苦艺术家而来 |
| 59 | And that is to say that mr. Mitchell is not only looking to replace our jobs. | 这就是说,先生。 米切尔不仅想取代我们的工作。 |
| 60 | He's even looking to replace the jobless | 他甚至想取代那个没工作的人 |
| 61 | But in all seriousness | 但从严肃的角度看 |
| 62 | We are very honored to have these very distinguished guests that will make for a very fine debate this evening and madam president | 我们非常荣幸有各位贵宾出席 今晚的精彩辩论 总统女士 |
| 63 | They are most welcome in the chamber | 他们最受欢迎 |
| 64 | Before I get into my argument | 在我开始争论之前 |
| 65 | Then I think we have to ask what it is that a GI can do because it is not a | 那我想我们得问一下GI能做什么 因为它不是 |
| 66 | Concept that comes out controversy that is it is best explained with contrast to narrow artificial intelligence that only does one big thing | 与狭隘的人工智能形成对照 只能做一件大事 |
| 67 | AGI is general purpose. | AGI为通用目的. |
| 68 | Not only is it actually smart in the intuitive sense | 不仅在直觉上很聪明 |
| 69 | But rather we're looking at AI that can learn and perform a wide range of tasks at the human level or above | 但是我们所看到的是人工智能 它可以学习和完成 广泛的人类层面或以上的任务 |
| 70 | That is an AI agent that is capable of advanced reasoning of learning of cross-applying skills and of pursuing its goals across the span | 这是一个人工智能代理人,能够对交叉应用技能的学习进行先进的推理,并在整个过程中追求其目标。 |
| 71 | Of time most importantly, this is also an agent that is even capable of editing itself | 时间上最重要的是,这个代理 甚至能够编辑自己 |
| 72 | That is we're looking at systems that are often capable of self improvement over time as in | 这就是我们所看到的系统 往往能够自我改进 随着时间的推移,像在 |
| 73 | This is a sort of development that we're likely to see at the current trajectories at some point within our lifetime | 我们有可能在生命的某个时间看到这种发展 |
| 74 | AGI is a sort of system then that can be as productive as your average human knowledge worker | AGI是一种系统,可以像普通的人类知识工作者一样有效 |
| 75 | Should it be achieved a very useful statistic to note here is that as far as Goldman Sachs is concerned the equivalent of | 如果能实现这一点,需要指出的是,就高盛而言, |
| 76 | 300 million full-time jobs will be subjected to automation pressure before long | 3亿个全职工作不久就会受到自动化压力 |
| 77 | The thing that we have to look at here then is whether this is a debate address too much in the definition | 我们在这里需要研究的是,这是否是定义中争论过多的话题? |
| 78 | There will be people that say that while large language models are only mimicking and don't actually have thoughts | 人们会说,虽然大语言模型只是模仿,实际上没有想法 |
| 79 | So they don't have internal states that qualify as agency | 所以他们没有内部的状态 有资格作为代理 |
| 80 | They're going to call these systems stochastic parrots that don't represent actual intelligence | 他们称这些系统为不代表实际智能的 浮雕鹦鹉 |
| 81 | That's not the point | 这不是重点 |
| 82 | However, and I want you to imagine like you're being attacked by a bear | 不过,我要你想象一下 你被熊袭击了 |
| 83 | even if it's the case that they tell you you have nothing to worry about because the big brown creature chasing you is | 即使他们告诉你你没什么好担心的 因为追着你的大棕色生物 |
| 84 | Actually nothing more than an arrangement of atoms and chemical reactions in the deterministic sense | 实际上,这只是一种 确定性意义上的原子和化学反应的安排 |
| 85 | It may well be that this can put you at ease in the philosophical sense | 从哲学的意义上来说,这很可能会让你轻松 |
| 86 | But this makes no difference when you become the beings dinner | 但是,这没有区别 当你成为生物晚餐 |
| 87 | That is if it looks like a duck and it quacks like a duck. | 如果它看起来像鸭子, 它像鸭子一样颤抖。 |
| 88 | We need to get more regulation going | 我们需要更多的管理 |
| 89 | Therefore, let's now look as to why it is that it is very likely. | 因此,现在让我们看看为什么这是非常可能的。 |
| 90 | We will see the development of AGI | 我们将看到AGI的发展 |
| 91 | the thing to note is that there is a pure economic incentive to bring about this technology and I understand the skepticism we will | 需要注意的是,有纯粹的经济刺激 带来这种技术, 我理解我们将会怀疑 |
| 92 | In terms of something is magnificent | 从某种意义上来说,是伟大的 |
| 93 | But I want you to note all of the different kinds of human innovations we have seen in the last 50 and 100 years | 但我要你注意到 过去50年和100年中 人类的各种创新 |
| 94 | Which we have merely brought about by throwing so much money at the problem | 我们只是把很多钱扔到这个问题上 |
| 95 | You only have to look at things like the Manhattan Project things like the moon landing all of these things may have seemed impossible | 你只需要看看曼哈顿计划之类的东西 像月球着陆这些东西 似乎是不可能的 |
| 96 | 10 15 20 50 years before the fact | 10 15 20 50年前的情况 |
| 97 | But none of this meant we didn't get there by the end and the problem here then is that when you play such a technology | 但这一切都没有意味着我们没有到达 到最后,这里的问题是 当你玩这种技术 |
| 98 | In the middle of a competitive market system when you dangle before the eyes of corporation owners that you can have employees that combine the skills | 在一个竞争的市场体系中间 当你在公司老板的眼皮底下 纠缠着 你可以有员工 结合技能 |
| 99 | Of an engineer and a doctor and a lawyer all in one that work 24 7 | A. 工程师、医生和律师 24 7 |
| 100 | You are going to throw so much investment at this and this is exactly what we're seeing | 你会投入这么多投资 这正是我们看到的 |
| 101 | That is why we're seeing so many billions being funneled into this industry | 这就是为什么我们看到这么多亿 被漏入这个行业 |
| 102 | And the thing is the first company to get there is going to achieve an advantage that no rival can afford to ignore | 这件事情是第一个到那的公司 将取得一个优势 任何对手都无法忽视 |
| 103 | Therefore then what we're seeing is a form of narrow sightedness that is permitting the sort of moral hazard | 因此,我们所看到的是 一种狭隘的目光 允许这种道德危害 |
| 104 | And even if it's the case that many corporations are comfortable with this risk | 即使很多公司都接受这种风险 |
| 105 | Even if they have faith that AI will not pose a risk | 即使他们相信AI不会构成风险 |
| 106 | I don't and the point I'm making here then is that it is not merely for these actors to decide where we go | 我并不知道,我在这里要说的是, 不仅仅是这些演员决定我们去哪里 |
| 107 | Why is it then that this is very likely? | 为什么这很有可能呢? |
| 108 | There is so much that is low-hanging in terms of research that we have yet to explore | 在研究方面还有许多事情有待探讨 |
| 109 | This is things like LLM's adapting to data in real time or expanding the size of data sets | 这是LLM实时适应数据或扩大数据集规模之类的东西 |
| 110 | Especially with new data that has been acquired through the use of artificial intelligence as in seeing how AI works in the workplace | 特别是借助人工智能获取的新数据,例如,在工作场所如何使用人工智能 |
| 111 | But also I don't think that it is very humble to say that human intelligence is all that advanced | 但我也不认为说人类的智慧是进步的 |
| 112 | That is there are very many things were already vastly outpaced by artificial intelligence | 人工智能已经远远超出了很多 |
| 113 | I posit it is only a matter of time before we get a single agent that is able to combine the capabilities of | 我推测我们找到一个能够结合 |
| 114 | All those different types of AI what we're likely to see then is a form of feedback cycle | 我们可能看到的是 不同种类的AI 是一种反馈循环 |
| 115 | We're not only is AI getting better, but AI becomes capable of soft improvement as well | 我们不但AI越来越好,而且AI也变得能够软改进 |
| 116 | What is the problem in this debate then it is precisely that we're not able to complete the task of AI alignment | 这场辩论到底有什么问题呢 正好是我们无法完成AI的校正任务 |
| 117 | That is it is not true that we can perfectly predict what it is that AI is going to do | 我们无法完全预测AI会做什么。 |
| 118 | It is not clear that our training processes can adequately get the agent to want the same things that we want | 不清楚的是,我们的培训过程能否让代理商像我们想要的一样想要 |
| 119 | And in a sense the problem is that if you create something that is highly agentic that can develop itself | 从某种意义上说,问题在于 如果你创造出 高度代理性的东西 可以发展自己 |
| 120 | We are most likely to impede the goals it sets for itself | 我们最有可能阻碍它为自己制定的目标 |
| 121 | Therefore, I don't think that uninterpretable systems are going to suffice in terms of the training that is necessary as | 因此,我认为,在必要的培训方面,无法解释的系统是不够的,因为 |
| 122 | Got Yarkowski and source put it if anyone builds it everyone dies | 找到雅科斯基了 消息源说如果有人建了它 每个人都会死 |
| 123 | And I don't know whether the obituary is going to prove correct | 我不知道讣告是否正确 |
| 124 | But I put it to you we are going to see AI | 但我告诉你,我们要去见AI |
| 125 | Within the span of our lives that is able to reach both a GI level and potentially even at some point of the future | 在我们的一生中,既能达到GI水平,也有可能在未来某个时刻 |
| 126 | ASI whatever assurances these labs periodically give in this market | 无论这些实验室定期在这个市场上做出何种保证 |
| 127 | Caution is an inconvenience rules or suggestions and accountability is only an optional extra | 谨慎是一种不便的规则或建议,问责制只是可选的额外规定。 |
| 128 | An attitude which I can say with only the greatest affection to this place | 我对这个地方最爱的态度 |
| 129 | In fact seems to be shared by many of the officers of the society | 事实上,许多社会官员似乎都认同 |
| 130 | What they would then say is that I don't think regulation is enough the time for caution was yesterday | 他们当时会说,我不认为 监管是足够的时间 谨慎是昨天。 |
| 131 | And today is the time for damage control | 今天是控制破坏的时候 |
| 132 | The question before you tonight then as people who are far more experienced and I are about to explain to you is | 今晚你们要面对的问题是 作为经验更丰富的人 我要向你解释 |
| 133 | Whether we can truly be optimistic about a system whose learned objectives | 我们是否能够真正乐观地看待一个其有经验的目标的体系? |
| 134 | We cannot truly interrogate to or reliably verify and if at the moment it matters most | 我们无法真正讯问或可靠地核查,如果目前最需要核查的话。 |
| 135 | We're not able to turn such a system off. | 我们无法关闭这样的系统。 |
| 136 | I put it to you that we should not hope for its development | 我告诉你们,我们不应该希望它的发展 |
| 137 | Thank you very much. | 谢谢 |
| 138 | I would like to call upon this a venture on Marty to open side opposition | 我想呼吁这个企业 马蒂打开侧面的反对 |
| 139 | Thank You madam president good evening ladies and gentlemen | 谢谢 各位总统夫人 晚上好 女士们先生们 |
| 140 | I must say that I am not at all surprised by how well populated | 我必须说,我一点也不惊讶 人多好 |
| 141 | The proposition ventures tonight because you see every time humanity | 今晚这个提议很冒险 因为你每次看到人性 |
| 142 | Invent something powerful you will always find a circle of nerds declaring that history will end with their egos | 发明一些强大的东西 你总能找到一群书呆子 宣称历史会以他们的自我结束 |
| 143 | They are the same people who weaponized their version of science to suffocate imagination and extinguish humanity's burning desire | 他们就是把科学版本武器化 扼杀想象力 灭掉人类燃烧的欲望的人 |
| 144 | To emancipate itself at every moment of breakthrough | 在突破的时刻解放自己 |
| 145 | But thankfully humanity has refused to be defined by the romanticization of their | 但幸好人类 拒绝被他们浪漫化 |
| 146 | intellectual stagnation disguised as wisdom or by the | 知识停滞 伪装成智慧或被 |
| 147 | anthropological insecurity fed by their selfish hope that nothing greater should exist beyond their time | 人类学的不安全感 源于他们自私的希望 |
| 148 | the rejection of such thinking as the philosophical foundation of our | 拒绝将这种思想作为我们 |
| 149 | understanding of this world is | 对这个世界的理解 |
| 150 | Precisely why humanity has progressed this far and why you and I are still here in this chamber tonight | 确切的说,为什么人类已经进步到这个地步,为什么你和我今晚仍然在这里 |
| 151 | AGI will not be humanity's last great invention | AGI不会是人类最后伟大的发明 |
| 152 | but only a step towards the eternity of memory and perhaps the first opportunity for us to enter a form of | 但只是迈向永恒记忆的一步 也许是我们第一次有机会 |
| 153 | civilization | 人类文明 |
| 154 | Civilizational transmission unknown in the entire history of our species because AGI transcends the biological boundary | 由于AGI超越了生物边界,文明传播在我们物种的整个历史中都是未知的. |
| 155 | That have until now limited our ability to conceive of the persistence of the mind beyond physical decay | 直到现在,这限制了我们 想象心灵的持久性的能力 超越了身体衰败 |
| 156 | and | 和 |
| 157 | Tonight I will not let the pessimism of the proposition attempt to crush that longing for something greater | 今晚我不会让这个提议悲观地 试图粉碎那个渴望更伟大的东西 |
| 158 | That I know exists within each of you | 我知道你们每个人都存在 |
| 159 | But before I begin it is indeed my duty to introduce to you the proposition speakers this evening | 但在我开始之前,我确实有义务向你们介绍今晚的演讲者 |
| 160 | 首先我们让Chris Mentis craveris 为那些不知道Chris是那个 带领我们穿过 | |
| 161 | Constitutional crisis of the Union last term and I would say that he is part of the reason this institution is still alive Chris | 联盟上届的宪政危机 我想说他是这个机构还活着的原因之一 克里斯 |
| 162 | I genuinely respect you for this now | 我现在真的很尊重你 |
| 163 | You know, I was not as interested in your professional career as you clearly were with mine, but I would just like | 你知道,我对你的职业生涯没有你对我的那么感兴趣 但我只想 |
| 164 | But I would just like to say that I don't think your knowledge in AGI | 但我只想说,我不认为 你在AGI的知识 |
| 165 | Development will ever be matched by your expertise in removing presidents of the Oxford Union | 与发展相匹配的是你在罢免牛津联合会主席方面的专门知识 |
| 166 | Then mr. Andersen Berg prominent Swedish researcher futurist and | 接下来是Mr. Andersen Berg 瑞典著名研究员未来学家 |
| 167 | Transhumanist mr. Sandberg what an oxymoron it is to see you seated on that side of the chamber | 跨人性主义者 Mr. 桑德伯格 看到你坐在房间的那一面真是令人难以置信 |
| 168 | I say good luck with your speech later because if you are not intending to deny that AGI | 我会说祝你的演讲好运 因为如果你不打算否认AGI |
| 169 | Constitutes a transhumanist instrument you would be stepping out of this chamber by the end of tonight having basically | 这是一种跨人类的乐器 你将在今晚结束前走出这个会议厅 |
| 170 | Deconstructed your entire career in front of our eyes | 在我们眼前解开你整个职业生涯 |
| 171 | Then mr. James Barrett prominent American filmmaker speaker and author | 接下来是Mr. 詹姆斯·巴雷特 美国著名电影发言人和作家 |
| 172 | I am frankly not quite sure as to how he stumbled into this debate, but I read from your presence | 坦白地说,我不确定他是如何 陷入这场辩论的, 但我从你那里读到的 |
| 173 | And I read from your presence on proposition site tonight a | 我从你今晚在提案网站上读到 |
| 174 | Remarkable lack of confidence in your own work and perhaps a fear that AGI may soon prove far more creative than your own mind | 对自己的作品缺乏信心,或许是担心AGI很快会比你自己的头脑更有创造力 |
| 175 | Then we have dr. Michael Woldridge | 然后我们有医生。 迈克尔·沃尔德里奇 |
| 176 | Renowned professor of artificial intelligence at our very own University of Oxford | 我们牛津大学著名的人工智能教授 |
| 177 | I respect you greatly and in fact, actually there are many I think of your students here watching us tonight | 我非常敬重你 事实上,我觉得今晚有很多学生在这里监视我们 |
| 178 | But for your own sake I do hope that after this debate your lectures would not be | 但为了你,我希望这次辩论之后 你们的演讲不会是 |
| 179 | Desserted because your students would have understood that part of the reason dr. | 因为你的学生会理解 原因的一部分 Dr. |
| 180 | Woldridge gets out of bed every day is to contribute towards building something that he himself is convinced will end humanity's | 伍德里奇每天都下床 是为了帮助建造他相信 将会终结人类的东西 |
| 181 | creativity and | 创造力和 |
| 182 | finally professor Roman Yampolsky | 终于罗曼·扬波尔斯基教授 |
| 183 | Computer scientists at the University of Louisville | 路易斯维尔大学计算机科学家 |
| 184 | I genuinely admire your concrete work on preventing AI from modifying its own source code and I look forward to hearing more about it | 我非常钦佩你为阻止AI修改其源代码所做的具体工作 我期待着听到更多关于它的信息 |
| 185 | Madam president, these are your speakers and they are the most welcome | 总统女士,这是您的发言人,他们是最受欢迎的 |
| 186 | now | 现在 |
| 187 | We can waste hours | 我们可以浪费时间 |
| 188 | Arguing how AGI will hardly be ever achievable because its energy consumption is unsustainable | 争辩AGI的能源消耗无法持续, |
| 189 | Or how microchips will never be able to scale efficiently, but that is not the purpose of this motion | 或者微芯片将永远无法有效放大, 但这不是这个运动的目的 |
| 190 | For this motion to even make sense one must assume that AGI is | 为了让这个动议有意义 必须假设AGI是 |
| 191 | Achievable and I think what we should really be asking is how much risk are we willing to accept an exchange for? | 我们是否愿意接受交换? |
| 192 | Transformation brought by AGI. | AGI带来的转变. |
| 193 | Are we ready to accept that AGI might indeed redefine humanity itself? | 我们是否准备接受爱奇艺国际确实可以重新定义人类本身? |
| 194 | And perhaps one should ask what defines humanity? | 也许我们该问一下人类的定义是什么? |
| 195 | Is it rationality? | 是理性的吗? |
| 196 | Consciousness moral autonomy if a GI were to embody these attributes what happens then? | 如果GI体现这些属性 那么会发生什么? |
| 197 | My answer is for the very first time in history | 我的答案是历史上第一次 |
| 198 | Humanity which has always been so so powerless before the erosion of mind in the face of the inevitable | 人类总是如此无能 在面对不可避免的心灵侵蚀之前 |
| 199 | Biological failure of organisms might glimpse something very radical | 生物体的生物衰竭 可能会发现非常激进的东西 |
| 200 | Which is the continuity of the mind beyond biology biology? | 生物生物学之外的思想的连续性是什么? |
| 201 | I think many people in this room probably have experienced the love of the loss of someone they loved | 我想这房间里的很多人 都经历过失去爱人的爱 |
| 202 | And I know that there is nothing more gut-wrenching than staring death | 我知道没有什么比看死更让人心痛的了 |
| 203 | Straight in the eyes and realizing how utterly incapable you are of stopping it from carrying away the consciousness of the person you cherished | 直视着眼睛 发现你完全无法阻止它 夺走你所珍惜的人的意识 |
| 204 | What if AGI is our only chance to subvert that finality as the closest functional pathway | 如果AGI是我们唯一的机会 破坏最终的功能路径 |
| 205 | We currently know towards modeling human level cognition beyond biology | 目前我们知道人类认知模式 超越生物学 |
| 206 | AGI is therefore the manifestation of our | 因此AGI就是我们 |
| 207 | Civilizational attempts to fight against this appearance that accompanies the decay of the human body | 文明主义试图对抗这种伴随人体腐烂的外表 |
| 208 | That is why we build pyramids libraries museum to preserve memory | 这就是为什么我们建造金字塔图书馆 保存记忆 |
| 209 | I mean, of course one could then argue that AGI may eventually attain a level of autonomy that surpasses us entirely that it might | 我的意思是,当然,人们可以这样认为,AGI最终可能获得的自主程度会完全超越我们,因为它可能 |
| 210 | Outthink us out innovate us and from that point on we're generate every subsequent great invention rendering our biological cognition | 想象我们创新 从那时起,我们产生 每一个伟大的发明 使我们的生物认知 |
| 211 | Comparatively obsolete and this is what I mean by risk tolerance | 比较过时,这就是我所说的风险承受能力 |
| 212 | What are we prepared to risk in pursuit of a more enduring preservation of human memory and pursuit? | 为了更持久地保存人类记忆和追求,我们准备冒什么风险? |
| 213 | ultimately of human emancipation | 最终实现人类解放 |
| 214 | So despite how much many scientists and I would say that the fetus of the proposition tonight seek to portray AGI as a mere | 尽管我和科学家们说 今晚提案的胎儿 试图将AGI描绘成一个 |
| 215 | Technological undertaking it has in my opinion in fact always been a fundamentally philosophical question | 我认为,技术实践实际上始终是一个基本哲学问题。 |
| 216 | It is about whether we are willing to make a decision in | 而是我们是否愿意在 |
| 217 | Inventing an intelligence whose evolution we cannot perfectly predict and may never fully control | 发明一种我们无法完美预测 也不可能完全控制的智能 |
| 218 | But with the aim of evolving towards a future in which our memory is no longer bound by the fragility of flesh | 但为了走向未来 我们的记忆不再被肉体的脆弱所束缚 |
| 219 | Humanity has always been risk takers | 人类一直是冒险者 |
| 220 | It is in our very own nature to evolve in the face of uncertainty about our very own existence | 面对我们自身生存的不确定性 进化是我们自己的本性 |
| 221 | From the Manhattan project to traveling to space you name it | 从曼哈顿计划到太空 |
| 222 | What do you think dr. Wodrich gets up to every day in his Oxford labs? | 你觉得呢,博士? 沃德里希每天在他的牛津实验室里? |
| 223 | He researches something called multi-age AI agent system that enable the | 他研究了一种叫做多年龄AI代理系统 能够使这个系统 |
| 224 | Collection of independent interacting AI agents in other words it directs contribution to AGI | 收集独立互动的AI代理,换句话说,它指导对AI的贡献 |
| 225 | So if someone like him tonight tries to convince you that AGI will be the last great invention of humanity | 所以如果有人像他一样 今晚试图说服你 AGI将是人类最后伟大的发明 |
| 226 | It is not because he seeks to gently warn you of its risks | 不是因为他想轻轻警告你 |
| 227 | But rather it is a weapon ization of his authority in this field to shut down | 但是,这是武器化 他的权力在这个领域关闭 |
| 228 | the imagination of younger generations who may wish to contribute further to humanity's | 年轻一代的想象力,他们可能希望为人类 |
| 229 | Emancipation through a GI because he wants to be amongst the last remember to have subverted our | 解放通过GI 因为他想成为最后的记忆之一 颠覆了我们 |
| 230 | existence and I will tell I will tell every single one of them that unfortunately | 我会告诉每个人 不幸的 |
| 231 | Humanity will certainly not end with their egos and its creativity not with their legacy | 人类当然不会以他们的自我和创造力结束,不会以他们的遗产结束。 |
| 232 | the future | 未来 |
| 233 | Fundamentally belongs to our generation | 基本上属于我们这代人 |
| 234 | Do not let anyone kill your longing to contribute to humanity when beyond these walls the field of AGI is | 不要让任何人杀 你渴望为人类贡献 当这些墙外的AGI领域 |
| 235 | Ava is advancing at extraordinary speed driven by researchers and leaders who are all far more | 阿瓦以超乎寻常的速度前进 由研究人员和领导者驱动 他们更远 |
| 236 | Acknowledgeable than the proposition tonight who dare even to flirt with the idea of telling you that what they do will be the last great | 比今晚的命题更能被认可 敢于调情的想法 就是告诉你,他们所做的将是最后伟大的 |
| 237 | Thing and not what you do | 事情不是你们做的 |
| 238 | AGI is not the last great invention of humanity. | AGI并不是人类的最后伟大发明. |
| 239 | It is only the beginning vote opposition tonight, ladies and gentlemen | 女士们,先生们,今晚只是开始反对投票 |
| 240 | Next I would like to call upon the second speaker for a side proposition | 下面我要请第二位发言者提出一个附带建议。 |
| 241 | The motion before us tonight is remarkable because it can be interpreted in two different ways | 今晚摆在我们面前的动议是了不起的,因为它可以有两种不同的解释方式 |
| 242 | We can interpret it in a very optimistic way if we get a GI | 我们可以以非常乐观的方式解释 如果我们得到GI |
| 243 | We will not have it to invent anything more human innovation is going to stop | 我们不会有它来发明 任何更多的人类创新会停止 |
| 244 | But innovation and humanity might very well keep on going | 但创新和人性可能继续 |
| 245 | But of course one can take the very pessimistic take that yeah the dark end | 但是,当然,人们可以采取 非常悲观的 采取是黑暗的结局 |
| 246 | It's the last great innovation because after that there is nobody around | 这是最后伟大的创新 因为在那之后,没有人在身边 |
| 247 | So the opposition has a problem here. | 所以,反对派有一个问题在这里。 |
| 248 | It kind of needs to defeat both | 这需要打败两个 |
| 249 | Statements I'm the optimist around here | 声明我是这里最乐观的 |
| 250 | I have other fellow panelists is way more pessimistic than me | 我还有别的小组成员 比我更悲观 |
| 251 | But I think even the optimistic take on this has some interesting | 但我认为,即使是乐观的 这一点有一些有趣的 |
| 252 | Implications, so AGI is a system that performs the full range of cognitive task | 影响,所以AGI是一个系统 履行所有认知任务 |
| 253 | for solving novel problems across different domains | 解决不同领域的新问题 |
| 254 | That in those involves identifying problems generating solutions evaluating them testing them and iterating and | 其中涉及查明产生解决办法的问题,对解决办法进行评估,对其进行测试,并延长和延长。 |
| 255 | Implementing them and it might be at human level. | 执行这些公约,可能在人一级执行。 |
| 256 | It might be above human level | 可能比人类还要高 |
| 257 | We don't need to assume anything about consciousness or feeling or that it enjoys it | 我们不需要任何关于意识或感觉 或它喜欢的东西 |
| 258 | That might be very well a discussion for another and evening in this house | 这也许是一个很好的讨论 在另一个晚上在这个房子里 |
| 259 | Whether these systems actually have an inner life. | 这些系统是否真的有内在生活. |
| 260 | The only thing that matters is that it might be able to do inventions | 唯一要紧的是 它可能能够做发明 |
| 261 | So what is an invention? | 发明是什么? |
| 262 | I would say a great invention | 我会说一个伟大的发明 |
| 263 | Substantially expands the range of the what is achievable it opens doors previously closed | 大幅扩大可以实现的目标的范围,打开以前关闭的大门 |
| 264 | The printing press it enabled mass knowledge the steam engine | 它的印刷机让大众了解蒸汽机 |
| 265 | Mechanization of labor the computer the automation of calculation each opened a specific door a | 劳动机械化 计算机 计算自动化 每门打开一个特定的门 a |
| 266 | Specific task although some of them very general and useful | 具体任务,尽管其中一些任务非常笼统和有用 |
| 267 | AGI opens the door to opening doors. | AGI打开开门门. |
| 268 | It makes the innovation process doable | 它使创新进程得以实现 |
| 269 | So when we think about how hammers extend our force and telescope extend our site, etc | 当我们思考锤子如何延伸 我们的力量和望远镜扩展 我们的场地等 |
| 270 | we find that what the AGI does is of course automation of | 我们发现AGI的工作当然是自动化的 |
| 271 | Formulating and problems and extends our capacity to extend the capacities | 提出问题,扩大能力 |
| 272 | What does that lead to? | 这导致什么? |
| 273 | Well, it's for the computer scientists around the room | 是给房间里的电脑科学家的 |
| 274 | This is a little bit like the universal Turing machine all specific divide | 这有点像通用的图灵机 所有特定的鸿沟 |
| 275 | Computational devices can be emulated by the universal Turing machine | 计算设备可由通用图灵机模拟 |
| 276 | There is no need anymore for those specific devices in principle. | 原则上不再需要这些具体装置。 |
| 277 | You only need the universal machine | 你只需要通用的机器就够了 |
| 278 | So it means that we no longer having to choose which doors to open | 也就是说我们不用再选择要打开的门了 |
| 279 | The choosing can be done by the opening all of them | 选择可以全部打开 |
| 280 | Now some people might object wait a minute. | 现在有些人可能会反对 等待一分钟。 |
| 281 | We humans are deciding what problems the AGI will solve | 我们人类在决定 AGI会解决什么问题 |
| 282 | That makes us the inventors but that's a little bit like saying that the inventor of the light bulb wasn't Edison | 这让我们成为发明家 但有点像是说 灯泡的发明人不是爱迪生 |
| 283 | But the first person complained about the darkness | 但第一个人抱怨黑暗 |
| 284 | Indeed the person who notices that there is a problem | 发现有问题的人 |
| 285 | It might be even the client not the inventor and indeed it might very well be a machine that notices that there are problems | 可能连客户都不是发明家 也可能是机器发现有问题 |
| 286 | So when we look at the empirical evidence today | 所以,当我们今天看 经验证据 |
| 287 | We see that systems like Alpha fold not did they just speed up the analysis of protein folding it more or less solved it | 我们看到,像Alpha折叠这样的系统 他们没有只是加快蛋白质折叠的分析 多少解决它 |
| 288 | Gnome | 格诺米 |
| 289 | Generated a hundred thousand the novel structure stable materials | 创造出十万个新结构的稳定材料 |
| 290 | Which is a combinatorial space much faster than a human mind can ever explore in mathematics | 这是一个组合空间 比人类心灵在数学中探索的速度要快得多 |
| 291 | We're seeing a novel and approves and conjectures that actually surprised professionals | 我们看到了一本小说 认可和猜想 这实际上令专业人士惊讶 |
| 292 | We're not quite there yet that at the state where the math department is seriously worried | 在数学系非常担心的州里 我们还没到 |
| 293 | But they're going to be more worried by the fall term | 但他们会更担心秋天 |
| 294 | And of course when it comes to code, we're again seeing in the systems that architect code | 当然,当它涉及到代码, 我们再次看到在系统中 建筑师代码 |
| 295 | They write it they debug it they iterate on it | 他们写它,调试它,他们在上面 |
| 296 | And it is transforming how we do their work in software. | 它正在改变我们在软件方面的工作方式。 |
| 297 | Isn't that invention? | 这不是发明吗? |
| 298 | Isn't that the form of innovation? | 这不是创新的形式吗? |
| 299 | It's doing | 还在做 |
| 300 | Now this gets to a dark point Michael Nielsen | 迈克尔·尼尔森(英语: |
| 301 | I have a dear friend of mine pointed out that this acceleration of innovation means that we get all the innovations in a much shorter time | 我有一个亲爱的朋友指出,这种加速创新意味着我们在更短的时间内获得所有创新 |
| 302 | Both the wonderful and the terrible ones and given how much problems we have had in dealing with some of our terrible inventions like | 无论是美妙的还是可怕的 也考虑到我们在处理一些可怕的发明时遇到的问题 |
| 303 | nuclear weapons | 核武器 |
| 304 | Bioweapons and various nerve agents or even foolish inventions like adding lead to gasoline | 生物武器和各种神经毒剂 甚至是愚蠢的发明 比如在汽油中添加铅 |
| 305 | If we get all of that | 如果我们得到这一切 |
| 306 | And greater things yet simultaneously that might easily overwhelm our ability to coordinate and deal with them | 同时还有更多的事情 可能轻易地压倒我们协调和处理它们的能力 |
| 307 | That is a real threat. | 这是一个真正的威胁。 |
| 308 | We're threatened by the cornucopia of innovation | 我们受到革新主义的威胁 |
| 309 | AGI compresses the timeline for all future inventions and that is something we might be seriously worried about | AGI压缩了未来所有发明的时间线 这是我们可能非常担心的 |
| 310 | Similarly, we have an economic argument. | 同样,我们有一个经济论点。 |
| 311 | I think which maybe | 我想也许 |
| 312 | You might appreciate | 你可能会感激 |
| 313 | One in eighteen hundred eighty percent of british were working agriculture today. | 18.8%的英国人今天务农。 |
| 314 | It's less than two percent | 这还不到2% |
| 315 | But we still need farming to get food, but we have automated it to a very large degree | 但我们还是需要农作来获得食物 但我们已经在很大程度上实现了自动化 |
| 316 | Muscles have become fairly irrelevant if you're a farmer | 肌肉已经变得相当无关紧要 如果你是一个农民 |
| 317 | And a gi does the same thing for innovation | 创新也一样 |
| 318 | Now the interesting thing here is that the labor going into innovation can of course be used individually anyway | 现在有趣的是 进入创新的劳动力 当然可以单独使用 |
| 319 | Many of us might still want to innovate | 我们很多人都想创新 |
| 320 | We might want to tinker we might want to do things | 我们可能想修补 我们可能想做一些事情 |
| 321 | But that's a little bit like how today many people appreciated hand woven clothing | 但有点像今天很多人 欣赏手织的衣服 |
| 322 | Even though we know most of them what we're wearing was made in a factory using a far more efficient machine | 尽管我们知道大多数我们穿的衣服 是工厂制造的 使用更高效的机器 |
| 323 | So even if you might have comparative advantages in some respect maybe taste maybe some other in the state | 所以即使你在某些方面有比较优势 也可能尝到其他的味道 |
| 324 | That might still be a microscopic aspect of a total innovation system | 这可能是整个创新体系的一个微观方面 |
| 325 | Most innovations are going to be done by a gi | 大部分的革新都要用一个基 |
| 326 | At least at first may be directed by humans seeing problems, but later on also directing themselves into finding problems | 人类至少会先发现问题 但后来也会发现问题 |
| 327 | Worthy solving which means that whatever great innovations are happening | 值得解决 意味着无论有什么重大创新 |
| 328 | They are going to be more properly ascribed to the to the a gi | 他们更应该归根结底 |
| 329 | so what new domains of inventions could emerge that | 因此,什么新的领域 发明可以出现, |
| 330 | Really generalizes | 真正的概括 |
| 331 | That really general intelligence system | 那个真正的普通情报系统 |
| 332 | Could not address and the problem is if there is such a domain, why do we think even humans could do it? | 不能解决,而问题是,如果有这样一个领域,为什么我们认为即使是人类也能做到这一点? |
| 333 | so | 这样 |
| 334 | About 3.4 million years ago in rift valley. | 约340万年前在裂谷. |
| 335 | The first tool was invented | 第一个工具是发明的 |
| 336 | Somewhere around there | 在附近某个地方 |
| 337 | One of our ancestors took a stone sharpened it and then made the first step on a long path forward | 我们的一位祖先用石头磨磨了它 然后在漫长的道路上迈出了第一步 |
| 338 | That set things into motion that stone tool led to other stone tool | 石头工具导致其他的石头工具 |
| 339 | It led to ways of tilling the soil to doing agriculture to the wheel to generally harnessing nature | 它导致将土壤耕田 做农业 轮子一般利用自然 |
| 340 | And shaping it into ever more sophisticated tools | 把它塑造成更精密的工具 |
| 341 | That's the process accelerated when we made tools for making tools leonardo da vinci's greatest invention might actually have been a device | 当我们制造工具 来制造Leonardo da vinci最伟大的发明时 这一过程就加快了 |
| 342 | for making better screws | 用来制造更好的螺丝 |
| 343 | In order to make that invention work you actually need to hand carve a wooden screw, but you can use that innovation | 为了让发明成功 你实际上需要手雕一个木螺钉 但你可以使用这个创新 |
| 344 | to make better screws | 做更好的螺丝 |
| 345 | But it can only make screws it helped eventually lead to the industrial revolution | 但是它只能制造螺丝 它帮助最终导致工业革命 |
| 346 | But it could only do one part of it the humans were still needed during the industrial revolution | 但是它只能做它的一部分 在工业革命期间人类仍然需要 |
| 347 | Our tools became better our precision became better. | 我们的工具变得更精准了 |
| 348 | We could automate things | 我们可以把事情自动化 |
| 349 | But eventually we are reaching a system that can do what our ancestor did and actually across every domain | 但最终我们到达了一个系统 能够做我们的祖先所做的 并且实际上跨越每一个领域 |
| 350 | at every scale | 每个尺度 |
| 351 | Because the big promise and the big threat here of agi is that it's scalable | 因为最大的承诺和最大的威胁是 它可以伸展 |
| 352 | And that we might indeed get an | 我们可能会得到一个 |
| 353 | Aware how a data center full of geniuses all of them working on coming up with solutions | 意识到一个充满天才的数据中心 如何让所有在寻找解决方案的人 |
| 354 | And of course one can then spend a career like me arguing that we should use technology | 当然,我们可以花一个职业生涯 像我一样 争论我们应该使用技术 |
| 355 | And to enhance humanity and I think this is actually a form of human enhancement. | 为了增强人性,我认为这其实是增强人性的一种形式。 |
| 356 | I'm an optimist | 我是一个乐观主义者 |
| 357 | I think we have a good chance of actually harnessing this well enough to get more of a control of the good technologies than the bad one | 我认为,我们有一个很好的机会 真正利用这个井 得到更多的控制好的技术 而不是坏的技术 |
| 358 | Intellectual honesty leads me to think that the era of human innovation is going to end | 理智诚实让我觉得人类创新的时代即将结束 |
| 359 | But that's fine | 没关系 |
| 360 | I think this is not | 我觉得这不是 |
| 361 | Disgrace not a tragedy a species that builds something smarter than itself is not diminished by it | 羞辱不是悲剧 一个比自己更聪明的物种 不会因为它而减少 |
| 362 | But in some sense defined by that agi is going to be the last great invention not because it ends our story | 但从某种意义上说 意气是最后一个伟大的发明 不是因为它结束了我们的故事 |
| 363 | But it generates a possible completion of that part of our story | 但是,它产生了一个可能的完成 我们的故事部分 |
| 364 | We invented the ability to invent and each previous great imagine left most of the work | 我们发明了发明的能力 之前的每个伟大的想象都留下了大部分作品 |
| 365 | For other inventors and other device | 其他发明者和其他装置 |
| 366 | This one is going to be the one that is the last because it does all the other inventions for us. | 这个将是最后一个,因为它为我们做了其他所有发明。 |
| 367 | Thank you | 谢谢 |
| 368 | Next I would like to call upon the second speaker for side opposition | 接下来,我要请第二位发言的一方反对者发言。 |
| 369 | ladies and gentlemen | 女士们,先生们 |
| 370 | Thank you for being here tonight. | 谢谢你今晚能来 |
| 371 | It's a great pleasure to be with you | 很高兴与你同在 |
| 372 | Nobody in this room can know what the future holds | 这个房间里没人知道未来会怎样 |
| 373 | it is | 没错 |
| 374 | Indeed unknowable and that's what we're here to debate | 确实不知道,这就是我们在这里的辩论 |
| 375 | But we can at least say something about | 但我们至少可以说说 |
| 376 | What needs to be assumed? | 需要假设什么? |
| 377 | 主张ai | |
| 378 | Will be our last great invention | 这将是我们最后的伟大发明 |
| 379 | to hold | 显示 |
| 380 | Indeed, we're already making an assumption here, right? | 事实上 我们已经在假设了 对吧? |
| 381 | We're making the assumption that agi has | 我们假设Agi已经... |
| 382 | Arrived is capable of planning. | 到达是能够规划的。 |
| 383 | It's capable of reasoning | 能够推理 |
| 384 | It's capable of coming up with new solutions | 它能提出新的解决方案 |
| 385 | but | 不过 |
| 386 | We have to make a number of other | 我们必须做一些其他的 |
| 387 | implicit | 隐含 |
| 388 | assumptions to get to the view that is going to be our very | 假设我们将会看到 |
| 389 | last invention | 上次发明 |
| 390 | And I think when you start to think about those implicit assumptions more seriously | 我认为当你开始思考 这些隐含的假设 更认真 |
| 391 | You will | 你会的 |
| 392 | Probably like me arrive at the conclusion that it is unlikely | 可能像我得出一个结论 这是不可能的 |
| 393 | And so let me go through | 让我过去 |
| 394 | What those implicit assumptions are? | 这些隐含的假设是什么? |
| 395 | Discovery is a long and complex process | 发现是一个漫长而复杂的过程 |
| 396 | and so if agi | 所以,如果爱姬 |
| 397 | Is responsible for generating new ideas? | 是否负责产生新的想法? |
| 398 | That's not all what discovery is about | 这不是所有发现 |
| 399 | The project is about defining the objective | 该项目旨在确定目标 |
| 400 | It is about measuring what is actually success. | 它涉及衡量什么是实际成功。 |
| 401 | It is about implementing it doing the field work | 它是为了执行它来进行实地工作 |
| 402 | Doing the experiments carrying out that process and doing all of that within | 进行实验 进行所有实验 |
| 403 | the norms and rules of regulations that society | 社会规章的规范和规则 |
| 404 | provides | 提供 |
| 405 | and | 和 |
| 406 | all of that | 这一切 |
| 407 | Means that it's tremendously complex and it varies from project to | 意味着它非常复杂 并且因项目而异 |
| 408 | project | 项目 |
| 409 | And I mentioned this because science science and innovation | 我提到这一点是因为科学和创新 |
| 410 | requires | 要求 |
| 411 | many different steps and tasks that are tightly interlinked | 许多不同的步骤和任务密切相关。 |
| 412 | And if any one any single one of those links fail | 如果其中任何一个环节都失败了 |
| 413 | That mean that the entire project fails, right? | 这意味着整个项目都失败了 对吧? |
| 414 | And so the space shuttle challenger | 所以航天飞机挑战者 |
| 415 | it | 这个 |
| 416 | crashed after 73 seconds | 73秒后坠毁 |
| 417 | It had some very | 曾经有过很多 |
| 418 | Good talent world-class astronauts. | 优秀人才世界级宇航员. |
| 419 | It had some of the | 它有一些 |
| 420 | best engines in the world best software | 世界上最好的引擎 最好的软件 |
| 421 | But failed because a small | 但失败了,因为一个小 |
| 422 | rubber seal | 橡胶密封 |
| 423 | failed | 失败 |
| 424 | And because of that the entire project the entire system failed | 整个计划失败了 |
| 425 | science is | 科学是 |
| 426 | like that, right even if | 像这样,即使 |
| 427 | ai | (单位:千美元) |
| 428 | Supercharges | 超额收费 |
| 429 | Medical discovery you still need to go through the clinical trials. | 医学发现你还需要通过临床试验 |
| 430 | You still need to go through | 你还需要经过 |
| 431 | running all the experiments and taking that | 运行所有实验 并带走 |
| 432 | innovation to | 创新 |
| 433 | Markets to patients and that's what's needed for | 市场给病人,这是需要的 |
| 434 | actual impact because to make an invention great | 实际影响,因为 使一个伟大的发明 |
| 435 | It needs to improve the human condition in some way | 它需要以某种方式改善人的状况 |
| 436 | And unless you're doing that what's great about it to begin with | 除非你在做这些 首先,这是伟大的 |
| 437 | So if you think that ai | 所以,如果你认为ai |
| 438 | Will supercharge discovery and generate new ideas | 将超量发现并产生新的想法 |
| 439 | That might be right | 这也许是正确的 |
| 440 | But what's going to happen then unless it automates the entire full process with all the linkages | 但是,会发生什么,除非它自动化 整个过程与所有的联系 |
| 441 | the hard bits | 硬点 |
| 442 | The part that has value | 有价值的部分 |
| 443 | Is going to migrate to actually | 要迁移到实际上 |
| 444 | Getting things done | 把事情办好 |
| 445 | Which involves a lot of tacit knowledge where humans are quite likely going forward to have the comparative advantage | 这涉及到许多隐性知识 人类很可能拥有比较优势 |
| 446 | Now you may not accept that and that takes me to the second implicit | 现在你可能不接受 这让我想到第二个隐含 |
| 447 | assumption | 假设 |
| 448 | We're not here to talk about whether agi | 我们不是来谈阿吉的 |
| 449 | Will be better at humans in some particular tasks | 在某些特殊的任务中,人类会更好 |
| 450 | We are even arguing that it's going to be the last invention which means that | 我们甚至争论说,这将是 最后一个发明,这意味着 |
| 451 | 阿吉语Name | |
| 452 | Also needs to be better than agi | 还要比阿吉好 |
| 453 | plus humans | 加上人类 |
| 454 | I think that is also unlikely because | 我觉得也不太可能 因为... |
| 455 | the ai that we're currently building | 我们目前正在建造的AI |
| 456 | Is very different from the human brain | 和人类大脑完全不同 |
| 457 | It has | 它已经 |
| 458 | certain strengths | 某些优势 |
| 459 | But it also has certain blind spots | 但它也有一些盲点 |
| 460 | and | 和 |
| 461 | Almost any domain I can think of whether it's board games whether it's medical diagnostics | 几乎任何领域,我可以想 它是不是棋局游戏 无论是医学诊断 |
| 462 | Whether it's coming up with new ideas in marketing campaigns | 市场营销运动中是否有新想法 |
| 463 | ai | (单位:千美元) |
| 464 | Plus humans do better than each | 加上人类比每个人做得更好 |
| 465 | Individually and if the human is the differentiating factor | 个人和如果人为区别因素 |
| 466 | In that partnership well, that's where the value resides | 在这种伙伴关系中,这就是价值所在 |
| 467 | And that's then primarily a human led | 这主要是人类的引领 |
| 468 | invention | 发明 |
| 469 | Which takes me to point number three | 这让我想到第三点 |
| 470 | Which is that? | 哪个? |
| 471 | the world | 世界大战 |
| 472 | Is not just the sum of scientific knowledge and hayek was early to recognize that | 不只是科学知识的总和 和干草很早就认识到 |
| 473 | There is a lot of knowledge | 有很多知识 |
| 474 | out there in the world | 外面的世界里 |
| 475 | Which is not scientific | 这不是科学 |
| 476 | It's not codifiable | 无法编译 |
| 477 | It's tacit | 这是默认的 |
| 478 | and | 和 |
| 479 | It's something that is known | 这是已知的东西 |
| 480 | by maybe just a few people in some pockets somewhere in | 也许只有几个人在某个口袋里 |
| 481 | the world | 世界大战 |
| 482 | and | 和 |
| 483 | that | 那个 |
| 484 | Uh is important because many of the great inventions that we've seen | 因为我们见过很多伟大的发明 |
| 485 | Historically have built on this kind of tacit knowledge, right? | 历史上这种隐性知识是建立在这样的基础上的,对吧? |
| 486 | Or a red red ration therapy for example, which saves millions of children | 比如说红色的配给疗法 拯救了数百万儿童 |
| 487 | from death due to | 死亡时死亡原因 |
| 488 | diarrhea and dehydration | 腹泻和脱水 |
| 489 | The chemistry is extremely simple. | 化学性质极其简单. |
| 490 | It's just mixing water | 只是搅拌水 |
| 491 | sugar | 糖类 |
| 492 | and salt | 盐 |
| 493 | The invention was making that | 发明就是这样 |
| 494 | process simple | 过程简单 |
| 495 | and trustworthy | 值得信赖的, |
| 496 | For families and households to do that at scale | 使家庭和住户大规模这样做 |
| 497 | in their homes without having to go to | 在他们家里,不用去 |
| 498 | a hospital that required local knowledge | 需要当地知识的医院 |
| 499 | And it required | 这需要 |
| 500 | the capacity | 能力 |
| 501 | To get things done at scale | 让事情在规模上完成 |
| 502 | Same is true in manufacturing | 制造业也是如此 |
| 503 | If you take the model t | 如果你把模型T |
| 504 | which | 哪个 |
| 505 | became the people's | 成为人民的 |
| 506 | vehicle | 车辆 |
| 507 | What did that require? | 这需要什么? |
| 508 | Well, it required four decades of experimentation with electricity on the factory floor to move | 嗯,它需要40年的实验 在工厂地板上的电力移动 |
| 509 | from | 从 |
| 510 | unit group drive to | 单位组驱动器 |
| 511 | unit right | 单位右 |
| 512 | It's required experimentation on the factory floor | 这需要工厂的实验 |
| 513 | to get assembly lines up and running | 使组装线上线和运行 |
| 514 | Same is true for teota lean production and if you look around the world, right? | Teota精益生产也是如此 如果你环顾世界,对不对? |
| 515 | The reason why most of the world's semiconductors are built in a small island that's threatened to be invaded | 世界上大部分半导体 都建在一个被入侵的小岛上 |
| 516 | Is tacit knowledge the reason that digital industries are | 暗中认识是数字工业 |
| 517 | Uh clustered in certain places like silicon valley | 聚集在硅谷之类的地方 |
| 518 | Is because that knowledge cannot easily be transferred from one place to another if it could be | 这是因为如果知识能够从一个地方转移到另一个地方是不容易的。 |
| 519 | We wouldn't have those kind of clusters | 我们不会有这种集群 |
| 520 | Now you might say we're just going to equip all spaces with sensors with cameras | 现在你可以说,我们只是要 装备所有空间的传感器 摄像头 |
| 521 | with | 与 |
| 522 | listening devices | 监听设备 |
| 523 | But you know there you run into eventually reality in terms of politics and law | 但是你知道你最终会遇到 政治和法律方面的现实 |
| 524 | The gdpr doesn't permit that | GDpr不允许这样 |
| 525 | Nor does the eu ai act and the gdpr is being implemented as a blueprint | euai法案和gdpr也没有作为蓝图实施 |
| 526 | rights across the world | 在世界范围内保护人权 |
| 527 | and so | 就这样 |
| 528 | Looking at these factors. | 审视这些因素。 |
| 529 | It seems to me highly unlikely that all of these | 在我看来 这一切不太可能 |
| 530 | Problems are going to be alleviated and for the proposition to hold | 问题将会得到缓解,并能够维持 |
| 531 | You need to essentially assume that | 你得从本质上假设 |
| 532 | Ai will be able to do every single link in the discovery process | Ai在发现过程中可以完成每一个环节 |
| 533 | You need to assume new complementarities no complementarities between humans and ai | 你需要承担新的互补 人类和Ai之间没有互补 |
| 534 | No tacit or local knowledge | 没有默示或当地知识 |
| 535 | You need to assume new no regulation and you need to assume that basically | 你需要假设新的没有规定, 你需要假设基本上 |
| 536 | All of this holds across the entire world across the entire planet | 所有这一切 遍布全世界 贯穿整个地球 |
| 537 | that to me | 对我来说 |
| 538 | Seems like a very strong | 似乎很强壮 |
| 539 | Assumption indeed and I would therefore advise the house to reject the motion. | 因此,我建议众议院驳回这项动议。 |
| 540 | Thank you | 谢谢 |
| 541 | I would next like to call upon the third speaker for a side proposition | 我接下来请第三位发言者提出一个附带建议。 |
| 542 | Madam president ladies and gentlemen of the house. | 总统女士们,先生们 |
| 543 | Thank you for inviting me to this historic | 谢谢你邀请我参加这个历史性的 |
| 544 | debate about | 辩论 |
| 545 | One of the most important questions of our time | 我们时代最重要的问题之一 |
| 546 | The american writer mark twain once said | 美国作家马克·吐温曾经说过 |
| 547 | committees are dark alleys down which ideas are led | 各委员会是领导思想的黑暗巷子 |
| 548 | and bludgeon | 和块棒 |
| 549 | Our opponents have proposed several dark alleys | 我们的对手提议了几条黑巷子 |
| 550 | My goal is to shed some light on them | 我的目标是让他们明白 |
| 551 | I want to I want to do three things clarify what are what we're claiming clarify the main challenges to our presentation | 我想做三件事 澄清我们所说的是什么 澄清我们陈述的主要挑战 |
| 552 | And affirm why agi will be the last great invention mankind ever makes | 并申明为什么阿吉将是人类最后的伟大发明 |
| 553 | First of all, what do we say? | 首先,我们该怎么说? |
| 554 | What are we saying when we say the last invention? | 我们说最后的发明是什么? |
| 555 | The last great invention | 最后的伟大发明 |
| 556 | We're not saying nothing will ever be built | 我们不是说什么也不会建成 |
| 557 | By humans, we're not saying that humans will never improve anything | 我们不是说人类永远不会改进任何东西 |
| 558 | we're not saying that um | 我们不是说,嗯 |
| 559 | Robots with ai for brains will chase us around until we die | 拥有大脑的机器人会追逐我们直到我们死去 |
| 560 | Although of course they will | 当然了 |
| 561 | Our claim is narrower | 我们的说法比较狭隘 |
| 562 | It's that agi is the last great invention and this is important where humans are the primary inventors | 是阿吉是最后的伟大发明 人类是主要发明者 这一点很重要 |
| 563 | after that the main engine of | 之后的主要引擎为: |
| 564 | Major breakthroughs becomes agi once you build the inventor | 发明者一旦造就 重大突破就会变成愤怒 |
| 565 | It does most of the inventing | 大部分发明都是这样 |
| 566 | Now what is a great invention my colleague anders went through that a little bit. | 现在,我的同事和人们经历了一点伟大的发明。 |
| 567 | It's the kind that changes civilization | 是那种改变文明的人 |
| 568 | It's a pivoting point | 这是一个支点 |
| 569 | Technical progress either accelerates | 技术进步或加快 |
| 570 | Radically or it takes a sharp turn | 绝对的,否则就要转弯了 |
| 571 | Dramatically and it throws everything a little bit off | 从戏剧角度来说,它把一切都抛了一点 |
| 572 | Fire was a great invention because it allowed us to digest more of food | 火是一个伟大的发明 因为它让我们能消化更多的食物 |
| 573 | And it it made digestion easier | 这让消化更容易 |
| 574 | So it freed up a lot of calories to give us our greatest innovation and that was our big brains | 所以它释放出很多卡路里 给我们最大的创新 那就是我们的大脑 |
| 575 | Our big brains that invent things | 我们发明事物的大大脑 |
| 576 | Farming created surpluses and allowed us to take jobs that weren't farmers | 农耕创造了余粮,让我们能找到不是农民的工作 |
| 577 | It allowed the rise of the rise of cities it arrived it really contributed to civilization | 它让城市的兴起 它的到来真正促进了文明 |
| 578 | The printing press spread knowledge drove literacy and scientific revolution | 印刷机传播知识,推动扫盲和科学革命 |
| 579 | AGI belongs here | AGI属于这里 |
| 580 | It won't just help us think it'll do much of the thinking | 它不会只是帮助我们 认为它会做很多的想法 |
| 581 | It won't be a mere tool. | 这不仅仅是个工具 |
| 582 | It will be the inventor of tools | 会是工具的发明者 |
| 583 | until now | 直到现在 |
| 584 | Intelligence has had to pass through the narrow human bottleneck the narrow bottleneck of human cognition | 情报部门不得不穿过人类认知的狭窄瓶颈 |
| 585 | Let me say that again | 让我再说一遍 |
| 586 | Intelligence has had to pass through the narrow bottleneck of human cognition with a gi | 情报部门不得不通过 人类认知的狭窄瓶颈 |
| 587 | Intelligence will step around that bottleneck and be liberated | 情报部门会绕过这个瓶颈 获得解放 |
| 588 | What does invention require? | 发明需要什么? |
| 589 | It's not quite as complicated as the opposition has said | 并不像反对派说的那么复杂 |
| 590 | invention requires | 发明需要 |
| 591 | generating ideas | 产生想法 |
| 592 | testing hypotheses | 测试假设 |
| 593 | learning from your mistakes and and moving forward into uh into progress | 从你的错误中吸取教训,然后走向进步 |
| 594 | If if a machine does this faster and better it will become the center of invention | 如果机器做得更快,更好 它将成为发明的中心 |
| 595 | Consider the differences humans take months or years for experiments | 考虑不同 人类需要数月或数年的实验 |
| 596 | AI could run thousands of experiments | AI可以运行上千个实验 |
| 597 | simultaneously | 同时 |
| 598 | Humans specialize humans specialize in one or two fields | 人类专门研究一两个领域 |
| 599 | AI or a gi could specialize in several fields at once | AI或Gi可以同时专攻几个领域 |
| 600 | Humans get tired and must rest | 人类累了,必须休息 |
| 601 | AGI must be maintained just at just a tiny fraction of the time | AGI必须维持在时间的一小部分 |
| 602 | Most importantly, perhaps a gi could improve itself designing better software and hardware | 最重要的是,Gi也许可以改进自己设计更好的软件和硬件 |
| 603 | Better systems create even better systems | 更好的系统创造了更好的系统 |
| 604 | Creating a cycle of recursive self-improvement | 创造循环自我改进 |
| 605 | This could lead to what we call the intelligence explosion | 这会导致我们所谓的情报爆炸 |
| 606 | But we don't even need unlimited improvement just enough for a gi for this proposition to be true just enough for a gi to be | 但是我们甚至不需要无限的改进 足够让这个提议成真 |
| 607 | Clearly to clearly exceed human capabilities across most domains AI only needs to become clearly better than human thinking | 很明显地,在大多数领域, 人的能力都超过人的能力。 |
| 608 | Which it already is in narrow domains | 它已经在狭窄的域中 |
| 609 | For one example of how the difference in in abilities | 举一个例子说明能力的差异 |
| 610 | In your brain a signal a signal moves at about 300 feet per second | 在你的大脑中,一个信号 移动在大约300英尺每秒 |
| 611 | From out coming out from a processor | 从处理器里出来 |
| 612 | an electronics transmission | 电子传输 |
| 613 | Travels at 500 million feet per second | 每秒旅行5亿英尺 |
| 614 | So let me address a couple of challenges to our position | 让我谈谈我们的立场面临的一些挑战 |
| 615 | It's too uncertain. | 这太不确定了 |
| 616 | You can't know if a gi will exist with some of our | 你不可能知道,如果一个Gi会存在 与我们的一些 |
| 617 | Opponents have agreed to the proposition actually assumes a gi exists | 反对者同意了这个提议 假设一个基存在 |
| 618 | It asks about the nature of invention after that point | 它问发明的性质 在那之后 |
| 619 | It doesn't really debate whether a gi will be possible it asks | 它不会真正争论 是否一个Gi是可能的 它问 |
| 620 | Will it be humanity's last great invention? | 这是人类最后的伟大发明吗? |
| 621 | But here's why a gi will likely exist | 但这就是为什么一个基会存在的原因 |
| 622 | In the last decade bigger models plus more data and better training have repeatedly produced unexpected | 在过去十年中,更大的模型加上更多的数据和更好的培训一再产生出意料之外的结果。 |
| 623 | general abilities | 一般能力 |
| 624 | You don't need a miracle. | 你不需要奇迹 |
| 625 | You don't even need | 你甚至不需要我帮忙 |
| 626 | The intelligence explosion you just need continued iteration on methods already producing broader | 情报爆炸,你只需要 继续重复的方法 已经产生了更广泛的 |
| 627 | capabilities broader competence | 扩大能力 |
| 628 | Unless there's some major defeater like a civilization collapse or a asteroid strike progress will lead to a gi you could call this | 除非有文明崩溃或小行星撞击的失败者 会导致你称之为"Gi" |
| 629 | compounding iteration | 复合迭代 |
| 630 | If a gi can exist it will be deployed | 如果一个基能存在,它将被部署 |
| 631 | businesses will | 企业会 |
| 632 | Use it to stay competitive with other businesses nations will use it to provide security against other nations | 利用它与其他企业保持竞争力 国家会利用它来提供对其他国家的安全 |
| 633 | Laboratories will use it to stay ahead with discovery | 实验室会用它来追踪发现 |
| 634 | If you still have doubts about whether a gi exists consider open ai and anthropics most identical almost identical definitions of a gi | 如果您仍然怀疑一个基是否存在 考虑开放的AI和人类学 最相同的 几乎相同的定义一个基 |
| 635 | Containing the corporation's defining documents it says they both say | 上面写着公司的定义文件 |
| 636 | A gi is quote a highly alternate autonomous system that outperforms human at most economically valuable work | 一个基就是引申出一种高度交替的自主系统,它表现优于人类最有经济价值的工作 |
| 637 | So what they've identified and why they're pouring billions of dollars into it is that in the very near future | 因此,他们已经确定 和为什么 他们向它注入数十亿美元 是在最近的将来 |
| 638 | You you'll be able to replace humans at economically valuable work and that's exactly what they're betting on | 你会在有经济价值的工作上取代人类 而这正是他们所相信的 |
| 639 | That's what they're betting billions and billions of dollars on | 他们就是这么赌上几十亿的 |
| 640 | They see in the future this giant opportunity opening up and what it is is a gi | 他们将来会看到这个巨大的机会 打开它是什么 |
| 641 | They know it's coming | 他们知道会来 |
| 642 | a lot of | 许多 |
| 643 | They know it's coming | 他们知道会来 |
| 644 | more information about | 更多信息,请访问 |
| 645 | What? | 你说什么? |
| 646 | Oh, sorry, go ahead | 对不起,请说 |
| 647 | Could it be that they're pouring money simply to hide the fact that they're not making as much? | 难道他们只是为掩盖自己没有赚到那么多钱吗? |
| 648 | Not make as much what? | 没赚多少? |
| 649 | They're pouring money into buying assets to hide the fact | 他们把钱倒进买资产来掩盖事实 |
| 650 | I you know, I I just don't think they're they're uh, I don't think they would be that obvious. | 你知道,我只是不认为 他们是呃, 我不认为他们会是那么明显。 |
| 651 | I think that | 我觉得 |
| 652 | Um | 垛 |
| 653 | They're trying to get ahead of obviously they're trying to pour money into a market that's coming up in the in the near future | 他们试图超越 显然他们试图把钱 倒到一个市场 在不远的将来 |
| 654 | They're uh, the billions they spend and how much they're spending on data centers | 他们是呃,他们花费的亿 和他们花费多少 在数据中心 |
| 655 | the exponential amount amounts only makes sense if you believe in the uh dramatic invention | 只有相信戏剧性的发明 指数数量才有意义 |
| 656 | possibilities of a gi | A. 放弃的可能性 |
| 657 | That's when they make sense if you follow the trajectory of the exponential, uh increases in in valuation. | 这时候,如果你遵循指数的轨迹, 呃,估值的增加。 |
| 658 | It's because | 因为 |
| 659 | The demand and the products are going to be astronomical the most most uh, a gi will be the most important invention ever made | 需求与产品将是天文上最强的,呃,一个基将是有史以来最重要的发明 |
| 660 | imagine taking 20 phd college a thousand phd quality brains and trading on things like | 想象一下,进入20个phd大学 一千个phd质量的大脑 和交易的东西喜欢 |
| 661 | Climate warming turning on things like cancer | 气候变暖导致癌症 |
| 662 | You can see the value of that of that product and that's what they're betting on | 你可以看到那些产品的价值, 这就是他们所赌的 |
| 663 | The second challenge i want to look at is humans, uh still choose goals. | 我想看的第二个挑战是人类, 嗯,仍然选择目标。 |
| 664 | So invention remains human | 所以发明还是人类 |
| 665 | This confuses two separate jobs. | 这混淆了两个不同的工作. |
| 666 | One job is deciding what we want. | 一个工作是决定我们想要什么。 |
| 667 | The other is creating a new method method that achieves it | 另一种是创建一种新的方法来实现它 |
| 668 | consider alpha fold | 考虑字母折叠 |
| 669 | Dimitris Abbas and john jumper of google deep mind built alpha fold to predict a protein's 3d structure based on its amino acids | Dimitris Abbas和John跳跃的Google深智构建了α折叠,以预测一个蛋白质的3d结构基于它的氨基酸 |
| 670 | now once they | 现在一旦他们 |
| 671 | Which is crucial for understanding things like neurodegenerative diseases like alzheimer's once they did that their roles switched they | 这对了解神经退化疾病 诸如阿尔兹海默的病症 至关重要 因为他们的角色被改变 |
| 672 | Uh | 对 |
| 673 | The the first inventors proposed the protein station that alpha fold should look at then alpha fold took over | 最早的发明家提出 蛋白质站,阿尔法折叠应该看 然后阿尔法折叠接管 |
| 674 | Generating structural hypotheses rapidly at scale so the jobs changed | 快速生成结构假设, 从而改变工作 |
| 675 | Now the human has become the director. | 现在人类变成了导演 |
| 676 | It says go look at that | 上面写着去看一下 |
| 677 | But the but the force of invention is on the cognitive architecture on the a gi | 但是,发明的力量 是在认知结构上 在一个基 |
| 678 | um | 垛 |
| 679 | In yeah, it's just like a gi but in a very narrow domain | 是的,它只是像一个基 但在一个非常狭窄的域 |
| 680 | So if a gi can produce designs hypotheses and solutions better and faster than humans across all domains | 因此,如果一个基可以产生 设计假设和解决方案 比人类在所有领域更好和更快 |
| 681 | It becomes the primary source of or most domains. | 它成为大部分域的主要来源。 |
| 682 | It becomes the primary source of new invention | 它成为新发明的主要来源 |
| 683 | Making it the last great invention humans ever need to make | 让它成为人类最后的发明 |
| 684 | Thank you very much. | 谢谢 |
| 685 | Next. | 下一个 |
| 686 | I would like to call upon the third speaker for side opposition | 我恳请第三位发言的一方反对。 |
| 687 | president | 庭长 |
| 688 | office of the house | 办公室 |
| 689 | Ladies and gentlemen, it's an absolute pleasure to be here to discuss one of the most important topics over the last 10 years | 女士们先生们 很高兴能来讨论 过去十年来最重要的话题之一 |
| 690 | Let's take on a journey | 让我们去旅行吧 |
| 691 | Let's go back a few years | 我们回去几年吧 |
| 692 | charles darwin | 查尔斯·达文 |
| 693 | You've all heard of him | 你们都听说过他 |
| 694 | He described this theory of evolution | 他描述了这个进化论 |
| 695 | he described | 他描述 |
| 696 | incremental changes | 增量变化 |
| 697 | Over thousands upon years if not millions of years where humanity evolved slowly but incrementally | 在数千年中,即使不是数百万年,人类在其中缓慢但渐进地发展 |
| 698 | It's a theory | 这是理论 |
| 699 | It's been obviously | 很明显 |
| 700 | Assessed and people also questioned it. | 评估,人们也对此提出质疑。 |
| 701 | Let's go to 1970 | 我们到1970年吧 |
| 702 | Stephen gould | 斯蒂芬·古尔德 |
| 703 | an evolution biologist | 进化生物学家 |
| 704 | Kept with another concept which he called | 坚持他所说的另一个概念 |
| 705 | punctuated equilibrium | 圆点平衡 |
| 706 | I'll repeat it punctuated equilibrium. | 我重复一遍 尖锐的平衡 |
| 707 | He describes humanity | 他描述人类 |
| 708 | Going slowly slowly similar to charles darwin's theory, but every now and then something changed in our trajectory | 慢慢地与查理斯·达温的理论相似 但偶尔我们的轨迹会有所改变 |
| 709 | Changing our direction forever | 永远改变方向 |
| 710 | Inflection points | 插点 |
| 711 | He described this concept. | 他描述了这一概念。 |
| 712 | For example, we move along slowly, but then suddenly we change our direction. | 例如,我们缓慢前进,但突然改变方向。 |
| 713 | Let's think about it fire | 让我们想想,火 |
| 714 | electricity | 电 |
| 715 | the motor car | 汽车 |
| 716 | The airplane the printing press the computer | 飞机 印刷机 电脑 |
| 717 | And now ai each of those moments of our history has changed our direction forever | 我们历史的每一个时刻 都永远改变了方向 |
| 718 | So that's the concept of punctuated equilibrium | 这就是点球平衡的概念 |
| 719 | If you talk about the concept of singularity | 如果你说到奇特的概念 |
| 720 | Which may have heard of singularity is the point at which computers and machines are better than human beings | 电脑和机器比人类好 |
| 721 | They say in 1995 the singularity is near | 据说1995年的奇异性已经接近了 |
| 722 | They say in 2025 the singularity is near | 他们说在2025年 奇点就快到了 |
| 723 | I'm hearing all the time. | 我常闻. |
| 724 | The singularity has been a problem. | 奇异性是一个问题。 |
| 725 | Is it coming now? | 现在来了吗? |
| 726 | 是个 gi | |
| 727 | The tech bros we call the tech bros people run the tech companies the tech bros | 我们称之为技术兄弟的人 管理技术公司 技术兄弟 |
| 728 | They tell us a gi is coming next year the year after may five years down the line lms were mankind's amazing invention | 他们告诉我们明年就要出动了 5年后 人类的惊人发明 |
| 729 | Lms aren't the answer. | 问题不是答案 |
| 730 | They've gone back. | 他们回去了 |
| 731 | So that's not the answer. | 所以这不是答案。 |
| 732 | That's not a gi | 这不是一个基 |
| 733 | They've also put a lot of money to this as we've just mentioned | 正如我们刚才提到的 他们也为此投入了很多钱 |
| 734 | It's called the circular economy | 这叫循环经济 |
| 735 | nvidia | 尼维迪亚 |
| 736 | promised 100 billion into open ai open ai promised 100 billion into oracle oracle | 承诺1000亿进入开放的Ai 开放的Ai 承诺1000亿进入甲骨文 |
| 737 | 一千亿回到呃 进入um nvidia | |
| 738 | Economy that's going to crash | 经济会崩溃 |
| 739 | They say this year actually ai may go bankrupt can imagine and you're right when you're saying that the economy is going to crash | 他们说今年可能会破产 你说经济会崩溃的时候 你是对的 |
| 740 | Mark these words as you go forward | 前进时标出这些字 |
| 741 | So we're talking about this concept of a gi as we go forward. | 所以,我们是在谈论这个概念,我们前进。 |
| 742 | Yes, it's a concept | 是的,这是一个概念 |
| 743 | Is it there? | 在吗? |
| 744 | Is it real? | 是真的吗? |
| 745 | We don't know yet. | 我们还不知道 |
| 746 | This is very superfluous and a nebulous term that we don't know what it means | 这是多余的, 模糊的术语 我们不知道是什么意思 |
| 747 | Let's go back and talk about | 我们回去谈谈 |
| 748 | my area of health care | 我的保健领域 |
| 749 | So radiology is obviously a profession that was ripe for disruption image-based image analysis | 因此,放射学显然是一个专业 已经成熟的破坏图像 图像分析 |
| 750 | Great if you look at feifei leaves work for image net going back. | 如果你看看Feifei的叶子 就能恢复形象网 |
| 751 | It was based around can we decipher images using pixel etc and recognition? | 它基于周围 我们可以破译图像 使用像素等 和识别? |
| 752 | Feifei's just raised now that one billion dollars for her company only a few days ago | 飞飞刚刚为她公司筹了十亿美元 就在几天前 |
| 753 | But radiologists let's go back in 2016 | 但是放射学家们 让我们回到2016年 |
| 754 | 在UCL的Ai Rey提示的教父 | |
| 755 | Said made the statement to the health care profession | 赛义德向保健界做了声明 |
| 756 | Within five years. | 五年内 |
| 757 | We won't need to train a radiologist | 我们不需要训练放射科医生 |
| 758 | in five years | 五年内 |
| 759 | In that five years the mayo clinic in america | 在这五年里,美国蛋黄酱诊所 |
| 760 | Increased radiologists by 55 percent | 放射科医生增加55% |
| 761 | They're 400 agents now in in the hospitals | 医院里有400名特工 |
| 762 | So that concept of saying we're going to be outperforming or not requiring doctors. | 所以这个概念是说,我们将 表现优于或不需要医生。 |
| 763 | For example, clearly was a misnomer | 例如,显然是一个错误的符号 |
| 764 | He came back this year and said actually it's augmentation | 他今年回来,说其实是扩容 |
| 765 | AI and radiology all met each other to make health care more efficient to do more work | AI和放射科都相遇,以提高保健效率,做更多的工作 |
| 766 | To allow for the discrepancy of global health workforce shortages on global level | 使全球保健人员短缺在全球一级出现差异 |
| 767 | AI undoubtedly is transforming our mankind. | 大赦国际无疑正在改变我们的人类。 |
| 768 | It's transforming health care and other verticals immediately. | 它正在立即改变医疗保健和其他纵向。 |
| 769 | We've seen that | 我们已经看到了 |
| 770 | It's not going away. | 不离灭. |
| 771 | The gene is out of the bottle | 基因从瓶子里出来 |
| 772 | But aji | 不过阿吉 |
| 773 | Is another matter | 是另一个问题 |
| 774 | If we look at my own professional surgery | 如果我们看看我自己的专业手术 |
| 775 | I'm a surgeon. | 我是个外科医生 |
| 776 | I've been around for 35 years in the nhs. | 我在这呆了35年了 |
| 777 | I've survived the nhs for that amount of time | 我活了这么久 |
| 778 | I know what you're thinking | 我知道你在想什麽 |
| 779 | He doesn't look old enough to be around for 35 years in health care. | 他看起来不够大 无法在35年的医疗保健。 |
| 780 | I know what you're thinking | 我知道你在想什麽 |
| 781 | But I have i've survived. | 但我已经活下来了 |
| 782 | I think about my own job as a surgeon a cancer specialist for three decades | 我想起我做外科医生三十年 癌症专家的工作 |
| 783 | What do we make decisions in health care? | 我们如何在保健方面作出决定? |
| 784 | Can ai replace this? | 爱能代替这个吗? |
| 785 | Would you like an ai to do your operation agi? | 你想Ai做你的手术吗? |
| 786 | An ai taking out scalpel or a robot to do your operation. | 一个AI取出手术刀或机器人做你的手术。 |
| 787 | Would you be comfortable? | 你会舒服吗? |
| 788 | It's even possible | 这甚至有可能 |
| 789 | Clinicians obviously require a different kind of mindset | 临床医生显然需要不同的思维方式 |
| 790 | It's a nuance of making clinical judgment. | 这是临床判断的细微差别. |
| 791 | It's the experience | 这是经验 |
| 792 | It's understanding when things go wrong how to deal with them. | 这是理解 当事情发生错误 如何处理它们。 |
| 793 | They say that for example the greatest aspect of surgeon | 他们说,例如外科医生的最大方面 |
| 794 | The most difficult thing | 最难的事 |
| 795 | Is when to say no to an operation? | 什么时候对手术说不? |
| 796 | What would you say? | 你会说什么? |
| 797 | Yes, that's easy | 对,很简单 |
| 798 | Will ai agi allow you to understand the concept of humankind the touch the emotion | Ai Agi会允许你理解人类的概念 感受情感 |
| 799 | The physical experience we have with our patients is agi are going to replace that is the last invention | 我们和病人的体能经验 将取代最后的发明 |
| 800 | I don't think so | 我不这么认为 |
| 801 | Humans will be in the loop | 人类将进入循环 |
| 802 | They have to be to maintain our direction to maintain sanity and to ensure that what we're doing is safe for not just patients | 他们必须坚持我们的方向,以保持理智,并确保我们所做的,不仅仅是病人的安全 |
| 803 | but mankind as overall | 但整个人类 |
| 804 | then the um | 礛琌 垛 |
| 805 | Of this conversation was of course and the question they asked is | 他们问的问题是 |
| 806 | Is agi humanity's last great invention? | 是人类最后的伟大发明吗? |
| 807 | What do we buy the last great invention ai undoubtedly powerful in health care | 我们买什么 最后一个伟大的发明 AI无疑在保健方面强大 |
| 808 | Let's take an example. | 举个例子 |
| 809 | They mentioned protein folding earlier | 他们之前提到蛋白质折叠 |
| 810 | Yes, of course when google deep minds team unraveled the mystery of science and biology | 是的,当然,当Google深智团队 解开科学和生物学的奥秘 |
| 811 | Unraveling one of the greatest mysteries of protein folding with alpha fault allows look at protein structure | 解开蛋白质与α断层叠叠的最大奥秘之一 就可以看蛋白质的结构 |
| 812 | What's that allowed us to do? | 那让我们做什么? |
| 813 | Is develop drugs much faster in the past to take a drug from inception to concept | 是过去研发的药物要快得多 服用药物从开始到概念 |
| 814 | Into clinical trials and to treatment it took 10 to 15 years at a cost of 1.5 billion dollars | 进入临床试验和治疗需要10至15年,花费15亿美元 |
| 815 | With ai fast tracking we now can produce drugs in one or two years | 只要快速追踪,一两年内就能生产毒品 |
| 816 | And actually at a cost of a few million | 其实花了几百万 |
| 817 | It democratizes health care. | 它使保健民主化。 |
| 818 | Of course it does | 当然 |
| 819 | What if the invention | 如果发明 |
| 820 | Great invention was a cure for cancer with ai driven protocols ai drug discovery tomorrow | 伟大的发明是治疗癌症的药 由Ai驱动的协议 AI药物发现明天 |
| 821 | Would that be a better invention than agi? | 那比阿吉的发明更好吗? |
| 822 | It helps humanity | 帮助人类 |
| 823 | What about if we think about our surgery? | 如果我们考虑我们的手术呢? |
| 824 | ai driven surgical practices navigation tools helping us become better surgeons or better doctors | 帮助我们成为更好的外科医生或更好的医生 |
| 825 | Having diagnostics by our side allowing us to diagnose pretty much faster | 我们身边有诊断 能够更快的诊断 |
| 826 | Is that better? | 这样好点了吗? |
| 827 | As we destroy our planet with climate change and everything else | 当我们用气候变化和一切毁灭地球 |
| 828 | We are going to be supposedly an interplanetary species. | 我们将会是行星际物种 |
| 829 | We are going to the moon | 我们要去月球 |
| 830 | Or mars according to musk | 或根据麝香 |
| 831 | If we are destroying our own planet | 如果我们在毁灭自己的星球 |
| 832 | And we move to the moon and colonize the moon of mars | 我们移动到月亮 殖民火星的月亮 |
| 833 | Is that humanity's last great invention or is the agi? | 这是人类最后的伟大发明 还是爱奇艺? |
| 834 | So we are on the precipice of change. | 因此,我们正处在变革的边缘。 |
| 835 | There's no question | ⊿拜肈 |
| 836 | Ai is giving us alternatives. | 艾给了我们其他选择 |
| 837 | It makes our lives a lot easier. | 这使我们的生活变得容易得多。 |
| 838 | It's automated our work | 我们的工作是自动化的 |
| 839 | It's made health care much more democratized. | 这让医疗保健更加民主化. |
| 840 | It helps understand our patients better. | 这有助于我们更了解病人 |
| 841 | There's no question. | 无问. |
| 842 | It's a huge advantage | 这是一个巨大的优势 |
| 843 | The question i'll ask you in the audience, of course is the question around the last great invention | 我在观众中问你的问题 当然是最后一个伟大的发明 |
| 844 | Is it like the human's last great invention or humanity's great invention? | 这是人类最后的伟大发明还是人类的伟大发明? |
| 845 | Or is it simply | 还是简单的 |
| 846 | Um what I would call it from the beginning you mentioned we spoke earlier | 嗯,我怎么称呼它 从一开始 你提到我们之前说过 |
| 847 | Is it merely just a punctuation? | 难道只是点缀而已? |
| 848 | Of our own equilibrium. | 我们自己的平衡。 |
| 849 | Thank you very much | 非常感谢 |
| 850 | Next I would like to call upon the fourth speaker for a side proposition | 下面我要请第四位发言者提出一个附带建议。 |
| 851 | Madam president, uh, ladies and gentlemen distinguished guests | 总统女士 女士们先生们 尊敬的客人们 |
| 852 | It may not be obvious in this | 这也许不明显 |
| 853 | glorious victorian debating chamber | 荣耀的胜利者辩论室 |
| 854 | The university of october is | 奥克托伯大学是 |
| 855 | That I have spent much of the last 15 years much of my working life building that academic center up | 过去15年中我度过了 大部分工作生活 建设了学术中心 |
| 856 | So let's talk about what some of the ai researchers | 所以,让我们谈谈一些AI的研究者 |
| 857 | What gets them out of bed in the morning since the subject was raised? | 是什么让他们在早上下床 从主题被提出? |
| 858 | What gets them out of bed in the morning is making a better world | 让他们早上下床的 是创造更美好的世界 |
| 859 | Let's talk about what some of them do | 让我们谈谈他们中有些人做了什么 |
| 860 | We have ai researchers in oxford who are developing techniques that will enable automatic analysis | 我们的Ai研究人员在牛福德 他们正在开发技术 能够自动分析 |
| 861 | Of heart scans and be able to detect abnormalities on those heart scans to be able to detect heart disease | 在心脏扫描中,能够检测这些心脏扫描中的异常情况,以便能够检测心脏病 |
| 862 | And get that technology out all over the world at virtually zero cost | 以零成本把技术推广到世界各地 |
| 863 | That technology has the promise to utterly transform | 这种技术有潜力彻底改造 |
| 864 | Cardiac the provision of cardiac health care across the globe. | 心脏在全球范围提供心脏保健。 |
| 865 | It's incredible work | 这是不可思议的工作 |
| 866 | We have researchers working on doing the automatic analysis of fetal brain scans to detect | 我们有研究人员在研究 胎儿脑扫描的自动分析 以检测 |
| 867 | Abnormalities in fetal brain scans long before a human being would be able to detect them. | 胎儿脑部的异常扫描 早在人类能够发现它们之前 |
| 868 | It's incredible work. | 不可思议功德. |
| 869 | It's good work | 干得不错 |
| 870 | We have researchers who are working on ai for developing drugs so that when there is another pandemic or whatever variant it takes | 我们的研究人员正在研究开发药物的Ai, 以便当有另一种流行病或任何改变时 |
| 871 | We'd be able to roll out drugs safely | 我们可以安全地推出毒品 |
| 872 | {\fn黑体\fs22\bord1\shad0\3aHBE\4aH00\fscx67\fscy66\2cHFFFFFF\3cH808080}那么我们就不会再做上一次大流行的噩梦了 | |
| 873 | We have researchers working on robotics | 我们有研究机器人的研究人员 |
| 874 | Assistive robotics so that elderly people can continue to live | 辅助机器人,使老年人能够继续生活 |
| 875 | Independent lives long beyond a point at which it would have been possible on their own | 独立生活远远超出自己能够实现的阶段 |
| 876 | We have AI researchers that are building robots that can go into disaster areas and explore and search for survivors and help those survivors. | 我们的AI研究人员正在建造机器人,可以进入灾区,探索和搜寻幸存者,帮助幸存者。 |
| 877 | It's incredible work. | 不可思议功德. |
| 878 | It's good work. | 功德圆满. |
| 879 | That is what gets AI researchers out of bed in the morning. | 这就是AI研究人员早上起床的原因。 |
| 880 | And to besmirch them and to attribute them the kind of motives that they were attributed to is insulting. | 嘲笑他们,并说出他们的动机是侮辱性的。 |
| 881 | Frankly, and it's patronizing to do that in the pursuit of a few childish debating point demeans this house. | 坦率地说,它很支持这样做, 追求一些幼稚的辩论点 贬低了这个房子。 |
| 882 | So that's what AI researchers are typically concerned with. | 这就是AI研究者通常关心的问题。 |
| 883 | Let's talk about me since the subject was raised and I was accused of being pursuing an AGI agenda. | 我们来谈谈我 自从提出这个话题, 我被指控追求一个AGI议程。 |
| 884 | Now, the remarkable thing about this is that even a modicum of research would have demonstrated that I am very publicly on record as being an AGI skeptic. | 现在,关于这一点的显著之处是,即使有一点研究, 也会证明我作为一个AGI的怀疑者, 被公开记录。 |
| 885 | A modicum of research apparently was not done in the preparation for this debate. | 在这次辩论的筹备工作中,显然没有做一点研究。 |
| 886 | As evidence of this, I would point out that one of the chapters in my 2020 book is in fact called The Singularity is Bullshit. | 作为这方面的证据,我要指出,我的2020年书中有一个章节实际上叫做"Singularity"是"废话". |
| 887 | And to add to that, I would point out that, and you are very welcome to do this immediately after the debate, | 此外,我要指出,辩论结束后,你们可以立即这样做。 |
| 888 | 我做了一个YouTube的采访 播客的同名标题, "Singularity is bullshit", 已经观看超过30万次。 | |
| 889 | The interviewer for that was extremely good. | 采访者对此极好。 |
| 890 | He'd done some research. | 他已经做了一些研究。 |
| 891 | Let's talk about AGI. | 我们来谈谈AGI |
| 892 | What really gets AI researchers out of bed in the morning is not AGI. | 真正让AI研究人员早上起床的不是AGI. |
| 893 | They're focused on the kind of things I was talking about. | 他们专注于我所说的那种事情。 |
| 894 | But the advances that we've seen in AI generally have brought AGI back into focus and made it the center of attention. | 但是我们在AI中看到的进展一般地将AGI重新集中起来,使其成为关注的中心. |
| 895 | That's why we're having this debate today. | 故今此论. |
| 896 | So what is AGI? | AGI是什么来着? |
| 897 | The problem is it's a very, very difficult subject to nail down. | 问题是它是一个非常,非常困难的 标的下来。 |
| 898 | It's kind of like art, right? | 这有点像艺术,对不对? |
| 899 | I mean, I know it when I see it and you know it when you see it. | 我的意思是,我知道它 当我看到它,你知道它 当你看到它。 |
| 900 | But if we try and agree on what it is, we're probably not going to reach an agreement. | 但如果我们试图就它是什么达成一致,我们可能不会达成协议. |
| 901 | So let's think about some of the varieties of AGI that we might contemplate. | 让我们来考虑一下AGI的一些品种。 |
| 902 | One variety of AGI is kind of chat GPT that just makes fewer mistakes. | 一种种类的AGI是一种聊天的GPT,只是犯较少的错误. |
| 903 | That would be a kind of a version of AGI that one could think about. | 这是一种AGI的版本,人们可以考虑。 |
| 904 | Now, the interesting thing about this is the most high profile AI product release of last year was GPT-5. | 有趣的是,去年最引人注目的AI产品发布是GPT-5. |
| 905 | And a lot of people seem to expect that that was going to be AGI. | 很多人似乎都认为那是AGI。 |
| 906 | I didn't, I have to say. | 我没有,我不得不说。 |
| 907 | And again, I'm very publicly on record as saying that. | 再说一遍,我公开公开说 |
| 908 | And if you'd asked me in advance what is the one thing I would have asked for in that product, | 如果你事先问我 产品中我要求的是什么 |
| 909 | it was just don't make so many bloody mistakes. | 只是别犯太多错误 |
| 910 | Don't enthusiastically come back and tell me things which are manifestly not true. | 别急着回来告诉我那些明显不真实的事 |
| 911 | That problem did not get fixed. | 这个问题没有得到解决。 |
| 912 | Despite the hundreds of millions, billions of dollars that were spent on that product, | 尽管在产品上花费了数亿,数十亿美元, |
| 913 | it still continually makes mistakes. | 它仍然不断犯错误。 |
| 914 | So even that very, very unambitious, seemingly unambitious version of AGI turns out to be a lot more difficult. | 因此,即使这个非常,非常不雄心勃勃的,似乎并不雄心勃勃的AGI版本也证明是困难得多的. |
| 915 | And then there's a more widely discussed version of AGI. | 之后还有较广泛的讨论版本AGI. |
| 916 | And that more widely discussed version of AGI is AI which is capable of any intellectual task that a human being is capable of. | 而较广泛讨论的AGI版本是AI,它能够完成一个人能够完成的任何智力任务. |
| 917 | Well, that's clearly further into the future. | 嗯,这显然更进一步的未来。 |
| 918 | But let me just point out a really, really big barrier. | 但我要指出一个非常非常大的障碍。 |
| 919 | I always took AGI as meaning AI that could do roughly speaking anything that a human being would be able to do | 我总是认为AGI意味着 AI,可以做 粗略的说任何事情 人可以做的 |
| 920 | with no more training than a human being would require in order to be able to do that. | 训练不足于人 才能做到这一点。 |
| 921 | And there is one big barrier, one big obstacle that we still haven't broken through yet in AI. | 还有一个大障碍,一个大障碍 我们在AI还没有突破。 |
| 922 | And that is the physical world. | 这就是物理世界。 |
| 923 | So let me illustrate the point as follows. | 因此,让我说明这一点如下。 |
| 924 | We have AI that you can talk about the history of the Roman Empire too. | 我们有AI,你可以谈论 罗马帝国的历史。 |
| 925 | And it will be confident, it will be fluent, it will make mistakes, but nevertheless it will be impressive. | 它会充满自信,会流利,会犯错误,但会给人留下深刻印象。 |
| 926 | And then you can flip immediately and start talking about quantum mechanics. | 然后你可以立即翻转,开始谈论量子力学. |
| 927 | And again, it will be fluent and confident. | 再说一遍,它将是流利和自信的。 |
| 928 | How many human beings could do that? | 有多少人可以这样做? |
| 929 | You can then immediately talk about quantum mechanics in Latin. | 然后可以马上用拉丁语谈论量子力学. |
| 930 | I've tried this and because we're in Oxford, I've actually given the text to a Latin expert and they said it was pretty good. | 我试过了,由于我们在牛津, 我实际上把短信给了一位拉丁裔专家, 他们说它很好。 |
| 931 | That's incredible, right? | 这是不可思议的,对不对? |
| 932 | That's an amazing intellectual feat, it seems like. | 这是一个惊人的智力成就, 它似乎是。 |
| 933 | And yet we don't have AI that could come into your home that it had never seen before, | 但是我们没有它从未见过的 进入你家的AI |
| 934 | locate the kitchen, clear the dinner table and load up the dishwasher. | 找到厨房 清扫餐桌 装上洗碗机 |
| 935 | Now that's firstly the most obvious thing to say. | 现在,这是第一个最明显的说法。 |
| 936 | So that's not a task which we would associate with intelligence, is it? | 所以,这不是一个任务 我们将与情报,是吗? |
| 937 | I don't think. | 我不觉得 |
| 938 | Actually, that's something that a minimum wage human worker could do, right? | 其实,这是一个最低工资 人类工人可以做的,对不对? |
| 939 | And this is not in any way to diminish minimum wage human workers. | 这绝不是降低最低工资标准。 |
| 940 | The point is that they are doing something which is much, much, much more complicated than it looks. | 重点是他们正在做一些比看起来复杂得多的事情。 |
| 941 | An AGI that could do anything that a minimum wage human worker can do, | 一个AGI,可以做任何事情 最低工资的人可以做, |
| 942 | and I offer that as a plausible definition of what AGI is, would be utterly, utterly transformative. | 我提出,作为一个合理的定义 AGI是什么, 将是完全,彻底的变革。 |
| 943 | But we don't have it yet and we are, I think, some distance from having it. | 但是,我们还没有拥有它,我想,我们距离拥有它还有一段距离。 |
| 944 | AI in the physical world is a difficult, difficult challenge. | 在物理世界中,AI是一个困难,困难的挑战. |
| 945 | But if we had that, it would, I think, be utterly, utterly transformative. | 但是,如果我们有, 我认为,它将是彻底的,彻底的变革。 |
| 946 | The discussion around AGI is difficult for an awful lot of reasons | AGI的讨论很困难 原因很多 |
| 947 | and really it is difficult for one reason because it is so hard for us to agree on what it really means | 实在很难 原因之一 因为我们很难 就它的真正含义达成一致 |
| 948 | and I think it is so hard for us to agree on what it really means because human intelligence is so multifaceted. | 我认为我们很难就它的真正含义达成一致, 因为人类的智力是如此的多元。 |
| 949 | And one of the reasons that I love AI as a subject is that it shines a light on that | 其中一个原因,我爱AI作为一个主题 因为它照亮了这一点。 |
| 950 | and helps me to understand just how rich and just how multifaceted human intelligence is | 帮助我了解人类的智慧是多么丰富 多么多元 |
| 951 | and how incredible we are as creatures on planet Earth. | 作为地球上的生物 我们是多么不可思议 |
| 952 | Let me wrap up with one final comment. | 最后,我要谈最后一点。 |
| 953 | The debate around AGI is often very difficult, but let me offer the following observation. | 围绕全球倡议的辩论往往非常困难,但让我提出以下看法。 |
| 954 | An awful lot of the debate around AGI is framed in such a way that if you replaced AGI with a magic wand, | 围绕AGI的很多辩论 被设计成这样 如果你用魔杖取代AGI |
| 955 | it would still make sense. | 那还是有道理的 |
| 956 | If we had a magic wand, that would be the last invention that we would ever need. | 如果我们有魔杖,那是我们最需要的发明。 |
| 957 | If we had a magic wand, then there would be an end to world hunger. | 如果我们有魔杖,那么世界饥饿就会结束。 |
| 958 | If we had a magic wand, there would be no potholes on the streets in Oxford. | 如果我们有魔杖,牛津的街道上就不会有坑洞. |
| 959 | That would be useful AI. | 这将是有用的AI。 |
| 960 | But we don't have a magic wand and we don't have AGI. | 但我们没有魔杖,也没有AGI. |
| 961 | So it remains a holy grail, a tantalising holy grail, but not one that excites me | 所以它仍然是圣洁的, 令人愉悦的圣洁, 但不是让我兴奋的 |
| 962 | and not one that excites a very large proportion of the AI community. | 而不是一个能激怒大部分AI社区的人 |
| 963 | If we had it, if we had a magic wand though, think how transformative that would be. | 如果我们有它,如果我们有魔杖, 想想那会如何变革。 |
| 964 | Is anybody else so confused as to what AGI is after all of that? | 还有谁对AGI是什么感到困惑吗? |
| 965 | Honestly, at this point, I feel like the G has to stand for gosh golly. | 老实说,现在我觉得G必须代表上帝 |
| 966 | I'll just make up the most complicated definition I can. | 我会编一个最复杂的定义 |
| 967 | And then they have to vote for us because everybody will think they're dumb if they don't. | 然后,他们必须投给我们 因为每个人都会认为 如果他们不这样做是愚蠢的。 |
| 968 | In all seriousness, the speakers on the proposition are doing incredible work. | 说实在的,提案的发言者正在做令人难以置信的工作。 |
| 969 | They really are, especially in AI alignment and AI safety. | 它们确实如此,特别是在AI的校正和AI的安全方面. |
| 970 | I applaud that work and we all should. | 我赞扬这项工作,我们大家都应该这样做。 |
| 971 | But in this debate and in this motion, they're basically asking for your faith in a doomsday prophecy. | 但是在这场辩论和动议中 他们基本上要求你相信一个末日的预言 |
| 972 | Really one of two doomsday prophecies. | 真正的两个末日预言之一。 |
| 973 | One doomsday prophecy where we just stop inventing one day, where all future human invention is redundant | 一个末日预言,我们只是停止发明 有一天,所有未来的人类发明都是多余的 |
| 974 | and we have this mass loss of meaning. | 而我们却失去了意义 |
| 975 | And another doomsday prophecy where, I don't know, somehow AGI through some unknown process | 另一个末日的预言,我不知道, 不知何故,AGI 通过一些未知的过程 |
| 976 | results in all of our deaths attacked by a bear. | 导致我们所有的死亡 被熊袭击。 |
| 977 | If anyone builds it, everyone dies. | 如果有人建造它,每个人都会死。 |
| 978 | Okay, let's get back to life, back to reality. | 好吧,让我们回到生活,回到现实。 |
| 979 | To believe in a story about the end of the world, you have to ignore what's happening in the real one. | 要相信一个关于世界末日的故事,就必须忽略真实中发生的事情. |
| 980 | In the real world, a billion people are using AI every day, every day people. | 现实世界中,每天,每天有十亿人使用AI. |
| 981 | And what are they doing with it? | 他们拿它做什么? |
| 982 | They're using it to learn and they're using it to invent. | 他们利用它来学习,并用它来发明. |
| 983 | They're using it to bridge skill gaps, skill gaps between themselves and an elite intellectual and commercial class | 他们利用它来弥补技能差距 和他们与精英知识分子和商业阶层之间的技能差距 |
| 984 | that seems to think they have a monopoly on intelligence and creativity. | 似乎认为他们垄断了智慧和创造力 |
| 985 | AI is fundamentally equalizing. | 大赦国际在本质上是平等的。 |
| 986 | It enables even the least resourced among us to learn, invent and achieve. | 它使我们中资源最少的人能够学习、发明和实现。 |
| 987 | But I guess we're not here to debate AI, right? | 但我想我们不是来辩论AI的 对吧? |
| 988 | 这是AGI。 | |
| 989 | 我想G代表了Gimmick的风格 | |
| 990 | And AGI, it may be the final frontier, but it's made in a Hollywood basement. | 和AGI,这可能是最后的前沿, 但它是在好莱坞地下室制造的。 |
| 991 | This version of evil super intelligence you keep reading about in articles, it isn't real. | 这个版本的邪恶的超级智能 你一直在读的文章, 它不是真的。 |
| 992 | It's science fiction. | 科幻小说. |
| 993 | In the movies, AI wakes up and just starts killing. | 在电影中,AI醒来后开始杀人. |
| 994 | But in the real world, we build technology iteratively, step by step, cognizant of the threats, | 但在现实世界中,我们反复地,一步一步地建设技术,意识到威胁, |
| 995 | including the threats studied and brought up by the proposition speakers. | 包括提议发言者研究和提出的威胁。 |
| 996 | And we correct problems as they emerge, however serious those problems may be. | 我们纠正出现的问题,不管这些问题多么严重。 |
| 997 | AGI will not arrive as a fully formed killer robot. | AGI不会作为一个完全成型的杀手机器人而来. |
| 998 | AGI will arrive as a misshapen clump of clay, imperfect, messy and crafted by anxious human hands. | AGI会像一团泥土 完美 混乱 由焦虑的人类手来设计 |
| 999 | And since people will form that clay into useful shapes and people will invent new ones, | 由于人们会把粘土形成有用的形状 人们会发明新的形状, |
| 1000 | then who is the inventor we're actually talking about tonight? | 那么谁是发明家 我们真正谈论今晚? |
| 1001 | Because it's not the clay. | 因为它不是粘土。 |
| 1002 | Semantic wrangling aside, last great invention is terminal in scope. | 语义争论不休,最后的伟大发明是范围上的终结. |
| 1003 | It assumes that no human will ever invent anything great with AGI. | 它假设没有人会发明 任何伟大的与AGI。 |
| 1004 | In reality, the great inventions of the future, inventions that will prevent our extinction, | 在现实中,未来的伟大发明, 防止我们灭绝的发明, |
| 1005 | inventions being studied that we just heard about, will be discovered by people using AGI. | 我们刚刚听说的发明 将被使用AGI的人发现 |
| 1006 | By framing AGI as the last great invention, | 通过将AGI设定为最后的伟大发明, |
| 1007 | the motion demands that you rob those future inventors of credit for their achievements | 动议要求你抢走那些未来的 信用发明者的成就 |
| 1008 | before they even have a chance to dream them up. | 在他们甚至有机会做梦之前 |
| 1009 | See, you should reject this motion not because of some sort of confused theory about the future, | 你看,你应该拒绝这个动议 不是因为某种关于未来的困惑理论, |
| 1010 | but you should reject this motion because of the real world implications of this motion today. | 但你应该拒绝这个动议 因为今天这个动议对现实世界的影响 |
| 1011 | This denial of authorship, this erasure of human expression with AI is a thin excuse. | 这种对作者的否认,这种对AI的人类表达的抹杀,是一种细微的借口. |
| 1012 | It's already happening. | 已经发生了 |
| 1013 | It's happening right now. | 正在发生 |
| 1014 | Part of why I'm here is I've been lobbying the U.S. | 我来这里的部分原因是我一直在游说美国 |
| 1015 | Copyright Office for years to recognize a simple fact, | 版权局几年来 承认一个简单的事实, |
| 1016 | which is that when people make music with AI, those are human creations that deserve protection. | 也就是说,当人们用AI制作音乐时,这些都是人类的创造,值得保护. |
| 1017 | When a parent and their kids make a funny song about their bedtime routine, | 当一个父母和他们的孩子们 做一首有趣的歌 关于他们的睡前常规, |
| 1018 | AI did not write that song. | 这首歌不是大赦国际写的。 |
| 1019 | Those people did. | 那些人就是这么说的。 |
| 1020 | We made some progress, but we've largely lost the plot. | 我们取得了一些进展,但我们基本上失去了这个阴谋. |
| 1021 | For example, in Chris's lovely introduction, he credited my work in AI-generated music, | 例如,在克里斯可爱的介绍中,他称赞了我在AI生成的音乐中的作品, |
| 1022 | which made me wince because that is not a concept I believe in. | 这让我迷惑,因为这不是我所信仰的概念。 |
| 1023 | What if I told you there is no AI-generated song? | 如果我告诉你没有AI生成的歌曲呢? |
| 1024 | There's no AI-generated anything. | 没有AI生成的东西。 |
| 1025 | This term AI-generated will be looked back on as a dehumanizing slur. | AI生成的这个术语将作为一个非人性的污秽来回顾. |
| 1026 | 我们不叫Jimi Hendrix做的吉他音乐 | |
| 1027 | any more than we call the works of Shakespeare quill pen-generated or the discoveries of Galileo telescope-generated. | 比我们称之为莎士比亚笔下的作品 或者伽利略望远镜的发现 |
| 1028 | When a person creates something with AI, creates something great, invents something great, | 当一个人用AI创造一些东西, 创造一些伟大的东西, 发明一些伟大的东西, |
| 1029 | we credit the individual. | 感谢个人。 |
| 1030 | When an AI-powered drone misidentifies a target and kills a civilian, do we blame the AI? | 当AI驱动的无人机误认目标并杀死平民时,我们是否要责怪AI? |
| 1031 | No, we blame the nation state. | 不,我们责怪国家 |
| 1032 | There is no such thing as an AI-generated civilian casualty, only humans directing tools of war. | 没有人工智能造成的平民伤亡,只有人类指挥战争工具。 |
| 1033 | There is no such thing as AI-generated media, only humans directing tools to express themselves. | 没有AI生成的媒体,只有人类引导工具表达自己. |
| 1034 | And when AGI does arrive, there will be no AGI-generated invention, | 等AGI到来 就不会有AGI产生的发明 |
| 1035 | only humans following their nature, which is to adapt and survive. | 只有人类遵循他们的本性 也就是适应和生存 |
| 1036 | So long as humans remain the rightful authors of what they produce with AGI, | 只要人类仍然是他们与AGI合作生产的东西的合法作者, |
| 1037 | every future AGI-assisted breakthrough will be a new human invention. | 未来AGI所协助的每一个突破都会是人类的新发明. |
| 1038 | So maybe it's not the last invention, but we have another burden to prove, and that's greatness. | 所以也许这不是最后的发明, 但我们还有另一个负担要证明,这就是伟大的。 |
| 1039 | Okay, maybe humanity invents stuff with AGI, but AGI will be so much better | 好吧,也许人类发明的东西 与AGI,但AGI会更好 |
| 1040 | and nothing we invent will ever be great anymore. | 我们的发明将不再伟大 |
| 1041 | Well, this is a really easy one. | 嗯,这是一个非常容易的。 |
| 1042 | I just want you to think about who is today's greatest chess player? | 我只想让你想想谁是当今最伟大的棋手? |
| 1043 | I don't have to say his name. | 我不必说他的名字。 |
| 1044 | You already know who that is. | 你已经知道那是谁了 |
| 1045 | He will lose to a 20-year-old computer program every single time. | 他每次都会输给一个20年的电脑程序 |
| 1046 | Is he not great just because, you know, like a computer can do what he does better than him? | 难道他不伟大 只是因为,你知道, 像计算机可以做 他做的比他更好? |
| 1047 | AGI may find solutions better than any person, but only people have the capacity to judge those solutions as great. | AGI可能比任何人找到更好的解决方案,但只有人们才有判断这些解决方案的能力。 |
| 1048 | See, in real life, greatness is obvious. | 在现实生活中,伟大是显而易见的。 |
| 1049 | Obviously, Magnus Carlsen is great at chess, even though he loses to a computer. | 显然,马格努斯·卡尔森虽然输给了电脑,但还是很擅长围棋. |
| 1050 | Obviously, I'm great at writing debate speeches, even though Chachi B.T. | 显然,我擅长写辩论演讲, 尽管Chachi B. T. |
| 1051 | could do a lot better. | 可能会做得更好 |
| 1052 | Back to life, back to reality, what do you see when you look around your real life? | 回到生活,回到现实,当你环顾现实生活时,你会看到什么? |
| 1053 | Not clickbait, not books, not these cynical articles, but your actual life. | 而不是书 也不是这些愤世嫉俗的文章 而是你真实的生活 |
| 1054 | Look past the screen for like a second, and do you actually see AI doom? | 看着屏幕像一秒钟, 你真的看到AI末日? |
| 1055 | Because everywhere I look, I see humanity. | 因为我所看到的每一个地方,我都看到了人性。 |
| 1056 | For all the hysterics about people falling in love with AI chatbots, my cousin just married a lovely human woman just a few weeks ago. | 对于所有关于人们爱上AI聊天机器人的歇斯底里,我的表弟刚刚嫁给了一个可爱的人类女人,就在几周前. |
| 1057 | Why would he do such a thing? | 他为什么要这么做? |
| 1058 | Is it because he's unaware of the loving nature of AI chatbots? | 是不是因为他不知道AI聊天员的爱意? |
| 1059 | Should we inform him that there's a much better alternative to his human wife out there in the market? | 我们是不是该告诉他 市场上还有比他妻子更好的选择? |
| 1060 | People have been using modern AI to make music for the better part of a decade, | 人们一直在使用现代AI 来制作音乐 在十年中更好的部分, |
| 1061 | and it's been common and popular to decry that AI will replace human musicians. | 并且很常见,很受欢迎地说,AI会取代人类音乐家. |
| 1062 | Well, that's a little strange because I saw a street musician in London earlier this week. | 这有点奇怪 因为我本周早些时候在伦敦见过一位街头音乐家 |
| 1063 | He was playing music on the street just like I used to do. | 他和以前一样在街上演奏音乐 |
| 1064 | He was playing saxophone. | 他在玩萨克斯风 |
| 1065 | It was wonderful. | 这是美妙的。 |
| 1066 | What's he doing? | 他在干吗? |
| 1067 | Does he not know he's been replaced? | 他不知道自己被替换了吗? |
| 1068 | Should we go tell him? | 我们应该告诉他吗? |
| 1069 | He showed no signs of AI replacement. | 他没有出现AI替换的迹象. |
| 1070 | In fact, I have yet to meet a musician who's been replaced by AI. | 事实上,我还没有见到一个音乐家,他已经被AI取代了. |
| 1071 | And for the medical researchers making breakthroughs with AI as we speak, should we just tell them to stop that they're no longer necessary? | 而对于医学研究者们,在我们发言的时候,在AI上取得了突破,我们是否应该告诉他们,他们不再需要了? |
| 1072 | Go ahead and pause that research because AGI is going to invent everything in the future and it's probably going to kill us all. | 继续暂停研究 因为AGI会在未来发明一切 它可能会杀死我们所有人 |
| 1073 | Now, the proposition might say and has said that like that's AI, not AGI with the spooky G. | 现在,这个提议可能会说,而且已经说,就像这是AI, 不是AGI与幽灵G。 |
| 1074 | But every step we've taken towards more capable AI has expanded human agency and invention, not diminished it. | 但是,我们朝着更有能力的AI迈出的每一步都扩大了人类的机构和发明,而不是削弱它. |
| 1075 | More people are making music. | 更多的人正在制作音乐。 |
| 1076 | More people are becoming radiologists. | 更多的人成为放射学家。 |
| 1077 | More people are creating small businesses. | 更多的人正在创办小企业。 |
| 1078 | More people are publishing papers. | 更多人发表论文. |
| 1079 | More people are inventing because AI empowers them to invent. | 更多的人正在发明,因为AI赋予他们发明的能力. |
| 1080 | The proposition has not even attempted to explain why that trend line of AI | 该提议甚至没有试图解释为什么大赦国际的这一趋势线 |
| 1081 | empowering invention would suddenly reverse the moment we slap the label AGI on the tools that people are already using. | 赋予发明权力会突然逆转 我们把AGI的标签 放在人们已经使用的工具上 |
| 1082 | And why would you want to reverse that trend line in the first place? | 那你为什么一开始就想扭转这个趋势? |
| 1083 | Why would you want to support discrimination against works and inventions that people create just because they use AI or AGI to do it? | 为什么你们要支持对人们仅仅因为使用AI或AGI来完成而创作的作品和发明的歧视? |
| 1084 | The answer is monopoly, exclusion, elitism. | 答案是垄断、排斥、精英主义。 |
| 1085 | The debate about AGI's future is inescapably a debate about who has the privilege to invent. | 关于AGI未来的辩论无疑是一场关于谁有权发明的争论. |
| 1086 | Who gets to call themselves an inventor? | 谁可以自称发明家? |
| 1087 | Who gets to call themselves a musician? | 谁会自称音乐家? |
| 1088 | Whatever you believe about the distribution of wealth, we should all be on the same page about a fair distribution of capability. | 无论你对财富分配的看法如何,我们都应当站在公平分配能力的同一页上。 |
| 1089 | Please reject this gatekeeping, this attempt to gatekeep human invention. | 请拒绝这种守门术 试图保持人类的发明 |
| 1090 | Stop thinking about everybody's death and start thinking about your own life, your own life. | 别再想每个人的死 开始想你自己的生活,你自己的生活 |
| 1091 | Think about the following. | 思惟以下诸法. |
| 1092 | What if after the debate tonight, and this is for you guys too, I put AGI in your hands. | 如果在今晚的辩论之后,这个也是给你们的, 我会把AGI交给你们。 |
| 1093 | The scariest version you can think of, a machine that could do anything, answer any question, discover any fact. | 你所能想到的最可怕的版本, 一个能做任何事情的机器, 回答任何问题,发现任何事实。 |
| 1094 | Would your honest reaction be to give up, to go home, to stop inventing, to resign from Oxford? | 你诚实的反应是放弃,回家,停止发明,从牛津辞职? |
| 1095 | No, you would use it to do incredible things. | 不,你会用它做不可思议的事情。 |
| 1096 | Before you vote tonight, I want you to ask yourself not what humanity will do with AI or AGI or whatever, but what will you do? | 在你今晚投票之前,我要你问自己 不是人类会用AI或AGI做什么,而是你会怎么做? |
| 1097 | What will you create? | 你要创造什么? |
| 1098 | What will you invent with AGI? | 你会用什么发明AGI? |
| 1099 | Even if you've lost your faith in humanity, which I get, don't be fooled. | 即使你已经失去了对人性的信心 我得到的,不要被愚弄。 |
| 1100 | Don't be fooled by some doomsday prophecy until losing faith in yourself. | 不要被一些末日预言所愚弄 直到失去对自己的信心。 |
| 1101 | Because as long as just one of you keeps inventing great things with AGI or without it, this motion is obviously wrong. | 因为只要你们中只有一个人 不断发明伟大的东西 或没有它, 这个运动显然是错的。 |
| 1102 | Thank you. | 谢谢 |
| 1103 | Next, I would like to call upon some members from the audience to deliver some floor speeches before I call upon the final speaker. | 接着,我请听众中的一些成员发言,然后再请最后一位发言者发言。 |
| 1104 | Yeah, is there a student speaker wishing to speak in favor of the motion? | 是的,有没有学生发言人想发言支持动议? |
| 1105 | Wow, seeing none, is there anyone for opposition? | 哇,没人反对吗? |
| 1106 | Right, the member in the back. | 对 后面那个成员 |
| 1107 | We heard the first speaker of the opposition asked, what is the purpose of humanity? | 我们听到第一位反对派发言人问道,人类的目的是什么? |
| 1108 | We heard the second speaker of the proposition asked, what is an invention? | 我们听到第二位发言者问道,什么是发明? |
| 1109 | Honorable members of the House, let me answer that with Caesar. | 尊敬的众议员,让我向凯撒回答 |
| 1110 | When he crossed the Rubicon, he said three things. | 当他穿过鲁比孔时,他说了三句话. |
| 1111 | (原始内容存档于2018-09-21). Inveni vidi vici. | |
| 1112 | Invention at the Ataman. | 阿塔曼发明. |
| 1113 | Invenire in, to go, to go into. | 进入,进入,进入。 |
| 1114 | And invention doesn't just mean to stumble upon something, it means to find, it means to enter a relationship with the thing. | 发明并不仅仅意味着偶然发现某种事物,它意味着发现,它意味着与某种事物发生关系. |
| 1115 | Invention is not brute discovery, it is attachment. | 发明不是野蛮的发现,而是依附 |
| 1116 | Because we don't invent as humans because we can or cannot, we invent because we care. | 因为我们不作为人类发明,因为我们可以或不能,我们发明是因为我们关心. |
| 1117 | Leonardo did not pursue flight as a nocturnization problem because he loved the sky, that's why he did it. | 莱昂纳多没有将飞行作为节日问题来追求,因为他热爱天空,这就是他这样做的原因. |
| 1118 | Tesla didn't choose current as a spreadsheet exercise. | Tesla没有选择电流作为电子表格练习. |
| 1119 | He adored the universal pulse of energy. | 他喜欢普遍的能量脉搏 |
| 1120 | And at the root of all of that, at the root of invention is love. | 而在这一切的根源, 发明的根源是爱。 |
| 1121 | And love here is not decorative sentiment, it is the orienting force behind our species. | 这里的爱不是装饰性的情感 而是我们物种背后的引领力量 |
| 1122 | And an AGI may search faster, it may compute deeper, it may generate designs at scale, but search without longing is directionless. | 一个AGI可能会更快的搜索,它可能计算得更深,它可能会在规模上生成设计,但是没有渴望的搜索是无方向的. |
| 1123 | Optimization requires a goal. | 优化需要一个目标。 |
| 1124 | Humans generate their goals often irrationally, beautifully, obsessively. | 人类往往以非理性,美丽,执着的方式产生目标. |
| 1125 | We build cathedrals that reach the sky not because acoustics or rocks require it, but because grief does. | 我们建造大教堂可以到达天空 不是因为声学或岩石需要它 而是因为悲伤 |
| 1126 | Because love does. | 因为爱情是。 |
| 1127 | And even if AGI accelerates scientific discovery, it does not originate desire. | 而即使AGI加速了科学发现,它也不起源于欲望. |
| 1128 | It does not wake in the night wanting to solve something for its own sake. | 它不会在夜里醒来 想解决一些事情 为了它自己。 |
| 1129 | It does not love the sky, the sea, or the symmetry of an equation. | 它不爱天空,海, 或一个方程式的对称。 |
| 1130 | Because invention for humanity is entering into a relationship with the world. | 因为人类的发明正在与世界建立关系。 |
| 1131 | And Twain wrote because he loved that world. | 而吐温写作是因为他喜欢那个世界. |
| 1132 | So, honorable members, when you live in your life, when you love in your life, you invent. | 所以,光荣的会员们,当你生活在你的生活中, 当你爱在你的生活中,你发明。 |
| 1133 | And as long as humans love ideas, beauty, justice, each other, we will continue to invent. | 只要人类爱思想,爱美,爱正义,彼此之间,我们就将继续发明. |
| 1134 | And AI might join us, but it won't replace the reason we begin. | 而AI可能加入我们,但它不会取代我们开始的原因. |
| 1135 | And so, I might advise the opposition in the middle of their solutions to maybe find some more love in their lives. | 因此,我建议反对派 在他们的解决方案中间 找到一些更多的爱 在他们的生活中。 |
| 1136 | Anyone wishing to speak for proposition? | 有谁想为提议说话吗? |
| 1137 | Ladies and gentlemen, I'm not afraid of AGI. | 女士们先生们 我不怕AGI |
| 1138 | What I'm afraid of is the reality that comes with it. | 我所担心的就是随之而来的现实. |
| 1139 | The fact of matter is, I am also a data scientist, someone trying to use AI for the benefit of fighting climate change. | 事实上,我也是一个数据科学家,有人试图利用AI来应对气候变化。 |
| 1140 | But in the end, most of the people in the community of whoever is trying to develop AI don't actually get to control what people do with it. | 但最终,在试图开发AI的人的社群中,大多数的人实际上都无法控制人们用它做什么. |
| 1141 | The fact of matter is, the people who really have a say are the people who own the vast computing resource, | 事实是,真正有发言权的人 是拥有巨大计算资源的人, |
| 1142 | the vast money of the world, the vast power of the world, and every single company they could buy. | 世界的巨大财富,世界的巨大力量, 以及他们能够买到的每一个公司。 |
| 1143 | The fact of matter is, AGI will not be owned by you and me. | 事实上 AGI不会归你我所有 |
| 1144 | AGI will be owned by Elon Musk, by Larry Ellison, by the same man who owns everything. | AGI将由埃隆·穆斯克拥有,由拉里·埃利森拥有,拥有一切的同一个人拥有. |
| 1145 | And that will be the end of invention. | 发明就结束了 |
| 1146 | Not because you and I don't have capacity to invent, but because there will be no economic payment for it. | 不是因为你和我没有创造能力 而是因为没有经济报酬 |
| 1147 | Because those people in power, when they have control of something to replace you, they will use it to drive their profit. | 因为当权的那些人,当他们控制了某种东西来取代你时,他们会用它来推动他们的利润. |
| 1148 | That's what AI is being used for. | AI正被用于此. |
| 1149 | I would rather use AI to fight climate change, but they are being used right now to write AI-generated songs | 我宁愿用AI来对抗气候变化,但现在他们正被用来写AI产生的歌曲。 |
| 1150 | and replace musicians from getting their paycheck. | 并代替音乐家获得薪水。 |
| 1151 | You talk about musicians playing on the street. | 你说音乐家在街上玩耍 |
| 1152 | They would rather be paid. | 他们宁愿得到报酬。 |
| 1153 | They would rather be paid by studios and labels, but instead they are going for AI-generated labels. | 他们宁可由工作室和标签付费, |
| 1154 | Ladies and gentlemen, the fundamental problem is that of a human, an AGI will be used against all of us. | 女士们,先生们,根本的问题是 一个人类,一个AGI将会被用来对付我们所有人。 |
| 1155 | Thank you very much. | 谢谢 |
| 1156 | Thank you, Madam Speaker. | 谢谢你,议长女士。 |
| 1157 | Ladies and gentlemen, we are having a very entertaining night, are we not? | 女士们,先生们,我们有一个非常有趣的夜晚,不是吗? |
| 1158 | Tonight's motion is highly seductive, but we must reject the motion, both positively and normatively. | 今晚的动议很诱人 但我们必须积极和规范地拒绝动议 |
| 1159 | Practically speaking, the claim that AGI has exceeded every aspect of our lives, | 实际上,关于AGI超越我们生活各个方面的说法, |
| 1160 | and the fact that further invention is impossible, these claims are just pure speculation. | 以及进一步发明是不可能的事实,这些说法只是纯粹的猜测. |
| 1161 | And the fact that AGI, at this rate, brilliant as it is, is still very much under human control, | 而AGI,以这个速度, 聪明的, 仍然非常在人类的控制之下, |
| 1162 | and the development of AGI is not unbridled, it's still very much governed. | 和AGI的发展并不是无节制的,它仍然受到很大的管理. |
| 1163 | And morally speaking, why should we celebrate this doomsday of human invention? | 从道德上讲,我们为什么要庆祝这个人类发明的末日呢? |
| 1164 | Human invention is the core of human flourishing, agency, creativity and purpose. | 人类发明是人类兴盛,代理,创造和宗旨的核心. |
| 1165 | We shall not celebrate that. | 我们不会庆祝这一点。 |
| 1166 | We shall rather prioritize AI governance, AI ethics. | 我们更应该优先考虑AI治理,AI道德. |
| 1167 | Perhaps we should even change AI from artificial intelligence to assistive intelligence. | 也许我们应该把人工智能变成辅助智能 |
| 1168 | Augmented intelligence, not once as a plan to replace as human beings. | 增强智力,而不是一次作为替代人类的计划. |
| 1169 | And with that moral and positive sense, I'm proud to oppose. | 有了这种道德和积极的感觉,我自豪地反对. |
| 1170 | Thank you very much. | 谢谢 |
| 1171 | And without further ado, I would like to call upon the final speaker for side proposition. | 无需赘言,我要请最后一位发言者提出附带建议。 |
| 1172 | I'm happy to have a mic. | 我很高兴有一个麦克风。 |
| 1173 | So, a piece of life advice, if you are ever invited to debate at Oxford, go first. | 所以,一个生活建议, 如果你曾经被邀请 在牛津辩论,先走。 |
| 1174 | I had some points prepared. | 我准备了几点 |
| 1175 | I have no idea what they are. | 我不知道他们是什么 |
| 1176 | I have eight people to debate and address and random number of students on top of it. | 我还有八个人要辩论和发言, |
| 1177 | So, let's look at some interesting points. | 所以,让我们来看看一些有趣的点。 |
| 1178 | One, I think we're debating like five different topics. | 一,我认为我们辩论 像五个不同的话题。 |
| 1179 | First, we didn't agree on the terms. | 首先,我们没有就条件达成一致。 |
| 1180 | What we have today, AI tools, has nothing to do with AGI, completely different technology. | 我们今天拥有的AI工具与AGI无关,技术完全不同. |
| 1181 | You have software. | 你有软件 |
| 1182 | You have tools. | 你们有工具 |
| 1183 | They are calculators. | 他们是计算器。 |
| 1184 | They prepare your taxes. | 他们准备你的税。 |
| 1185 | That's great. | 赞曰. |
| 1186 | We are talking about a paradigm shift to systems which are as capable as any human being, AGI. | 我们指的是向像任何人一样有能力的系统转变。 |
| 1187 | We don't have it yet. | 我们还没拿到呢 |
| 1188 | So, any criticism you have of modern system, oh, it hallucinated. | 所以,你对现代系统的任何批评, 哦,是幻觉。 |
| 1189 | It made a mistake. | 它犯了一个错误。 |
| 1190 | 因此,AGI XYZ不适用. | |
| 1191 | AGI has nothing to do with super intelligence, a different technology, which gives a singularity. | AGI与超级智能无关,是一种不同的技术,它给出了一个奇特性. |
| 1192 | Singularity is not BS. | 声调不是BS. |
| 1193 | Would you like to sit on the other side? | 你想坐在另一边吗? |
| 1194 | Please join them. | 请加入他们。 |
| 1195 | And if you join them, I would wish you luck because I want the other side to win. | 如果你加入他们 我会祝你好运 因为我希望对方赢 |
| 1196 | I want good for humanity. | 我想为人性好 |
| 1197 | I want free stuff. | 我想免费的东西。 |
| 1198 | I want to cure all the diseases. | 欲治一切诸病. |
| 1199 | But it's a pipe dream. | 但它是一个烟斗梦。 |
| 1200 | We have a bunch of comedians who prepare text and deliver it beautifully, very funny. | 我们有一群喜剧演员, 他们准备了文本,并提供了它漂亮,非常有趣。 |
| 1201 | I appreciate it. | 谢谢 |
| 1202 | But it has nothing to do with science, with technology, with reality, with facts, with experiments. | 但与科学,技术,现实,事实,实验无关. |
| 1203 | Every model released today comes with a report from the right team which tested it and found it to be lying, cheating, trying to escape. | 今天公布的每个模特儿 都会收到一个正确的团队的报告 测试后发现它是假的 作弊 试图逃跑 |
| 1204 | Every prediction, AI safety community made in this space has come true with respect to technology we already have. | 在这个空间里所做的每一个预测, AI安全社区 已经实现了 与我们已经拥有的技术。 |
| 1205 | Then the rest of the predictions will come true. | 然后,其余的预测将会实现. |
| 1206 | There won't be anyone to tell you I told you so. | 没人会告诉你我告诉过你的 |
| 1207 | Some people have a bucket list. | 一些人有一个桶列表。 |
| 1208 | I have an AGI list. | 我有一个AGI名单。 |
| 1209 | A list of things I wanted to do before AGI. | 在AGI之前我想做的事情列表. |
| 1210 | The reading here was one of them. | 这里的阅读是其中之一。 |
| 1211 | But I don't think we got a debate. | 但我觉得我们没有辩论 |
| 1212 | We got random prepared texts. | 我们得到了随机准备的文本。 |
| 1213 | Maybe generated by GPT. | 也许由GPT生成. |
| 1214 | I have no idea. | 我不知道,我还不知道 |
| 1215 | But we haven't addressed the actual arguments. | 但是,我们还没有解决 实际的论点。 |
| 1216 | We haven't addressed what needs to be addressed. | 我们没有解决什么需要解决。 |
| 1217 | The proposition in front of the house is a tautology. | 殿前的命题是同义词. |
| 1218 | It doesn't say can we create AGI. | 它没有说我们能创建AGI. |
| 1219 | It's not asking how soon before AGI. | 它没有问 AGI多久前。 |
| 1220 | It says if AGI, then will it be the last invention? | 它说如果AGI, 那么这将是最后的发明吗? |
| 1221 | And it's true because if it's beneficial and it can actually invent and it automates that part of our cognitive labor, it will make all the inventions. | 之所以如此,是因为如果它有益,而且它能够真正发明, 并且它能实现我们认知劳动的自动化, 那么它就能实现所有发明。 |
| 1222 | And if it's actually not a beneficial AGI, it's malevolent and it kills everyone, well, dead people don't invent stuff. | 如果它实际上不是一个有益的AGI, 它的恶意和杀死每个人, 好了,死人不发明的东西。 |
| 1223 | So by definition, this is always going to be true. | 因此,从定义上讲,这永远是真实的。 |
| 1224 | This debate is not even discussing what is important. | 这次辩论甚至没有讨论什么是重要的。 |
| 1225 | What is important is the question of if we continue on this trajectory, if we continue making more and more advanced AGI, | 重要的是,如果我们继续沿着这一轨迹前进,如果我们继续不断提高非洲一体化倡议, |
| 1226 | whatever mistakes are happening right now, they are going to be corrected. | 无论现在发生什么错误,它们都将得到纠正。 |
| 1227 | We have scaling laws which work. | 我们拥有行之有效的法律。 |
| 1228 | The reason we invest billions of dollars is because we know that with that amount of compute, we get that amount of performance. | 我们投资数十亿美元的原因是, 因为我们知道,通过计算, 我们得到了如此多的业绩。 |
| 1229 | It's very predictable and we can predict how much money we need to get to human level. | 它非常可预测,我们可以预测我们需要多少钱才能达到人类的水平。 |
| 1230 | It's not when AGI, it's how much to AGI. | 这不是当AGI, 它是多少AGI。 |
| 1231 | It used to be a trillion dollars. | 昔为万亿. |
| 1232 | Now it's probably a hundred billion. | 今约千亿. |
| 1233 | It will be a billion soon. | 很快就会有十亿了 |
| 1234 | And then we get to that level. | 然后我们到达那个水平。 |
| 1235 | The process will not stop. | 这一进程不会停止。 |
| 1236 | It will continue. | 它将继续下去。 |
| 1237 | Today we have reports that 100 percent of code for the next A.I. | 今天,我们有报道说,100%的代码 对于下一个人工智能。 |
| 1238 | models is written by A.I. | 模型由A.I.编写. |
| 1239 | This is the early stages of recursive self-improvement. | 这是递归自我改进的初级阶段. |
| 1240 | And as they get better at creating new models, | 随着他们更好地创造新模式, |
| 1241 | they will get better at creating new models. | 他们将更好地创造新的模式。 |
| 1242 | It's not an exponential process. | 这不是一个指数化的过程。 |
| 1243 | It's a hyper-exponential process. | 这是一个超责任过程. |
| 1244 | In data, in human resources, in compute, every resource allocated to those systems. | 在数据、人力资源和计算中,分配给这些系统的每一项资源。 |
| 1245 | So I think it's already smarter than most students at my university. | 所以我觉得这已经比大学里大多数学生都聪明了。 |
| 1246 | I'm not a tax-word. | 我不是一个税字。 |
| 1247 | But I think it will soon be smarter than all of us. | 但我认为它很快就会比我们所有人聪明。 |
| 1248 | And then all of us combined. | 然后我们所有的人结合。 |
| 1249 | IQ tests are meaningless in many cases and they're especially meaningless at high levels of IQ. | 智商测试在很多情况下都是毫无意义的,在智商较高的水平上尤为毫无意义. |
| 1250 | They don't measure anything above 200, let's say. | 比方说,它们不能测量超过200的任何东西. |
| 1251 | But you can kind of imagine what I mean when I say | 但你可以想象 当我说 |
| 1252 | you will not control machines with IQ of a thousand, a million, a billion. | 你将无法控制机器的智商为千,百万,十亿。 |
| 1253 | You won't even be able to tell the difference between them. | 你甚至无法分辨他们之间的区别 |
| 1254 | To you, it's all the same. | 对你来说,一切都是一样的。 |
| 1255 | It's not an infinite growth curve. | 并非无限增长曲线. |
| 1256 | It's impossible. | 这是不可能的。 |
| 1257 | But to us, it's all the same. | 但对我们来说,一切都是一样的。 |
| 1258 | So if we are creating those machine gods, those super-intelligent devices, | 如果我们要创造那些机器神 那些超级智能设备 |
| 1259 | I don't know if I'm allowed to do like surveys and raise your hand things, but think to yourself. | 我不知道我是否被允许 做调查 举起你的手的东西, 但想想你自己。 |
| 1260 | Do you think it's possible to indefinitely control a super-intelligent machine? | 你认为可以无限期地控制一台超级智能机器吗? |
| 1261 | Just think to yourself. | 想想你自己吧,好吗? |
| 1262 | If you have a solution, if you have a working mechanism, you can sell it for billions of dollars. | 如果你有解决方案,如果你有工作机制,你可以以数十亿美元的价格出售. |
| 1263 | Because not a single AI lab in the world has one. | 因为世界上没有一个AI实验室有 |
| 1264 | They don't have a paper published on how to do it, | 他们没有发表论文 如何做到这一点, |
| 1265 | not a patent, not a rigorous blog post. | 并非专利, |
| 1266 | The best they did so far is say, | 他们迄今做得最好的是说, |
| 1267 | when we build it, we'll figure it out. | 当我们建造它,我们会想出办法。 |
| 1268 | Or they say, AI will help us make safe AI. | 或者他们说,AI会帮助我们 安全AI。 |
| 1269 | Those are actual statements. | 这些是实际陈述。 |
| 1270 | Every single leader of a large AI lab is on record as saying, | 一个大型AI实验室的每个负责人 都会说 |
| 1271 | it will probably kill everyone. | 它可能会杀死所有人。 |
| 1272 | Those are the people actually building this technology, | 这些人实际上 建设这种技术, |
| 1273 | experts on how to do it. | 专家如何做到这一点。 |
| 1274 | They all have probability of doom, 20, 30 percent. | 他们都有毁灭的概率,20%,30%。 |
| 1275 | Average for machine learning researchers, average for normal researchers at an AI conference is 30 percent | 机器学习研究人员在AI会议上的正常研究人员平均为30% |
| 1276 | that AI kills everyone. | AI杀死每个人。 |
| 1277 | Not so funny, huh? | 没那么好笑吧? |
| 1278 | If it was 1 percent, it would be insane. | 如果百分之一,那会是疯了。 |
| 1279 | You would never take a bet where you bet your life at those odds. | 你从不赌命赌命 |
| 1280 | They're not betting their lives, | 他们不是在赌命 |
| 1281 | they're betting everyone's lives. | 他们打赌每个人的生命。 |
| 1282 | Eight billion people, future generations, all the kids, everyone you know. | 80亿人,后代,所有的孩子,你知道的每一个人。 |
| 1283 | It's an unethical experiment on human beings. | 这是对人类不道德的实验. |
| 1284 | And it's without consent. | 亦不获同意. |
| 1285 | Even if you like this technology and you want it, you can't consent to it, | 即使你喜欢这个技术,你想要它, 你不能同意它, |
| 1286 | because you don't understand how it works. | 因为你不明白它是如何运作的。 |
| 1287 | You cannot explain it, you cannot predict it, you cannot control it. | 你无法解释 无法预测 也无法控制 |
| 1288 | I know because I published papers on all of those topics. | 我知道是因为我发表了关于所有这些主题的论文。 |
| 1289 | We have a paper surveying | 我们有一个纸调查 |
| 1290 | 50 different impossibility results in AI safety in the top AI survey journal. | 50种不同不可能的结果 AI安全 在最高AI调查期刊。 |
| 1291 | No one has shown that what we are saying is wrong. | 没有人表明我们所说的是错误的。 |
| 1292 | No one published a rebuttal. | 没有人发表反驳。 |
| 1293 | No one has a working system which scales to any level of intelligence. | 任何人都没有能够达到任何水平的智能系统。 |
| 1294 | I love interesting points of information. | 我喜欢有趣的信息点。 |
| 1295 | I don't even know what that means. | 我什至不知道 这意味着什么。 |
| 1296 | So what is important? | 那又怎样? |
| 1297 | We explain the confusion between AI, GI, and artificial super intelligence. | 我们解释AI,GI和人工超级智能之间的混淆. |
| 1298 | We need to understand the difference between tools and agents. | 我们需要了解工具和代理人之间的区别。 |
| 1299 | Tools are wonderful, we need more AI tools. | 工具很棒 我们需要更多的人工智能工具 |
| 1300 | I'm a computer scientist, I'm an engineer, I love technology. | 我是个计算机科学家,我是工程师,我爱科技 |
| 1301 | I use AI tools every day. | 我每天都用人工智能工具 |
| 1302 | They make our lives better, they solve problems. | 他们让我们的生活更好, 他们解决问题。 |
| 1303 | We solve protein folding problem. | 我们解决蛋白质折叠问题。 |
| 1304 | We can solve lots of medical problems, cure diseases, probably achieve immortality with nothing but super intelligent tools. | 我们可以解决很多医疗问题,治疗疾病, 可能实现不朽,只有超智能的工具。 |
| 1305 | We should not create general super intelligence, which is an agent, which works independently of any human. | 我们不应该创造一般的超级情报,它是一个特工,它的工作独立于任何人类. |
| 1306 | No one controls it. | 没有人控制它。 |
| 1307 | No one has to tell it what to do. | 无人得道. |
| 1308 | No one sets goals for it. | 没有人为此设定目标。 |
| 1309 | It does so independently. | 它这样做是独立的。 |
| 1310 | It's an agent and it's an explicit goal of most AI labs. | 这是一种毒剂,也是大多数AI实验室的明确目标. |
| 1311 | They have super intelligence team, they have super intelligence alignment team. | 他们有超级情报团队,他们有超级情报协调团队. |
| 1312 | For whom does the bell toll, I ask you? | 钟声会给谁打响呢? |
| 1313 | For all of us. | 为诸众生. |
| 1314 | I was recently in another debate and there was a student and she shared her story. | 我最近参加了另一场辩论 有一个学生 她分享了她的故事 |
| 1315 | She spent four years of school learning to become an artist and now no one wants her art. | 她花了四年时间学习成为一名艺术家,现在没有人想要她的艺术. |
| 1316 | An AI can replace her, do this job a lot easier, cheaper, faster. | 人工智能可以替代她,做这个工作 更容易,更便宜,更快。 |
| 1317 | She has no idea what to do, she was crying. | 她不知道该怎么办,她在哭。 |
| 1318 | And being the sensitive person that I am, I said, who cares, it's going to kill everyone. | 而我的敏感人物,我说,谁在乎, 它会杀死每个人。 |
| 1319 | We talk about algorithmic bias, we talk about diversity, discrimination, | 我们谈论的是算法偏见, 我们谈论的是多样性,歧视, |
| 1320 | we talk about problems with allocation of energy towards warehouses. | 我们谈论了分配能源给仓库的问题。 |
| 1321 | This has potential of being the last invention we ever make. | 这有可能是我们最后的发明 |
| 1322 | We're going to lose meaning, we're going to lose purpose, and we are very likely to lose our lives. | 我们将失去意义,失去目的,我们很可能失去生命. |
| 1323 | I'm sorry about that. | 对不起 |
该视频共有字幕 1325 条。解锁更多字幕为会员功能，请移动到 价格