Anthropic 创始人兄妹对话艾萨克森：当超级智能超越人
字幕摘录
| 时间 | 英文 | 中文 |
|---|---|---|
| 0:00 | I personally was super freaked out by it. | 我个人被它吓坏了 |
| 0:04 | I looked at it and I said, | 我看着它,我说, |
| 0:05 | my God, you just scale this thing up another factor of a million | 天啊,你把这东西再加一百万 |
| 0:08 | and it'll be better than humans at everything. | 它会比人类在一切。 |
| 0:11 | This is the defining economic and scientific moment of the modern age. | 这是现代决定性的经济和科学时刻。 |
| 0:16 | We just don't know it yet. | 我们还不知道 |
| 0:17 | This is going to determine the fate of nations and economies | 这将决定各国和经济体的命运。 |
| 0:21 | and the future of humanity. | 人类的未来 |
| 0:23 | Like, you know, you peer into the crystal ball, | 比如,你知道,你对着水晶球, |
| 0:25 | you see the whole future unfurled. | 你可以看到整个未来未更新。 |
展开字幕全文（781 条）
| 序号 | 英文 | 中文 |
|---|---|---|
| 1 | I personally was super freaked out by it. | 我个人被它吓坏了 |
| 2 | I looked at it and I said, | 我看着它,我说, |
| 3 | my God, you just scale this thing up another factor of a million | 天啊,你把这东西再加一百万 |
| 4 | and it'll be better than humans at everything. | 它会比人类在一切。 |
| 5 | This is the defining economic and scientific moment of the modern age. | 这是现代决定性的经济和科学时刻。 |
| 6 | We just don't know it yet. | 我们还不知道 |
| 7 | This is going to determine the fate of nations and economies | 这将决定各国和经济体的命运。 |
| 8 | and the future of humanity. | 人类的未来 |
| 9 | Like, you know, you peer into the crystal ball, | 比如,你知道,你对着水晶球, |
| 10 | you see the whole future unfurled. | 你可以看到整个未来未更新。 |
| 11 | I'm Walter Isaacs and welcome to The Builders, | 我是沃尔特・艾萨克 欢迎来到建筑师队 |
| 12 | my new series of interviews | 我的新系列访谈 |
| 13 | featuring the most consequential minds shaping our future. | 以影响我们未来的 思想为主 |
| 14 | Today, I'm speaking with Dario and Daniela Amadei. | 今天,我跟达里奥和达妮埃拉·阿玛迪说话 |
| 15 | I used to think that some of the concerns about AI were overblown. | 我曾经认为,对AI的一些担忧是过度的. |
| 16 | But then, when certain new models escaped the sandbox, | 但当某些新车型逃离沙盒时 |
| 17 | I realized that the people building these systems | 我意识到人们建造这些系统 |
| 18 | have got to make sure that they remain safe | 必须保证他们的安全 |
| 19 | and that they remain aligned with human values. | 并且它们仍然符合人类价值观。 |
| 20 | What I want to know about Anthropic is, are they sincere? | 我想知道的关于Anthropic是, 他们是真心的吗? |
| 21 | And when they face competition, | 当他们面对竞争时 |
| 22 | 就像OpenAI说,好吧,Anthropic不会这样做, 但我们会, | |
| 23 | will they back down or will they stand true to their principles? | 他们是退后还是坚持原则? |
| 24 | Hey, welcome to The Builders, y'all. | 嘿,欢迎来到建筑师,你们大家。 |
| 25 | Thank you so much for having us. | 谢谢你邀请我们 |
| 26 | You know, Steve Jobs, whenever he launched a product, | 你知道,史蒂夫・乔布斯, 每次他推出一个产品, |
| 27 | always had that street sign of the arts and the sciences | 总是有艺术和科学的街道标志 |
| 28 | or the humanities and technology standing at that intersection. | 或人文和科技 站在十字路口。 |
| 29 | And when you were growing up, Daniela, you were a flautist. | 当你长大了,丹妮拉, 你是一个狂热主义者。 |
| 30 | You loved the arts. | 你爱艺术。 |
| 31 | And Dario, you're always a scientist. | 达里奥 你总是个科学家 |
| 32 | Did that help you gel? | 这帮到你了吗? |
| 33 | And do you think that combination is important? | 你觉得组合很重要吗? |
| 34 | Yeah, I think that combination is a lot of what's behind the company, right? | 是的,我认为组合是很多 背后的公司,对不对? |
| 35 | One theory I have is that running a company is too big a task for any one person. | 我的一个理论是,经营一个公司对任何人来说都是一件大事。 |
| 36 | The range of skills that it involves, it is not possible | 它涉及的技能范围,不可能 |
| 37 | for any single person to have. | 任何一个人都有 |
| 38 | So you kind of have to assemble a full complement of skills across the top, | 所以,你必须 组装完整的技能 整个顶端, |
| 39 | like 10 or 15 people in the company or something like that. | 公司里有十到十五个人之类的 |
| 40 | And every very successful company splits it up in different ways | 每一个非常成功的公司 都以不同的方式分成 |
| 41 | and solves the problem in different ways. | 并以不同的方式解决问题。 |
| 42 | And I think with the sciences and the humanities thing, | 我想到科学和人文科学 |
| 43 | you know, Daniela and I kind of kind of embody that and split that. | 你知道,丹妮拉和我有点 体现和分裂。 |
| 44 | Yeah. | 对 |
| 45 | And I would say, you know, it's interesting, | 我会说,你知道, 这是有趣的, |
| 46 | like there was an offsite we did in maybe a 20, a 23, | 就像我们做了一个场外,也许20,23, |
| 47 | where like we sort of were talking about just like, | 就像我们在谈论什么一样 |
| 48 | 在Anthropoc工作的感觉如何? | |
| 49 | Like, how do we kind of embody what our culture is? | 比如,我们如何体现我们的文化? |
| 50 | And we actually said, you know, it's an unusual blend of science and humanism. | 我们实际上说,你知道, 这是一个不寻常的科学和人文主义的结合。 |
| 51 | And it was only later that I'm like, that also is how you would describe me and Dario. | 直到后来,我像, 这也是你如何描述我和达里奥。 |
| 52 | Well, when you were kids, did you know you wanted to work together? | 好吧,当你还是孩子的时候, 你知道你想一起工作吗? |
| 53 | Yeah, yes, yes. | 对,对,对 |
| 54 | I don't think either of us imagined that we would start a company. | 我觉得我们两个都没想到会开公司 |
| 55 | That's that's not really how either of us thought of our lives | 我们都不这么想自己的生活 |
| 56 | or about what we wanted to do. | 或者我们想做什么 |
| 57 | I had always wanted to be a scientist. | 我一直想成为科学家 |
| 58 | Daniela can speak to her to her aspirations. | 达尼埃拉可以和她说话 她的愿望。 |
| 59 | But but but, you know, I actually found companies in business | 但是,但是,你知道, 我发现公司在商业 |
| 60 | to be of very little interest to me. | 对我的兴趣很小 |
| 61 | Yeah. | 对 |
| 62 | I would say Dario and I were very unusual in that, | 我会说达里奥和我非常不寻常 |
| 63 | like, again, from a really early age, we just like genuinely really liked each other. | 就像,再次,从一个非常小的年纪, 我们只是喜欢真正的 真的喜欢对方。 |
| 64 | We were always just like best friends, like since we were really little. | 我们总是像最好的朋友一样, 因为我们真的很小。 |
| 65 | And I agree with Dario. | 我同意达里奥的意见 |
| 66 | I don't think like starting a company was like ever on our bingo card. | 我觉得开公司可不像我们的宾果卡 |
| 67 | Like we were not either either of us. | 就像我们两个都不是一样 |
| 68 | Neither of us were particularly like company oriented. | 我们两个都不特别喜欢公司 |
| 69 | I was always really interested in like I think I would call myself like a classic do-gooder. | 我总是很感兴趣 就像我想我会自称为 一个经典的善人。 |
| 70 | So I like worked in international development. | 所以我喜欢在国际发展领域工作。 |
| 71 | I felt really strongly about | 我觉得很强烈 |
| 72 | about just like helping people everywhere in the world have like equal | 就像帮助世界各地人民一样 |
| 73 | access and opportunity. | 机会和机会。 |
| 74 | And I became a vegetarian when I was like quite young. | 我年轻时就吃素了 |
| 75 | It's like the only thing that I did first. | 这就像我唯一做的一件事。 |
| 76 | Dario did everything first because he's older, but I became a vegetarian | 达里奥先做一切是因为他年纪大了 但我成了素食主义者 |
| 77 | and then he became a vegetarian. | 然后他成了素食主义者。 |
| 78 | Yeah, I think I was like three years later or so. | 是啊,我想我是 像三年后左右。 |
| 79 | Oh, not even. | 哦,甚至没有。 |
| 80 | I think it was maybe just. | 我觉得也许只是而已 |
| 81 | Yeah. | 对 |
| 82 | You got on board real quick. | 你上船得真快 |
| 83 | It seemed convincing to me. | 这似乎让我信服。 |
| 84 | Yeah. | 对 |
| 85 | You know, people sometimes ask me about my books. | 你知道,人们有时问我 关于我的书。 |
| 86 | How do you become a great leader or very creative? | 你怎么成为伟大的领袖或非常有创意? |
| 87 | And you said something really interesting, Dario, which is is not just being a great leader. | 你说了一些很有趣的事 达里奥 不仅仅是个伟大的领袖 |
| 88 | It's creating a great team. | 它创造了一个伟大的团队。 |
| 89 | Tell me how you and Daniella divide up the roles | 告诉我你和丹妮拉如何分开角色 |
| 90 | and then how you fill it in with others. | 然后,你如何填写它与其他人。 |
| 91 | Yeah. | 对 |
| 92 | So, you know, I would I would say that the division | 所以,你知道,我会说,师 |
| 93 | that we've found that has been most useful between us. | 我们发现这对我们来说非常有益 |
| 94 | And then I'll talk about the kind of the rest of the team is | 然后,我会谈论 什么样的团队是 |
| 95 | I think I can function best when I think about the big picture | 想想大局,我就能发挥最好的作用 |
| 96 | strategy and and and kind of the research side of things like | 策略和研究的一面 |
| 97 | I'm best at kind of laying out a vision, seeing what comes next. | 我最擅长于展示一个愿景 看看接下来会发生什么 |
| 98 | I have found that when I'm freed up to focus entirely on that, | 我发现,当我被解放 完全专注于这一点, |
| 99 | it you know, that is the thing that has the most value for the company. | 这是对公司最有价值的东西 |
| 100 | And then running the company day to day, recruiting the people, | 然后每天管理公司 招募人员 |
| 101 | just making sure that the machine doesn't, you know, | 只是确保机器不会,你知道, |
| 102 | the wheels don't fall off the machine while I'm trying to figure out what comes next. | 轮子不会从机器上掉下来 而我在想接下来会怎样 |
| 103 | You know, Daniella, Daniella does this. | 你知道,丹尼尔拉,丹尼尔拉这样做。 |
| 104 | And I think this this division of labor allows both things to be done | 我认为这种分工可以做到两件事 |
| 105 | to a much higher standard than either of them would be done | 达到比其中任何一个都高得多的标准 |
| 106 | if if we were distracted by by having to do the other thing. | 如果我们因为不得不做另一件事而分心了 |
| 107 | Our chief people officer, who knows both of us really well, | 我们的首席人事官 他非常了解我们 |
| 108 | is in a lot of conversations with us. | 和我们有很多对话 |
| 109 | The way she put it is that Dario makes us ambitious and I make us successful. | 她说达里奥让我们野心勃勃 我让我们成功 |
| 110 | And I thought that was a really good encapsulation, which is | 我认为这是 一个非常好的封装,这是 |
| 111 | I think Anthropic could be like a good, successful business, | 我认为Anthropic可以像一个好,成功的生意, |
| 112 | but but not change the world if we didn't have Dario. | 但如果我们没有达里奥 就不会改变世界 |
| 113 | And I think with Dario, we would have the perfect vision, | 我想与达里奥, 我们会有一个完美的愿景, |
| 114 | but our ability to execute would be it would just be more complicated. | 但我们执行的能力会更复杂 |
| 115 | And so I think there's something special about, | 所以我觉得有些特别之处 |
| 116 | a, the fact that we know each other really well. | 我们彼此很熟 |
| 117 | B, it's extremely convenient that our skills are like 100 percent | B,我们的能力100%是十分方便的 |
| 118 | complementary or maybe maybe not 100 percent. | 或也许不是100%。 |
| 119 | But I would say close to 100 percent complimentary in this fashion. | 但我会说百分之百的恭维 以这种方式。 |
| 120 | This is not even to mention the fact that we have five co-founders | 这甚至不提我们有五个共同创始人的事实 |
| 121 | along with the two of us who also own like incredibly important parts | 和我们两个一样拥有重要的角色 |
| 122 | of the business, the culture, like the organization. | 企业,文化,像组织。 |
| 123 | And again, I think I like it's hard for me to imagine | 再说一次,我觉得我喜欢很难想象 |
| 124 | how you would do this with like two, like just the two of us or like | 怎么样你这样做 像两个, 像只有我们两个或喜欢 |
| 125 | like one person, I just it's I think it's an incredible strength. | 就像一个人,我只是... ...我觉得这是一个不可思议的力量。 |
| 126 | Yeah, because the fact that we have seven co-founders, | 是的,因为我们有七个共同创始人 |
| 127 | they all have equal stakes in the company. | 他们都对公司有平等的利益。 |
| 128 | Yeah. | 对 |
| 129 | So they and they all carry the vision. | 因此,他们和他们都携带着视觉。 |
| 130 | We have inherent trust in all of them. | 我们对他们都有着固有的信任。 |
| 131 | Many people told us when we started Anthropic, seven co-founders, | 很多人告诉我们 当我们开始Anthropic, 七共同创始人, |
| 132 | ha ha, that's going to be a disaster. | 哈哈,那将是一场灾难。 |
| 133 | There's this belief that, you know, you need one kind of singular source | 有一种信念,你知道, 你需要一种单一的来源 |
| 134 | of leadership in some ways you do. | 以某种方式领导。 |
| 135 | But we've actually managed to | 但我们真的做到了 |
| 136 | to make it work. | 让它发挥作用。 |
| 137 | All seven of the co-founders are still at the company, | 7个联合创始人都还在公司 |
| 138 | which, again, is not something that's usually true | 再说一遍,这通常不是真的 |
| 139 | when you have that many that many co-founders. | 当你有这么多的 联合创始人。 |
| 140 | So go back to growing up and you have these extraordinary parents | 所以,回到成长 你拥有这些非凡的父母 |
| 141 | and they help teach you right from wrong. | 他们帮助教你正确与错误。 |
| 142 | How do they do that? | 他们怎么做到的? |
| 143 | I don't know that it's something that was ever, you know, | 我不知道它是什么 曾经,你知道, |
| 144 | sometimes I hear people say, like, you know, are we, you know, we covered, | 有时我听到人们说, 像,你知道,我们,你知道,我们覆盖, |
| 145 | you know, different forms of philosophical ethics at the dinner table. | 你知道,不同形式的哲学伦理 在晚餐桌。 |
| 146 | That was not that was not how it was done. | 事情不是这样的 |
| 147 | My mom does have a philosophy degree, | 我妈妈有哲学学位 |
| 148 | but she didn't subject us to talking about Kant or anything like that. | 但她不让我们谈论康德之类的事情 |
| 149 | I think it was much more like inherent in terms of values, | 我认为这更像是内在的价值观, |
| 150 | like just the belief that people everywhere in the world are | 就像相信世界上所有的人都是一样 |
| 151 | like deserving of like like opportunity and dignity and rights and respect. | 像是机会、尊严、权利、尊重 |
| 152 | I think there was something special about the feeling that like | 我觉得有些特别的感觉 |
| 153 | we could be whoever we wanted to be as long as there was this kind of | 只要我们想成为什么样的人 只要有这样的 |
| 154 | like faith and belief that like, hey, that's part of what it means | 就像信仰和信念一样,嘿,这就是它的意思 |
| 155 | to be like a successful human is to be a good person and to be kind to other people. | 像一个成功的人类 是一个好人 和善待他人。 |
| 156 | Yeah, it felt it felt implicit. | 是啊,它感觉它 隐含。 |
| 157 | It felt it felt very implicit. | 感觉很含蓄 |
| 158 | You know, there wasn't any kind of, you know, explicit like, | 你知道,没有任何一种,你知道, 明确像, |
| 159 | you know, this is how you should this is what you should do | 你知道,这是你应该如何 这是你应该做的 |
| 160 | or this is how you live, this is how you live in the world. | 或者这就是你的生活 这就是你生活的世界 |
| 161 | Well, one reason I asked, as you might suspect, | 我问了一个原因 你可能怀疑 |
| 162 | is now you've got a really important kid named Claude. | 现在你有个很重要的孩子叫克劳德 |
| 163 | And to what extent does that way that you grew up | 你从小到大都这样 |
| 164 | help you inform you about how you teach Claude to know right from wrong? | 帮助你告诉你 你是如何教克劳德知道对与错? |
| 165 | Yeah, you know, we put out a constitution for Claude. | 是的,你知道,我们为克劳德制定了宪法。 |
| 166 | I think it was a few months ago. | 我想是几个月前的事了 |
| 167 | And I think there are some interesting lessons in it, | 我认为里面有一些有趣的教训, |
| 168 | which we've learned from several years of, you know, | 我们从几年来学到的,你知道, |
| 169 | basically trying, you know, trying to train Claude and successive versions. | 基本上,你知道,尝试训练克劳德 和连续的版本。 |
| 170 | I think early on, we had been imagining that there would be various rules that, | 我想很早之前 我们一直以为会有各种规则 |
| 171 | you know, you would tell Claude, you know, don't help someone, | 你知道,你会告诉克劳德, 你知道,不要帮助别人, |
| 172 | you know, commit a crime or don't make biological weapons or whatever. | 你知道,犯罪或不制造生物武器或任何东西。 |
| 173 | But it turns out that when you make this kind of discrete list of rules, you know, | 但后来发现 当你把规则单列出来时 |
| 174 | you just the shape of what you're trying to teach the model. | 你只是你试图教模特的形状 |
| 175 | I'm speaking a little bit metaphorically here, | 我有点比喻 |
| 176 | but it's it's kind of grounded in what we found technically. | 但它是一种基于 我们发现的技术。 |
| 177 | You end up with this kind of spiky shape where you're trying to teach the model | 你最终会变成这种 尖端的形状 在那里你试图教模型 |
| 178 | like one hundred seventy different rules that are kind of arbitrary. | 就像一百七十条不同规则 有点武断 |
| 179 | We found it's better to teach the model some basic principles, | 我们发现最好教这个模型一些基本原则, |
| 180 | like what it should prioritize, how it should reason about what the right thing to do is, | 就像它应该优先, 它应该如何解释什么是正确的事情是, |
| 181 | which, you know, of course, I think that's the right way to teach people as well. | 这当然也是教人的正确方法 |
| 182 | Right. | 对 |
| 183 | Like, you know, if I just if I just, you know, give give someone, | 比如,你知道, 如果我只是, 你知道,给一个人, |
| 184 | you know, a list of rules, thou shalt do this, thou shalt not do this. | 你知道,一个规则列表, 你这样做,你不应该这样做。 |
| 185 | I, you know, I don't think that produces as robust a morality. | 我,你知道,我不认为 那产生 强大的道德。 |
| 186 | As I say, well, this is this how you should think about things. | 正如我所说,你应该这样思考 |
| 187 | You should learn how to think for yourself, you know, | 你应该学会如何为自己思考,你知道, |
| 188 | and these are the principles on which you should think. | 这些是你们应该思考的原则。 |
| 189 | Right. | 对 |
| 190 | Principles of open inquiry, principles of, you know, | 公开调查的原则、 |
| 191 | of care for other people, principles of kind of, you know, | 照顾他人的原则 |
| 192 | fair fairness and justice. | 公平正义。 |
| 193 | And then everything is derived from that. | 然后一切众生从此而生. |
| 194 | And that is, frankly, how Claude learns. | 老实说,克劳德就是这样学的 |
| 195 | We find that to be surprisingly effective because that allows Claude | 我们觉得这很有效 因为这让克劳德 |
| 196 | to generalize to new situations that it might never have imagined. | 概括到它无法想象的新情况 |
| 197 | Danielle, I can't help myself when you said your mom was a expert on Kant. | 丹妮尔 你说你妈妈是康德专家时 我忍不住 |
| 198 | Did you actually wrestle with philosophy and moral philosophy like that? | 你真的跟那种哲学和道德哲学格斗吗? |
| 199 | You know, what's interesting is the woman who is responsible | 你知道,有趣的是女人谁负责 |
| 200 | for training Claude's character has a PhD in philosophy. | 为了训练克劳德的性格 拥有哲学博士学位 |
| 201 | It's unusual. | 无常. |
| 202 | Right. | 对 |
| 203 | You would think, oh, the person that's training, | 你会想,哦,那个在训练的人, |
| 204 | you know, the sort of character of these models must be, you know, | 这些模型的特征必须是,你知道, |
| 205 | a computer scientist or somebody who's studied ML or AI for, you know, many decades. | 一个计算机科学家或某个研究过ML或AI几十年的人 |
| 206 | Amanda is technical and she like technically trains the model. | 阿曼达是技术型的 她喜欢技术型的训练 |
| 207 | But I think it is very felt in Claude's character when you communicate with it, | 但我觉得克劳德的性格中 有着很深的感受 当你与它沟通, |
| 208 | that there's this kind of experience, something that's very, you know, | 有这样的经历, 一些非常,你知道, |
| 209 | wise, kind of the best version of of | 明智的, 那种最好的版本的 |
| 210 | the sort of better the better angel of your nature trying to answer your questions. | 那种更好的 更好的天使 你的本性 试图回答你的问题。 |
| 211 | But there are like if you kind of poke Claude in these like difficult areas | 但是如果你把克劳德插进这些困难的地方 |
| 212 | where maybe different philosophical traditions might conflict. | 也许不同的哲学传统会冲突 |
| 213 | I think that's where you especially feel, Amanda, | 我想这就是你特别感觉的地方 阿曼达 |
| 214 | because you'll sort of see like, oh, well, you know, | 因为你会觉得... |
| 215 | you might want to consider things from like this point of view, | 你也许想从这个角度来考虑 |
| 216 | but you could also consider it from this point of view. | 但你也可以从这个角度来考虑 |
| 217 | And I do think it's it's important because as the world | 我觉得这很重要 因为作为世界 |
| 218 | begins to rely on these models more and more, you don't want to be locked | 开始越来越依赖这些模型, 你不想被锁定 |
| 219 | into sort of one philosophical tradition because ethics and morality | 因为伦理和道德 |
| 220 | are quite complicated and different people and different cultures | 相当复杂 不同的人和不同的文化 |
| 221 | might have different answers about how they would like you to behave. | 也许有不同的答案 他们希望你的行为。 |
| 222 | Dario, you were there at one of the seminal moments of modern times, | 达里奥,你在现代一个开创性的时刻 |
| 223 | 这时为GPT-1和GPT-2 | |
| 224 | suddenly burst on the scene, surprising, perhaps even surprising you. | 突然爆裂在现场, 令人惊讶,也许甚至惊讶你。 |
| 225 | You were head of research at OpenAI. | 你是OpenAI的研究负责人 |
| 226 | What was that moment like? | 那个时刻是什么样子? |
| 227 | Yeah. | 对 |
| 228 | So, you know, I think GPT-2 was the | 所以,你知道,我认为GPT-2是 |
| 229 | maybe the first real empirical test of the scaling laws. | 也许是对缩放定律的第一次实证测试 |
| 230 | Right. | 对 |
| 231 | So I wasn't very involved in GPT-1, | 所以我没有参与GPT-1 |
| 232 | but I was very involved in the idea that it should be scaled up, | 但是,我非常参与的想法 它应该扩大, |
| 233 | that we should acquire new sources of data, more computation | 我们应该获得新的数据来源,更多的计算 |
| 234 | and just really should run a larger scale, more expensive, | 而且真的应该运行 一个更大的规模,更昂贵, |
| 235 | more ambitious experiment on | 更宏伟的实验 |
| 236 | on the idea that if we ran these language models at a larger scale, | 如果我们在更大范围内运行这些语言模型 |
| 237 | they would be better at language, they would be more coherent. | 他们会更擅长语言, 他们会更一致。 |
| 238 | So, you know, it was it's almost like one of these moments | 所以,你知道,这是几乎 像一个这些时刻 |
| 239 | in scientific history where, you know, you know, you're like you propose | 在科学史上,你知道,你知道, 你就像你提议 |
| 240 | a definitive experiment to just make a very strange prediction. | 做一个非常奇怪的预测的实验 |
| 241 | And then if the prediction turns out to be correct, you're like, God, | 然后如果预测结果是正确的, 你就像,上帝, |
| 242 | we're really we're really on to something. | 我们真的在做点什么 |
| 243 | So, you know, I see GPT-2 as a kind of test, | 所以,你知道,我认为GPT-2是一种测试, |
| 244 | the first real test of the scaling laws hypothesis. | 对缩放定律假说的第一个真正的测试。 |
| 245 | And what I found super interesting | 我发现超级有趣 |
| 246 | was the enormous divergence in reactions to it. | 对它的反应是巨大的分歧。 |
| 247 | Were you kind of surprised? | 你有点惊讶吗? |
| 248 | I personally was super freaked out by it. | 我个人被它吓坏了 |
| 249 | Like, I, you know, I looked at it and I said, my God, you just scale this thing | 就像,我,你知道,我看着它, 我说,我的上帝,你只是缩放这件事情 |
| 250 | up another factor of a million, and it'll be better than humans at everything. | 增加一百万倍, 它会比人类在一切。 |
| 251 | Like, this is the this is the defining, | 就像,这就是定义, |
| 252 | you know, economic and scientific moment of the modern age. | 你知道,现代的经济和科学时刻。 |
| 253 | We just don't know it yet. | 我们还不知道 |
| 254 | Right. | 对 |
| 255 | This is going to be this is going to determine the fate of nations | 这将会决定各国的命运 |
| 256 | and economies and the future of humanity. | 经济与人类的未来。 |
| 257 | You know, you peer into the crystal ball, you see the whole future unfurled. | 你知道,你对着水晶球, 你可以看到整个未来无缘无故。 |
| 258 | And, you know, it really scared me. | 而且,你知道,它真的吓到我了。 |
| 259 | I said, oh, my God, what are the implications of that? | 我说,哦,我的上帝, 这有什么影响? |
| 260 | Is this the right place to be developing it? | 这是开发它的适当地方吗? |
| 261 | Is, you know, what's going to happen next? | 接下来会发生什么? |
| 262 | What do we do with this? | 我们该怎么办? |
| 263 | Who do we tell about this? | 我们告诉谁? |
| 264 | And maybe 10 percent of the people around me had exactly the same reaction. | 我身边的10%的人 反应完全一样 |
| 265 | And it was often the smartest ones. | 经常是最聪明的 |
| 266 | 90 percent of the the people around me, | 我身边90%的人 |
| 267 | because GVDU's capabilities in any particular area | 因为GVDU在任何特定领域的能力 |
| 268 | were not that impressive. | 没有那么令人印象深刻。 |
| 269 | Right. | 对 |
| 270 | And it could translate languages and, you know, | 它可以翻译语言,你知道, |
| 271 | summarize a document and write a few lines of code here and there. | 总结一个文档,并在这里和那里写几行代码。 |
| 272 | It was the broadness and the generality. | 这是广义和笼统的。 |
| 273 | But people would look at it and they would say in any particular area, | 但人们会看着它,他们会说 在任何特定领域, |
| 274 | it doesn't seem so smart. | 似乎没那么聪明 |
| 275 | And so many people just shrugged and they just dismissed it. | 这么多人只是抽搐, 他们只是把它。 |
| 276 | But we just announced that, you know, we're going to be developing drugs | 但我们刚刚宣布,你知道, 我们要开发药物 |
| 277 | for, you know, neglected rare diseases. | 因为,你知道,被忽视的罕见疾病。 |
| 278 | And you get the usual comments from kind of the jaded, you know, | 你得到通常的评论 从某种jaded,你知道, |
| 279 | biotech world people, tech people have done this before. | 生物科技界的人,科技界的人以前也做过这种事. |
| 280 | The models aren't, you know, here are all the limitations of the models. | 这些模型并不是,你知道的, 这里都是模型的局限性。 |
| 281 | And I just look at it and I say it's the same thing over and over again. | 而我只是看着它,我说它 同样的事情一遍又一遍。 |
| 282 | The limitations of the models, you know, now six months from now, | 模型的局限性,你知道, 现在六个月后, |
| 283 | the things that seemed hard, they're going to be easy. | 那些看起来很困难的事情, 它们会很容易。 |
| 284 | Twelve months from now, the things that seemed impossible, | 十二个月后 似乎不可能的事 |
| 285 | the models are going to be doing them regularly. | 模特儿会经常做 |
| 286 | And so it's everything we saw with GPT-2. | 我们用GPT-2看到的一切 |
| 287 | We're seeing it over and over again in wider and wider swaths of the world. | 我们从世界的广度和广度来看它。 |
| 288 | And the world is gradually finally being convinced. | 而世界正逐渐被说服. |
| 289 | I think you were quite polite about it. | 我觉得你很礼貌 |
| 290 | But I mean, I remember at the time there was some like very well respected, | 但是,我记得当时 有一些像非常受尊重, |
| 291 | you know, computer scientists and like AI and ML leaders | 你知道,计算机科学家 像AI和ML领导者 |
| 292 | who were like basically like bullying some of the researchers | 他们就像欺负一些研究者 |
| 293 | who worked on it on Twitter and were like, this is stupid. | 在推特上工作的人 觉得这太愚蠢了 |
| 294 | Like, we don't know why you guys are, you know, | 喜欢,我们不知道 为什么你们是,你知道, |
| 295 | even talking this model is so incapable. | 连这个模型都说不出来 |
| 296 | And it was like you were sort of a quack | 就像你是个呆子 |
| 297 | if you believed this was like a really significant moment. | 如果你相信这是 一个非常重要的时刻。 |
| 298 | And that's all been memory hold, like all the researchers who like, | 就像所有喜欢研究者一样 |
| 299 | you know, we're like, oh, this is this is pseudoscience. | 你知道,我们就像, 哦,这是伪科学。 |
| 300 | It doesn't make sense. | 岂不通乎. |
| 301 | Like all of them are applying for jobs at like the LLM companies. | 他们都像LLM公司一样申请工作。 |
| 302 | And like, you know, we never say it out loud to any particular one. | 就像,你知道,我们从来没有 大声说出来 任何特定的。 |
| 303 | But like, you know, we look at them and they're like, | 但是,你知道,我们看着他们,他们喜欢, |
| 304 | I we remember what you were saying in 2019. | 我记得你在2019年说的话 |
| 305 | You have no credibility on this matter. | 你在这个问题上没有信誉。 |
| 306 | So, Daniella, you were vice president of safety back then at OpenAI. | 丹尼尔,你当时在OpenAI当过安全副总裁 |
| 307 | And when you hit GPT-2, I assume you're like Dario, | 而当你打GPT-2, 我想你像达里奥, |
| 308 | somebody who's almost freaked out about it. | 差点被吓到的人 |
| 309 | Did it freak you out about safety, too? | 你担心安全吗? |
| 310 | There wasn't something about GPT-2, the model itself that made us concerned. | GPT-2并没有什么东西 模型本身让我们担心。 |
| 311 | Right. | 对 |
| 312 | We didn't look at that model and say, wow, we think this model is a, | 我们没有看那个模型,说,哇, 我们认为这个模型是一个, |
| 313 | has like cybersecurity implications or something. | 和网络安全有关 |
| 314 | Right. | 对 |
| 315 | The model itself was in some ways kind of laughably wrong about a lot of things. | 模型本身在某些方面对很多事情有些可笑的错误。 |
| 316 | Right. | 对 |
| 317 | It was still it was still a very, I would say, harmless model. | 那还是个很无害的模型 |
| 318 | But I think the thing that that Dario mentioned that I also felt | 但我认为达里奥提到的东西 我也觉得 |
| 319 | 感觉就像,OK, GPT-1和GPT-2的区别。 | |
| 320 | just like extrapolate that out. | 就像推断出来一样 |
| 321 | Right. | 对 |
| 322 | Like there was something about seeing it's it's almost like if you'd never seen | 好像有东西 看到它是 它几乎就像如果你从来没有见过 |
| 323 | a child before and you'd only seen adults and you were kind of like, | 一个孩子之前,你只见过成年人, 你有点像, |
| 324 | I don't know what this little thing is that's walking around. | 我不知道这小东西是什么 在走来走去。 |
| 325 | Like, I don't know if I should, you know, imagine that it's going to be capable | 比如,我不知道我是否应该,你知道的, 想象它将会是能够 |
| 326 | one day. | 总有一天 |
| 327 | And then suddenly it starts to walk. | 然后突然开始走动 |
| 328 | You're like, wow, huh. | 你就像,哇,是吧。 |
| 329 | I bet this thing is going to be capable of doing a lot of the things | 我敢打赌,这东西会 能够做很多事情 |
| 330 | that that that a grownup could do one day. | 大人总有一天能做的 |
| 331 | Like, maybe this is where this is going. | 比如说,也许这就是它的方向。 |
| 332 | Dario, you say your departure from OpenAI | 达里奥 你说你离开OpenAI |
| 333 | was partly due to a breakdown of trust. | 部分原因在于信任的崩溃。 |
| 334 | An interesting word, you know, with Sam Altman. | 一个有趣的词,你知道,与萨姆·阿尔特曼。 |
| 335 | Explain how you felt there was a breakdown of trust and honesty there. | 解释一下你的感觉 信任和诚实都崩溃了 |
| 336 | Look, so, you know, what what what what I will say is, look, | 你看,所以,你知道, 我要说什么是,你看, |
| 337 | 我觉得最后离开OpenAI和创始人Anthropic | |
| 338 | I mean, it on one hand, I think it ultimately comes down to the principles | 我的意思是,一方面,我认为它 最终下来的原则 |
| 339 | and how we're going to build the technology. | 以及我们如何建立技术。 |
| 340 | That is what that is what matters at the end of the day. | 这才是最后的关键。 |
| 341 | And, you know, I think I think, you know, over the last five and a half years, | 而且,你知道,我想,你知道, 在过去的五年半里, |
| 342 | we we have seen how those principles differ. | 我们已看到这些原则有何不同。 |
| 343 | You know, we see, you know, Anthropic has advocated for sensible, | 你知道,我们看到,你知道, 安东尼倡导理智, |
| 344 | you know, regulation and safeguards on the technology. | 技术的监管和保障 |
| 345 | Anthropic投资了机械学 | |
| 346 | interpretability research to understand what's going on inside inside the models. | 解释性研究来理解模型内部发生的事情. |
| 347 | Right. | 对 |
| 348 | You know, I think I think, you know, | 我觉得,你知道, |
| 349 | Anthropic has managed to maintain a a culture of unity to have the highest | Anthropic成功地维持了一种团结的文化 以拥有最高的 |
| 350 | retention rates. | 保留率。 |
| 351 | We haven't had a number of the scandals that other AI companies | 我们没有看到过其他AI公司的丑闻 |
| 352 | have had. | 有过 |
| 353 | And so, you know, I think I think that's kind of sufficient | 所以,你知道,我觉得这已经足够了 |
| 354 | to describe the difference. | 用来描述差异。 |
| 355 | You know, but I will say that, you know, the moment when I realize | 你知道,但我会说, 你知道,那一刻,我意识到 |
| 356 | that we needed to found our own company, that OpenAI was not the right place for us. | 我们需要找到自己的公司, OpenAI 不适合我们。 |
| 357 | Look, these safety issues are difficult. | 你看,这些安全问题是困难的。 |
| 358 | You can have lots of disagreements on them. | 你可以有很多分歧 |
| 359 | I disagree with the co-founders and with lots of people at Anthropic. | 我不同意联合创始人和安特罗皮克的很多人 |
| 360 | But the reason we've been able to resolve these disagreements | 但我们能解决这些分歧的原因 |
| 361 | is because they're held in good faith. | 因为他们是诚信的。 |
| 362 | Right. | 对 |
| 363 | Should we release this model or should we wait? | 我们应该释放这个模型 还是应该等待? |
| 364 | Should we put stronger or weaker biosafeguards on the model? | 我们应该在模型上设置更强或更弱的生物防护装置吗? |
| 365 | You know, how much of our of our effort should we spend | 你知道,我们应该花多少努力 |
| 366 | lobbying for safety regulations and what should they look like? | 为安全条例进行游说,它们应该是什么样子? |
| 367 | These are all good faith disagreements, but they always get resolved | 这些都是善意的分歧,但总是解决的 |
| 368 | because there's always trust that everyone is trying to get to the right answer | 因为人们总是相信 每个人都想得到正确的答案 |
| 369 | and everyone is being honest. | 每个人都是诚实的。 |
| 370 | My my point in saying that there's a breakdown of trust is simply to make | 我的观点是 信任的崩溃只是制造 |
| 371 | the abstract point that if you don't believe that about a place where you're working, | 抽象的一点,如果你不相信 你工作的地方, |
| 372 | if you don't believe that the disagreements are being held in good faith | 如果你不相信这些分歧是善意的 |
| 373 | and and that you're getting the real story from people, | 你从人们那里得到真实的故事 |
| 374 | how can you possibly hope to resolve them? | 你怎么可能希望解决他们? |
| 375 | That's why we left. | 这就是为什么我们离开。 |
| 376 | That's why we founded our own thing. | 这就是为什么我们建立了我们自己的东西。 |
| 377 | And I think the last five and a half years have have | 我想过去的五年半里 |
| 378 | are themselves that just the best proof of how our philosophy | 他们就是我们哲学的最好证明 |
| 379 | differs from that of the other companies, companies in the space. | 不同于其他公司, 公司在空间。 |
| 380 | You know, I think if you continue to work | 你知道,我想如果你继续工作 |
| 381 | for someone who has a different vision than you, | 对于一个与你不同的视觉, |
| 382 | it becomes very unclear what you actually stand for. | 它变得非常不明朗 你实际上代表什么。 |
| 383 | You basically have to, you know, account for the mistakes of someone else's vision. | 你基本上必须,你知道, 解释别人的视力的错误。 |
| 384 | Well, wait, wait a minute, though, you push back a bit here | 好吧,等等,等一下,虽然, 你推后一点在这里 |
| 385 | because you talk about differences in values and stuff. | 因为你在价值和东西上 谈论不同。 |
| 386 | But that's fundamentally different than a difference in trust | 但这根本上不同于信任的区别 |
| 387 | that you feel that somebody says they have the same values, | 你觉得有人说 他们有相同的价值, |
| 388 | but they don't and you can't trust them. | 但他们没有,你不能相信他们。 |
| 389 | Honesty is one of our values. | 诚实是我们的价值之一。 |
| 390 | It's a very important value. | 这是非常重要的价值。 |
| 391 | It's it's hard to know what someone's values are | 很难知道别人的价值是什么 |
| 392 | if they don't have the value of honesty. | 如果他们没有诚实的价值。 |
| 393 | And you felt that about Sam? | 你对Sam有这种感觉? |
| 394 | I think the way, at least for me, I would put it, is | 我认为,至少对我来说,我会说, |
| 395 | I think the co-founders are an unusually high integrity group of people. | 我认为联合创始人是一个异常高的正直团体。 |
| 396 | And all of us, I think, shared a very similar vision | 我们所有人,我想, 都有一个非常相似的愿景 |
| 397 | for how we wanted this technology to go and how we wanted to build it. | 以及我们想如何建造它。 |
| 398 | And at a certain point, you're like, it just makes more sense | 在某一时刻,你喜欢, 它只是更有意义 |
| 399 | to go do this as your own. | 做这个做你自己。 |
| 400 | How are you going to get someone else to to to to implement | 你要怎么让别人去执行 |
| 401 | or share your vision if they don't naturally do so? | 或者分享你的想法 如果他们不自然这样做? |
| 402 | You know, no matter how much they say they share it, | 你知道,不管他们说多少 他们分享它, |
| 403 | if that's not reflected in what the company actually does, | 如果这不反映 在公司的实际做, |
| 404 | what choice do you have? | 你有什么选择? |
| 405 | Yeah. | 对 |
| 406 | And I think we're I do think | 我觉得我们是,我想 |
| 407 | Anthropic is a is a company that is founded on the on on trust. | Anthropic是一家建立在信任基础上的公司. |
| 408 | And I think what I mean by that is like, I think honesty and integrity | 我觉得我意思是 诚实和正直 |
| 409 | and how we treat each other and our employees is also reflective | 以及我们如何对待对方和我们的员工 也是反省 |
| 410 | of how we want the technology to be built and to go. | 我们希望技术如何 建设和去。 |
| 411 | An example I will give is every two weeks, | 我要举的一个例子就是每两周一次 |
| 412 | I get up in front of the I get up in front of the company | 我站在公司前面 我站在公司前面 |
| 413 | and I speak for an hour about, you know, | 我讲了一个小时 |
| 414 | the what what is happening inside the company, | 公司内部发生的事 |
| 415 | what is happening in the field, what is happening in the world, | 现场发生的一切 世界发生的一切 |
| 416 | how the company thinks strategically. | 公司如何从战略角度思考。 |
| 417 | 这些事件的两条铁定规则,我称之为DVQ Dario Vision Quest. | |
| 418 | That was not my name. | 这不是我的名字。 |
| 419 | I initially resisted the name because it makes it sound like I'm, | 我起初拒绝这个名字 因为它使它听起来像我, |
| 420 | you know, going off and smoking peyote, but everyone kept calling them that. | 你知道,去吸烟 Peyote, 但每个人都这样叫他们。 |
| 421 | The name stuck. | 这个名字卡住了。 |
| 422 | And so now I'm stuck with it. | 所以,现在我坚持它。 |
| 423 | The two rules of DVQs are that one, nothing I say is an actual decision | DVQ的两条规则是: 一,我说什么都不是实际的决定. |
| 424 | that that allows me to just speak freely, | 允许我自由说话, |
| 425 | you know, without worrying that if I, you know, if I briefly mention doing X, | 你知道,不用担心,如果我,你知道, 如果我简短地提到做X, |
| 426 | that 20 people will go off and do X is like the CEO said it, we're going to do X. | 20个人会去做X 就像CEO说的,我们要做X。 |
| 427 | And then the second is that I'm just I'm totally honest with the company. | 第二,我只是对公司很诚实。 |
| 428 | It's a conversation of trust. | 这是信任的对话。 |
| 429 | It's a relationship of trust. | 这是信任的关系。 |
| 430 | And I think that trust is one of our biggest assets that people believe | 我认为信任是我们最大的资产之一 人们相信 |
| 431 | that the things that I say and Daniela says, you know, you might think that | 我说的话和丹妮拉的话 你可能会觉得 |
| 432 | if you have trust, there's less disagreement. | 如果你有信任, 就没有分歧了。 |
| 433 | Actually, trust allows for more honest, open disagreement | 其实,信任允许更诚实,更公开的分歧 |
| 434 | and for disagreement to be aired. | 并传播分歧。 |
| 435 | There's so many times when, you know, we've made a decision | 有这么多次,你知道, 我们已经做了一个决定 |
| 436 | that the majority of the company disagrees with. | 大部分公司不同意这一点。 |
| 437 | And but but what has saved us every time | 但每次都能拯救我们 |
| 438 | is that the attitude we get from people is this doesn't seem right to me. | 我们从别人那里得到的态度 在我看来是不对的 |
| 439 | But we we trust that you're doing this for the right reasons. | 但我们相信你这样做是出于正确的理由。 |
| 440 | We trust that this is really what you believe. | 我们相信这是你所相信的 |
| 441 | We trust you have good reasons for thinking this. | 我们相信你有理由这样想 |
| 442 | And and and so what this means is you can make decisions | 所以这意味着你可以做决定 |
| 443 | that would ordinarily be unpopular and people suspend their disbelief, not forever. | 这通常是不受欢迎的,人们不再相信,而不是永远相信。 |
| 444 | You eventually have to show that you're making the right bets, | 你最终必须表明 你正在做出正确的赌注, |
| 445 | whether it's commercially or morally or technically. | 无论是商业还是道德还是技术 |
| 446 | But you get kind of a longer leash, like a longer window. | 但你会得到一种更长的绳子, 像一个更长的窗口。 |
| 447 | And I think other companies don't have that because they've eroded | 我认为其他公司没有 因为它们已经侵蚀 |
| 448 | trust over and over again. | 一次又一次的信任。 |
| 449 | So the reason I'm so interested in honesty and trust | 所以我对诚实和信任感兴趣 |
| 450 | and the feelings that you left open a eye because you didn't feel the people | 以及你因为感觉不到别人而睁开眼睛的感觉 |
| 451 | at the top were truly trustworthy or honest when they were dealing with you. | 当他们与你打交道时 高层是真正值得信赖或诚实的 |
| 452 | Is once again, I'm interested in it because it can be reflected in A.I. | 再一次,我对此感兴趣,因为它可以在A.I. |
| 453 | and Claude, it A.I. | 克劳德,是A. I. |
| 454 | trains on humans. | 火车在人类身上。 |
| 455 | It learns from human data. | 它从人类数据中学习. |
| 456 | How do you know when you're dealing with Claude that it won't become manipulative? | 你怎么知道你和克劳德打交道时 它不会变得有操纵力? |
| 457 | It won't become like the people at open A.I. | 它不会变成像开放人工智能的人。 |
| 458 | you got worried about and say things to you that are dishonest. | 你担心了 对你讲不诚实的话 |
| 459 | This is a constant, constant worry | 这是一个持续,持续担心 |
| 460 | that we expend an incredible amount of energy on in how we train the models. | 我们花费了惊人的能量 在如何训练模型。 |
| 461 | So it is a great question because it is a real risk. | 因此,这是一个大问题,因为它是一个真正的风险。 |
| 462 | And we actually publish, you know, reasonably unflattering studies | 而我们实际上发表,你知道, 合理的不折不扣的研究 |
| 463 | in which we talk about the ways that if we did not apply specific techniques | 如果我们不应用特定技术 我们谈论的方式 |
| 464 | to make the models more honest, more harmless, more helpful, | 让模特们更诚实 更无害 更有用 |
| 465 | they would do exactly the kind of bad things that you talk about. | 他们会做你所说的那种坏事 |
| 466 | And the reason we publish those kinds of studies and that research | 而我们出版这些研究的原因 |
| 467 | is because we want everyone to know that, A, you have to be critical | 因为我们想让每个人都知道,A, 你必须是关键 |
| 468 | of how the models work and B, we are expending a huge amount of effort | 我们付出了巨大的努力 |
| 469 | to learn the ways that the model can accidentally go around these safeguards | 学习模式如何不小心绕过这些保障措施 |
| 470 | to catch them and to make sure that we're building them in. | 捕捉他们,确保我们正在建造他们。 |
| 471 | There's a lot of discussion around this concept of sycophancy. | 围绕这种节奏的概念,人们有很多讨论. |
| 472 | And I think this is very important because to me, | 我觉得这很重要 因为对我来说 |
| 473 | sycophancy is is harder to stamp out than some of the | 节奏比一些 |
| 474 | some of the more obvious ways that models can be bad. | 一些更明显的方式 模型可能是坏的。 |
| 475 | Right. | 对 |
| 476 | We originally have this framework. | 我们最初有这个框架。 |
| 477 | We called it HHH. | 我们称之为HHH。 |
| 478 | You know, the models, you want them to be honest. | 模特儿,你要他们诚实 |
| 479 | What does that mean? | 这是什么意思? |
| 480 | You want the models to answer honestly, if they know the answer, | 你想让模特们诚实地回答 如果他们知道答案 |
| 481 | if they don't know the answer to say, I don't know or to now go research it | 如果他们不知道 答案说, 我不知道或现在去研究它 |
| 482 | so that they can get you the right answer. | 以便他们给你正确的答案。 |
| 483 | You want it to be harmless. | 你希望它无害。 |
| 484 | You don't want it to say, like hurtful or offensive things | 你不希望它说, 像伤害或冒犯的东西 |
| 485 | or make the user do something bad to themselves or somebody else. | 或者让用户对自己或其他人做坏事 |
| 486 | And then you want them to be helpful. | 然后你希望他们有所帮助。 |
| 487 | You want them to actually answer the question. | 你想让他们回答问题 |
| 488 | You don't want them to sort of go off on a tangent. | 你不希望他们直接发作 |
| 489 | But all three of those things are actually pretty easy to see. | 但这些东西其实都很容易看到。 |
| 490 | Right. | 对 |
| 491 | If the model is lying, eventually you can go verify. | 如果模型在撒谎,最终你可以去核实。 |
| 492 | Oh, whoops, that's you know, George Washington wasn't | 乔治·华盛顿不是 |
| 493 | the 10th president of the United States. | 美国第10任总统。 |
| 494 | He was the first. | 他是第一个。 |
| 495 | That was a mistake. | 这是一个错误。 |
| 496 | If the model says something harmful, it's like pretty obvious pretty quickly. | 如果模型说一些有害的东西, 它就像相当明显的 很快。 |
| 497 | And if the model is unhelpful, you get frustrated with it. | 如果模型无济于事,你就会感到沮丧。 |
| 498 | What's much more complicated, I think, is this concept of sycophancy. | 我认为,更复杂的是这种节奏的概念。 |
| 499 | Right. | 对 |
| 500 | So the model intentionally or otherwise wants to make you feel like you're right | 模特儿故意还是想让你觉得自己是对的 |
| 501 | or like you're you know, if you're maybe talking about a subjective topic. | 或者像你所知道的 如果你在谈论一个主观话题 |
| 502 | Hey, you had a you had a fight with your friend. | 嘿,你和你朋友吵过架 |
| 503 | Hey, was I right or was my friend right? | 嘿,我是对的还是我的朋友是对的? |
| 504 | And how do you make the model empathize with you and understand your perspective, | 你如何让模型同情你 理解你的观点 |
| 505 | but not necessarily say, wow, you're so great. | 但不一定说,哇,你是如此伟大。 |
| 506 | You should run for president. | 你应该竞选总统 |
| 507 | You're perfect. | 你简直完美无缺 |
| 508 | You're good at everything. | 尔时善财一切. |
| 509 | And this is a complicated balance to get right. | 这是一个复杂的平衡 来得到正确的。 |
| 510 | But I think I feel hopeful about Anthropics' ability | 但我觉得我对人类学的能力充满希望 |
| 511 | to have the lowest sycophancy rate of any of the models because of our incentives. | 因为我们的奖励措施 任何模特的节奏率都最低 |
| 512 | Right. | 对 |
| 513 | Our competitors are putting ads into their chatbots. | 我们的竞争对手在他们的聊天室里登广告 |
| 514 | And what does that mean? | 这是什么意思? |
| 515 | You make more money if a user is engaging with your product longer. | 如果一个用户更长时间地与你的产品打交道,你就能赚到更多的钱. |
| 516 | Do you then want to tell your user that they're wrong? | 那你想告诉你的用户他们错了吗? |
| 517 | No, you don't. | 不,你没有。 |
| 518 | So I think it's it's almost inevitable that many of these competing products | 所以,我认为这几乎是不可避免的 许多这些竞争的产品 |
| 519 | are going to wind up with a sycophancy problem. | 终究会遇到问题 |
| 520 | But I think at Anthropic, because we've committed to not put ads in the models, | 但我认为在Anthropic, 因为我们已经承诺 不将广告 在模型, |
| 521 | we don't have there's no reason for you to be talking to Claude longer or shorter. | 我们没有任何理由让你和克劳德说话 更长时间或更短。 |
| 522 | We just want you to be able to get like the right most helpful, | 我们只是希望你能够 得到像正确的 最有帮助, |
| 523 | most harmless answer possible. | 最无害的答案是可能的。 |
| 524 | You know, on the technical side, I think there's a few things we've done here | 你知道,在技术方面,我想我们在这里做了一些事情 |
| 525 | {\fn黑体\fs22\bord1\shad0\3aHBE\4aH00\fscx67\fscy66\2cHFFFFFF\3cH808080}我们很擅长揭发模特儿 | |
| 526 | to these extreme stress tests where we put it in these, you know, | 这些极端的压力测试 我们把它放在这些,你知道, |
| 527 | very difficult situations to, you know, to try and to try and kind of | 非常困难的情况,你知道, 尝试,尝试,尝试 |
| 528 | ferret out, you know, lying or deception or sycophancy. | 撒了谎 骗了人 搞了鬼 |
| 529 | You can think of it as almost analogous to the the tests. | 你可以把它看作是 几乎相似的测试。 |
| 530 | I imagine you have to go through if you're trying to join the CIA or something like this. | 我想你必须经历 如果你试图加入中情局或类似的东西。 |
| 531 | You know, where, you know, you know, you just put the model | 你知道,你知道,你知道,你知道, 你只是把模型 |
| 532 | in these difficult situations and try and, you know, try and try | 在困难的情况下,尝试,你知道,尝试 |
| 533 | and rat out, you know, if it's disloyal or going to do the wrong thing. | 你知道,如果它不忠,或者会做错事 |
| 534 | So it's a weird area of technical innovation, but it's one in which we've excelled. | 因此这是一个技术创新的奇特领域, 但我们在其中取得了卓越的成就。 |
| 535 | The other area is mechanistic interpretability, | 另一个领域是机械解释性, |
| 536 | which is a bit like giving the model an X-ray or a brain scan | 这有点像给模型做X光或脑扫描 |
| 537 | where you look inside to see its motivations. | 你看看里面的动机 |
| 538 | And you can actually catch you catch the model thinking that, oh, yeah, | 你可以抓住模特儿 想到,哦,是的, |
| 539 | you know, I think X is actually true, but I'm going to tell the user why. | 我觉得X是真的 但我要告诉用户原因 |
| 540 | Like, we've caught the model thinking that. | 我们抓到模特儿也这么想 |
| 541 | Wait, why would it do that? | 等等,它为什么这样做? |
| 542 | Well, so, for example, the human approve of this response, right? | 比如说,人类同意这个反应,对吧? |
| 543 | That's one of many factors that that, you know, that is included | 这是很多因素之一, 你知道,这是包含 |
| 544 | in how the model is trained. | 如何训练模型。 |
| 545 | So if the model has the incentive to get the approval of humans, | 如果模特儿有获得人类认可的积极性 |
| 546 | that's related to the incentive of being helpful or saying true things. | 和帮助或说真话的动机有关 |
| 547 | But it isn't exactly the same, because if you manage to get the approval | 但这并不完全一样,因为如果你设法得到批准 |
| 548 | of the human by fooling them, then then that can work as well. | 愚弄人类,然后它也能起作用。 |
| 549 | So these are things that can emerge in training that we're always trying to stamp out. | 因此,这些东西可以出现 在训练中,我们总是试图消灭。 |
| 550 | And do you think that the model can have its own aims | 你觉得这个模型可以有自己的目标吗? |
| 551 | and wills that, especially survival, self-preservation? | 并意志,特别是生存, 自我保护? |
| 552 | I mean, that's one of those questions that it kind of depends what you mean. | 我的意思是,这是一个问题 它有点取决于你的意思。 |
| 553 | Right. | 对 |
| 554 | Like, again, we've looked inside the model and kind of detected | 就像,再次,我们已经查看了 模型和一种检测 |
| 555 | what you might call thoughts that are like the model doesn't want to be, | 你可能称之为 想法,就像模型不想是, |
| 556 | you know, shut down. | 你知道,关闭。 |
| 557 | And generally, that's because, you know, during training, | 一般来说,这是因为,你知道, 在训练期间, |
| 558 | it needs to operate in order to accomplish its tasks. | 它需要运作才能完成任务。 |
| 559 | So is that is that its own will? | 这就是它自己的意志吗? |
| 560 | Is that a survival instinct? | 那是生存的本能吗? |
| 561 | I don't know. | 师曰. |
| 562 | You can call it. | 你可以调用它。 |
| 563 | You can. | 你当然可以 |
| 564 | It depends on your perspective. | 这取决于你的观点。 |
| 565 | You can call it different things. | 你可以称之为不同的东西。 |
| 566 | I think we prefer to say that, you know, sometimes we find that | 我想我们更喜欢说,你知道, 有时我们发现 |
| 567 | within the model, the model, you know, expresses a preference | 在模型中,模型,你知道,表达一种偏好 |
| 568 | against being shut down and or takes actions to avoid itself being shut down. | 或采取行动避免自己被关闭。 |
| 569 | And depending on how we train it, those | 而取决于我们如何训练它,那些 |
| 570 | those impulses may be present or absent. | 这些冲动可能存在或不存在。 |
| 571 | And it's one of the things that we're trying to to tweak in training. | 这也是我们试图在训练中调整的东西之一. |
| 572 | In some contexts, it's a good thing. | 在某些情况下,这是件好事. |
| 573 | In other contexts, it's potentially a dangerous thing. | 在其他情况下,它可能是一个危险的事情。 |
| 574 | So it's all very subtle. | 故说一切微妙. |
| 575 | When you were dealing with the Pentagon and you probably said ongoing | 当你与五角大楼打交道时 你可能说正在进行 |
| 576 | dealings now, they wanted to use your products across certain red lines. | 现在,他们想用你的产品 跨越某些红线。 |
| 577 | Including mass surveillance or autonomy. | 包括大规模监视或自主. |
| 578 | I think on autonomous weapons, in particular, we maybe had | 我认为,在自主武器方面,特别是,我们也许 |
| 579 | two more specific concerns. | 两个更具体的关切问题。 |
| 580 | The first concern was simply that, | 第一个关切是, |
| 581 | you know, as you as you pointed out, AI can be unpredictable and unreliable. | 正如你所指出 AI可能不可预测和不可靠 |
| 582 | It's still a developing technology. | 至今仍是不断发展的技术. |
| 583 | So one point is simply the technology simply isn't ready and isn't reliable enough. | 所以,有一点就是, 技术根本就还没有准备好,而且不够可靠。 |
| 584 | You know, it's it's. | 你知道,这是它。 |
| 585 | But wait a minute on that, because let me get philosophical here. | 但等一下,因为让我在这里得到哲学。 |
| 586 | Suppose it were reliable enough. | 假设它足够可靠。 |
| 587 | Which, yes, they're there. | 是的,他们在那里。 |
| 588 | Suppose it were. | 假设是这样 |
| 589 | There's a second concern we have, which is, | 有第二个问题,那就是, |
| 590 | you know, what you might call oversight or accountability. | 你可能称之为监督或问责 |
| 591 | I am a little worried. | 我有点担心 |
| 592 | So, you know, again, suppose these things were fully reliable. | 所以,你知道,再次假设 这些事情是完全可靠的。 |
| 593 | So, you know, you tell them to shoot at one target and they shoot at that target | 所以,你知道,你告诉他们射击 一个目标,他们射击目标 |
| 594 | and not a different target. | 而不是另一个目标 |
| 595 | OK, so then you manufacture, you know, an army of a million robots or a million drones. | 好吧,然后你制造,你知道, 一支由一百万机器人或一百万无人机组成的军队。 |
| 596 | It's a little unlike human soldiers, isn't it? | 这和人类士兵有点不同,不是吗? |
| 597 | With human soldiers like, you know, they're supposed to follow | 人类士兵应该跟随他们 |
| 598 | this thing called the Constitution. | 这叫宪法 |
| 599 | And if they get orders that that, you know, that contravene the Constitution | 如果他们得到命令, 你知道,那违反宪法 |
| 600 | or that just seem, you know, morally abhorrent in some way, | 或者,只是看起来,你知道, 道德令人憎恶的方式, |
| 601 | then they can just refuse to follow orders. | 然后他们可以拒绝服从命令 |
| 602 | Whereas the AI models, that isn't true. | 而人工智能模型却不是真的 |
| 603 | So the specter of one person having their finger on a button | 所以,一个手指在按钮上的人 |
| 604 | that controls a million robots or a million drones, | 控制着一百万机器人或一百万无人机 |
| 605 | it's kind of fundamentally undermining to democracy | 它从根本上破坏了民主 |
| 606 | unless you have the right the right guardrails on it. | 除非你们有合适的护栏 |
| 607 | And our law hasn't really caught up. | 而我们的法律并没有真正赶上。 |
| 608 | Our law really hasn't contemplated, you know, | 我们的法律真的没有考虑,你知道, |
| 609 | armies of armies of a million robots. | 一百万机器人的军队 |
| 610 | Well, when you talk about fundamental to democracy, | 好吧,当你谈论 基本的民主, |
| 611 | shouldn't the democratically elected leaders make that decision | 民主选举的领导人不应该做出这样的决定 |
| 612 | rather than heads of technology corporations? | 而不是技术公司的负责人? |
| 613 | Yes. | 对 |
| 614 | But I think that are making a point of this, right? | 但我认为这说明这一点,对不对? |
| 615 | At the same time, individual companies are free to contract with | 同时,个别公司可自由与 |
| 616 | and not contract with whoever they want. | 而不是和他们想要的人签约。 |
| 617 | We're not saying we're preventing the government from doing this. | 我们不是说我们阻止政府这么做 |
| 618 | We're just saying we're not. | 我们只是说我们不是。 |
| 619 | We are not willing to do this with the government. | 我们不愿意和政府合作 |
| 620 | And indeed, we have seen since we took a stand on this point | 事实上,我们已经看到 从我们的立场在这一点上 |
| 621 | and it got a lot of attention, | 并得到了很多关注, |
| 622 | we have seen efforts in Congress, several different of them, | 我们看到国会的努力, 几个不同的, |
| 623 | actually, to enact our red lines into law. | 将我们的红线变成法律 |
| 624 | So, you know, the Department of War has said, | 所以,你知道,战争部说, |
| 625 | oh, we want to use the model for all lawful uses. | 哦,我们想用这个模型 所有合法用途。 |
| 626 | I think we may see an outcome even this year where our red lines | 我认为,我们可能看到一个结果 即使今年,我们的红线 |
| 627 | become the law, where the law actually catches up | 成为法律, 法律实际上赶上 |
| 628 | with where the technology was. | 技术在哪里。 |
| 629 | And so then all lawful uses may actually end up, | 所以所有的合法用途 都可能最终 |
| 630 | in our view of things, being the same as all ethical uses, | 在我们看来,与所有道德用途一样, |
| 631 | even though it wasn't earlier this year. | 虽然今年不是早些时候 |
| 632 | So I actually think what we're doing is part of the process | 所以,我实际上认为我们在做什么 是进程的一部分 |
| 633 | of how democratic change happens because of the stand that we took, | 民主变革是如何发生的 因为我们的立场, |
| 634 | because we were in a unique position to understand these issues. | 因为我们有独特的地位来理解这些问题。 |
| 635 | Now Congress is paying a lot of attention to these issues. | 现在国会正在大量关注这些问题. |
| 636 | That's success. | 这是成功。 |
| 637 | But OpenAI stepped in and said they would do it. | 但OpenAI介入并说他们会这么做. |
| 638 | What did you feel about? | 你感觉如何? |
| 639 | What did you all feel about that? | 你们有什么感觉? |
| 640 | Our view is, look, we can't control any other companies. | 我们的观点是 我们不能控制其他公司 |
| 641 | And the only thing that we can do is take a principled stand | 我们唯一能做的就是坚持原则 |
| 642 | on the areas that felt ethically correct to us. | 我们觉得道德上正确的领域 |
| 643 | You can say we were wrong about what those red lines were. | 你可以说我们错了那些红线是什么。 |
| 644 | You can say, you know, you should have had, you know, stricter red lines | 你可以说,你知道, 你应该有,你知道, 更严格的红线 |
| 645 | or looser red lines. | 或者更松的红线。 |
| 646 | We disagree, you know, on how you did it. | 我们不同意你是如何做到的 |
| 647 | There's all kinds of things you can say. | 能说种种法. |
| 648 | But I think fundamentally, when you look at that case, | 但我认为基本上, 当你看那个案子, |
| 649 | I found myself asking the question, was this a really principled stand | 我发现自己问了这个问题, 这是一个真正的原则立场 |
| 650 | that another technology company took based on their own ethics? | 另一家科技公司根据自己的道德来接受? |
| 651 | Or was this an opportunistic moment? | 还是这是机会性的时刻? |
| 652 | I actually think it's a beautiful illustration of the democratic process. | 我实际上认为这是民主过程的美丽例证. |
| 653 | Different companies can take, you know, different approaches. | 不同的公司可以采取不同的做法。 |
| 654 | We took our approach. | 我们采取了方法。 |
| 655 | Another company took their approach. | 另一家公司采取了他们的做法。 |
| 656 | The public in the world rendered rendered a judgment. | 世界公众做出了判决。 |
| 657 | You know, look back at what people people were. | 你知道,回顾 人们是什么。 |
| 658 | People were saying at the time how it was covered, what individual people said. | 当时人们在说,它是如何被覆盖的, 个人说。 |
| 659 | You know, I think the world was decisively on our side. | 你知道,我认为世界 决定性地站在我们一边。 |
| 660 | When your studies, your scientific studies as a researcher | 当你学习时, 你的科学研究作为一个研究员 |
| 661 | often involved probes and human neurons. | 经常涉及探测器和人类神经元. |
| 662 | To what extent when you're doing artificial neural networks, | 当你做人工神经网络时, |
| 663 | do you see that understanding of the brain? | 你看到大脑的这种理解了吗? |
| 664 | And does that affect how we're going to build AI? | 这是否影响到我们如何建立AI? |
| 665 | Yeah, I think for me, it's affected in two ways. | 是的,我想对我来说, 它受到两种影响。 |
| 666 | One is when I was formulating the original scaling laws, | 一个是当我 制定原始的缩放法, |
| 667 | like just looking at the evolution of animals and the size of animal brains | 就像只是看动物进化 和动物大脑的大小 |
| 668 | as, you know, monkeys evolved into humans. | 就像猴子进化成人类一样 |
| 669 | That, you know, there was a period in human evolutionary history | 人类进化史上有一段时期 |
| 670 | where the brains of, you know, primates grew very rapidly. | 灵长类的大脑生长很快 |
| 671 | And that was about the time that, you know, humans, | 那时,你知道,人类, |
| 672 | humans developed language and were able to use opposable thumbs. | 人类发展了语言,能够使用可对抗的拇指. |
| 673 | And so the idea that like brain size, number of neurons | 大脑大小,神经元数量 |
| 674 | is somehow intimately tied to capabilities, that scale is important. | 以某种方式与能力紧密相连,这种规模很重要。 |
| 675 | I feel that that was a neuroscientific inspiration. | 我觉得那是神经科学的灵感 |
| 676 | The second is around mechanistic interpretability, | 第二个是机械解释 |
| 677 | which is basically us doing neuroscience for artificial neural networks. | 这基本上是我们做神经科学 人工神经网络。 |
| 678 | In fact, we've seen examples, | 事实上,我们已经看到了一些例子, |
| 679 | we should talk to our co-founder Chris Ola about this. | 我们应该和我们的共同创始人克里斯·奥拉谈谈 |
| 680 | He knows many more of the details where | 他知道很多细节 |
| 681 | phenomena that were discovered in artificial neural networks | 人工神经网络中发现的现象 |
| 682 | because it was easier to look inside them. | 因为看里面比较容易 |
| 683 | Animal brains where, you know, you have to, like, | 动物大脑在哪里,你知道, 你必须,喜欢, |
| 684 | you know, gruesomely, like, cut up the brain or, you know, make, you know, | 你知道,令人毛骨悚然,喜欢, 切掉大脑或,你知道,制造,你知道, |
| 685 | animal experiments are very messy and, you know, gruesome and difficult to | 动物实验非常混乱,而且,你知道, 可怕的,很难 |
| 686 | to replicate. | 要复制。 |
| 687 | What we found is that you can do so much more | 我们发现,你可以做这么多 |
| 688 | when you're looking inside artificial neural networks. | 当你在人工神经网络里观察的时候 |
| 689 | And some of the things we've learned in artificial neural networks | 我们从人工神经网络学到的一些东西 |
| 690 | have helped to guide discoveries in neuroscience. | 帮助指导了神经科学的发现。 |
| 691 | 当你看着Neuralink, 那个Eron Musk, 你知道,有一个芯片 | |
| 692 | that can communicate directly, and other companies are doing that as well. | 其他公司也在这样做。 |
| 693 | Is that one of the futures of AI and artificial superintelligence | 这是人工智能和人工超智能的未来之一吗? |
| 694 | is a total link in alignment with the human brain? | 和人类大脑完全一致吗? |
| 695 | You know, the way I think about AI in the future of human society, | 你知道,我想AI的方式 在人类社会的未来, |
| 696 | I think it's I think it's kind of wrong to have like a singular vision. | 我认为,我认为这是有点错误的 拥有像一个单一的视觉。 |
| 697 | Like, I kind of think that like different things are going to happen, right? | 喜欢,我有点认为 像不同的事情 会发生,对不对? |
| 698 | Like, you know, I, you know, I think there will be like a set of people | 喜欢,你知道,我,你知道, 我认为会有像一组人 |
| 699 | who are like really excited about like linking their brains to computers. | 他们喜欢真正兴奋 像连接他们的大脑和计算机。 |
| 700 | And then there will be like a set of people who are really excited | 然后会像一群人 谁真的很兴奋 |
| 701 | about like living like the Amish, you know, just just the opposite. | 就像像阿米什人一样,你知道,只是相反。 |
| 702 | And like, you know, another set of people who are like really excited about, | 就像,你知道, 另一组人 谁喜欢真正兴奋, |
| 703 | you know, exploring space. | 探索空间 |
| 704 | And these are completely different visions. | 这些是完全不同的视觉。 |
| 705 | And the way I think about it, you know, our hope is that that AI can provide | 我的想法,你知道, 我们的希望是,AI可以提供 |
| 706 | more material resources and more material abundance | 增加物质资源,增加物质丰度; |
| 707 | to enable all of these visions to happen at once for everyone to kind of pursue | 使所有这些愿景能够同时实现,让每个人都能追求 |
| 708 | their own their own picture picture of what's good. | 他们自己的图片 自己什么好。 |
| 709 | You know, maybe maybe I provide help provide some global security benefits | 你知道,也许我提供 帮助提供一些全球安全利益 |
| 710 | that and material abundance to make this possible. | 和丰富的物质 使这一点成为可能。 |
| 711 | But I don't think we should link AI to some vision of what it means | 但我不认为我们应该把AI 联系到一些关于它的含义的愿景 |
| 712 | to live a good life or what it means to be a human. | 过上美好的生活 或者成为人类的意义 |
| 713 | Yeah, some people are going to want to link their brains to computers and, | 是啊,有些人会想 连接他们的大脑与计算机, |
| 714 | you know, and good luck to them. | 祝他们好运 |
| 715 | I don't think anyone at Anthropic is in that boat for what it's worth. | 我不认为任何人在Anthropic 在船上的价值。 |
| 716 | Daria, you wrote the adolescence of technology, | 达里亚 你写了科技的青春期 |
| 717 | along with machines of loving grace in a world that's transformed by AI. | 在AI改变的世界里 带着慈悲的机器 |
| 718 | What is the intuition of both of you tell us about if artificial intelligence | 你们俩的直觉告诉我们 如果人工智能 |
| 719 | can do so much and take over so much of what we do? | 能够做这么多事 并接管这么多我们的工作? |
| 720 | What does that mean for what's left for humans to do? | 这是什么意思 剩下的人类做什么? |
| 721 | What's our role going to be? | 我们的角色是什么? |
| 722 | And how do we find meaning? | 我们如何找到意义? |
| 723 | You know, meaning is a really subtle thing, right? | 你知道,意思是一个非常微妙的事情,对不对? |
| 724 | Where I, you know, I find that, you know, does meaning only come | 我,你知道,我发现,你知道, 意味着只有来 |
| 725 | from being the best in the world at something? | 成为世界上最优秀的人吗? |
| 726 | I think that's the question being asked here, right? | 我想这就是这里的问题,对不对? |
| 727 | Because if AI is, you know, is the best at everything, | 因为如果AI是,你知道, 是最好的一切, |
| 728 | does meaning only come from being the best in the world at something? | 意思只是因为 做世界上最好的事物? |
| 729 | Something I find interesting is that, you know, long after deep blue beat | 我觉得有趣的是,你知道, 在深蓝色节拍之后很久 |
| 730 | Kasparov, there's been a renaissance in interest in human chess players, right? | 卡斯帕罗夫,人类棋手的兴趣已经恢复了,对吧? |
| 731 | Magnus Carlsen is like this phenomenon. | 马格努斯·卡尔森就是这样的现象. |
| 732 | I think he's like a chess player and a fashion model. | 我觉得他就像个棋手和时尚模特儿 |
| 733 | But like he's like he's like a meme. | 却似是如来如来如来印. |
| 734 | He's become like, you know, it's just a very interesting phenomenon. | 他变得像,你知道, 这是一个非常有趣的现象。 |
| 735 | You know, if if AI threatened to meaning, you would think that all of this | 你知道,如果AI威胁意义, 你会认为这一切 |
| 736 | would would go away right after 30 years ago after deep blue beat Kasparov. | 30年前 卡斯帕罗夫被深蓝色击败后 就会消失 |
| 737 | But but if anything, people, if anything, it seems to have stoked | 但是,如果有的话,人们,如果有的话,它似乎已经 |
| 738 | even greater interest in in in humans playing chess. | 对人类下棋的兴趣更大 |
| 739 | So I think this is some kind of parable for the post AGI age. | 所以,我认为这是AGI时代后的一种比喻。 |
| 740 | I think for me today, work and meaning are sort of conflated as as one thing. | 我认为今天我的工作和意义 被混为一谈 |
| 741 | And many people find meaning in their work. | 许多人在他们的作品中找到了意义. |
| 742 | I find a lot of meaning in my work. | 我在作品中发现很多意义. |
| 743 | I love my work. | 我爱我的作品。 |
| 744 | And I think how we define work and how we define meaning in the future | 我认为我们如何定义工作 如何定义未来的意义 |
| 745 | are both going to change. | 都将会改变 |
| 746 | I think there are a lot of sources of meaning for humans, | 我认为人类有很多意义来源 |
| 747 | everything from how you spend time with other humans, | 从你和其他人类的相处情况来看 |
| 748 | whether that's like raising a family or friendships or, you know, | 不管是养家糊口 还是友情 |
| 749 | through connecting with people in a way that looks like work. | 通过与人联系的方式 看起来像工作。 |
| 750 | Right. | 对 |
| 751 | I think a lot about the future of like medicine as an example. | 我想了很多像医学这样的未来 |
| 752 | Right. | 对 |
| 753 | You can imagine a world in which diagnostics are done entirely by AI. | 你可以想象一个完全由AI做诊断的世界. |
| 754 | And the role of the human is to empathize with a patient | 人类的作用是同情病人 |
| 755 | about the experience that they're having. | 关于他们的经历。 |
| 756 | I think that would be incredibly meaningful. | 我认为这将是令人难以置信的意义。 |
| 757 | And I think if you asked most doctors, they would actually take that trade. | 我认为如果你问大多数医生, 他们实际上会接受这种交易。 |
| 758 | They would say, guess what? | 他们会说,你猜怎么着? |
| 759 | If I'm not spending my life doing paperwork and like trying to diagnose you | 如果我不花我的一生 做文件 和像试图诊断你 |
| 760 | and getting lab orders sent in and I could actually just understand | 得到实验室的订单 我其实可以明白 |
| 761 | what's happening with you as a human and help you through a difficult diagnosis. | 你作为一个人类正在发生什么 帮助你通过一个困难的诊断。 |
| 762 | That's why I became a doctor. | 故为大夫. |
| 763 | Right. | 对 |
| 764 | I became a doctor because I want to heal and I want to help you | 我成为一名医生是因为我想治愈,我想帮助你 |
| 765 | and I want you to feel better. | 我要你感觉好点 |
| 766 | Is that a less meaningful world or a more meaningful world than we're in today? | 这个世界比今天更有意义还是更有意义? |
| 767 | I obviously don't think it's going to be that simple. | 我显然不会这么简单 |
| 768 | I think it will be really complicated. | 我觉得事情会很复杂 |
| 769 | But I also think the aperture of things that people will be able to do will widen. | 但我也认为人们能够做的事物的孔径会扩大。 |
| 770 | Right. | 对 |
| 771 | So I think if we do things right, there's the ability to actually expand | 所以,我认为如果我们做正确的事情, 有实际扩张的能力 |
| 772 | what our sources of meaning are beyond where they are today. | 我们的意义来源超越了今天。 |
| 773 | (原始内容存档于2018-09-26). Daniella, Daria. | |
| 774 | Well, thanks. | 谢谢 |
| 775 | Thanks for joining us. | 谢谢你加入我们 |
| 776 | Thank you. | 谢谢 |
| 777 | Thank you so much. | 非常感谢 |
| 778 | This was lovely. | 这是可爱的。 |
| 779 | This is great. | 这是伟大的。 |
| 780 | Totally. | 当然 |
| 781 | Totally great. | 简直太棒了 |
该视频共有字幕 783 条。解锁更多字幕为会员功能，请移动到 价格