【中英双语】彭博社:李开复:中国AI模型够用又便宜
字幕摘录
| 时间 | 英文 | 中文 |
|---|---|---|
| 0:00 | Kaifu Li has been president of Google China, held senior roles at Microsoft and Apple, | 李开富曾任Google中国总裁,曾在微软和苹果公司担任高级角色, |
| 0:04 | 现为Cinevation Ventures和O1.ai的首席执行官. | |
| 0:09 | Well, the AI pioneer is also a best-selling author, and his new book, AI Native, | AI先驱也是畅销的作者 他的新书"AI Indigen" |
| 0:14 | looks at how business leaders should navigate the tech transformation. | 企业领导人应该如何引导技术转型。 |
| 0:17 | Well, I'm delighted that Kaifu Li is joining us now. | 很高兴李开福现在加入我们 |
| 0:20 | Kaifu, thank you so much for joining us. | 开福,谢谢你加入我们 |
| 0:22 | It's a great book. | 其书甚盛. |
| 0:23 | It's a great read. | 诵经甚妙. |
| 0:24 | It's short. | 其脉短. |
| 0:25 | I urge everyone who's a business leader to actually go and read it. | 我敦促每个身为商业领袖的人 都去读一下 |
展开字幕全文(277 条)
| 序号 | 英文 | 中文 |
|---|---|---|
| 1 | Kaifu Li has been president of Google China, held senior roles at Microsoft and Apple, | 李开富曾任Google中国总裁,曾在微软和苹果公司担任高级角色, |
| 2 | 现为Cinevation Ventures和O1.ai的首席执行官. | |
| 3 | Well, the AI pioneer is also a best-selling author, and his new book, AI Native, | AI先驱也是畅销的作者 他的新书"AI Indigen" |
| 4 | looks at how business leaders should navigate the tech transformation. | 企业领导人应该如何引导技术转型。 |
| 5 | Well, I'm delighted that Kaifu Li is joining us now. | 很高兴李开福现在加入我们 |
| 6 | Kaifu, thank you so much for joining us. | 开福,谢谢你加入我们 |
| 7 | It's a great book. | 其书甚盛. |
| 8 | It's a great read. | 诵经甚妙. |
| 9 | It's short. | 其脉短. |
| 10 | I urge everyone who's a business leader to actually go and read it. | 我敦促每个身为商业领袖的人 都去读一下 |
| 11 | Thank you. | 谢谢 |
| 12 | Are we approaching an era where a company needs to literally fundamentally reinvent its business | 我们是不是要进入一个时代 公司需要彻底地 重塑企业 |
| 13 | instead of adapt to AI or technology? | 而不是适应AI或技术? |
| 14 | Yes, yes. | 对,对。 |
| 15 | The time has come. | 现在是时候了。 |
| 16 | If people will think back when internet first arrived, | 如果人们会回想 当互联网刚到来的时候, |
| 17 | and then most people just glue internet to the company, use a browser, put up a website, | 然后大多数人只是粘上 互联网到公司, 使用浏览器, 建立一个网站, |
| 18 | and then if that's all they did, they're dead now. | 如果他们只这么做 他们就死定了 |
| 19 | Now, AI revolution is 10 times bigger and 10 times faster than the internet revolution, | 现在,AI革命比互联网革命大10倍,快10倍, |
| 20 | and now I see most CEO are just gluing AI, do an agent here, do customer service, | 现在,我看到大多数CEO 只是粘着AI, 在这里做一个代理,做客户服务, |
| 21 | and those are window dressing. | 那些是窗纱 |
| 22 | They are not meaningful AI transformation. | 它们不是有意义的AI转变。 |
| 23 | So what distinguishes these companies that I guess are AI native? | 那么,是什么区别 这些公司,我想是AI的本土? |
| 24 | And honestly, | 说实话 |
| 25 | it's very difficult because if you're a large corporation, how do you do it? | 很难 因为如果你是大公司 你怎么做到的? |
| 26 | I mean, do you really start from the bottom or from the top? | 我的意思是,你真的从下面开始 还是从上面开始? |
| 27 | Okay, so two questions. | 好吧,那么两个问题。 |
| 28 | So who are the AI native companies? | 谁是AI本地公司? |
| 29 | I will tell you now, | 我现在告诉你 |
| 30 | only the AI companies are AI native because we understand what it means. | 只有AI公司是AI本地人,因为我们理解它的含义. |
| 31 | Just like only the internet companies like early days, Google, Amazon, became internet centric. | 就像早期的互联网公司一样, Google, Amazon 成为了互联网的中心。 |
| 32 | So a company can do it early and lead the competition or can do it late and watch it | 所以公司可以早做,领导比赛,也可以晚做看 |
| 33 | become too late. | 变得太晚了。 |
| 34 | Now, how to do it? | 现在,如何做到这一点? |
| 35 | Well, ultimately, AI native means if you remove the AI, | 好吧,归根结底,AI本土的意思是 如果你删除AI, |
| 36 | the company falls apart. | 公司倒闭。 |
| 37 | It means the AI is the core thing that's running the company where | 也就是说人工智能是管理公司的核心 |
| 38 | everything is interconnected. | 一切都是相互关联的。 |
| 39 | And also it means the AI will learn from itself and become better | 也意味着人工智能会从自己身上吸取教训 变得更好 |
| 40 | and better at servicing our company's needs. | 更能满足我们公司的需求 |
| 41 | So doing something that big is a fundamental change. | 所以做这样大的事情是一个根本性的变化。 |
| 42 | It means we need to, the way we make decisions, the way we have meetings, the way we have | 这意味着我们需要,我们做决定的方式, 我们开会的方式,我们拥有的方式。 |
| 43 | organizational charts need to all change. | 组织结构图需要全部改变。 |
| 44 | Therefore, it can only be done from the top. | 因此,它只能从头来做。 |
| 45 | But Kaifu, we have a lot of chief executives watching, right? | 但是开福,我们有很多主管在看,对吧? |
| 46 | And then I speak to a lot | 然后,我对很多 |
| 47 | of chief executives that say, I understand, I just don't know how to start. | 首席执行官说,我知道, 我只是不知道如何开始。 |
| 48 | So if you had to give | 所以,如果你必须给 |
| 49 | the leader of a FTSE 100 company or a big American firm today three priorities on what they need to | 一家FTSE 100公司或一家美国大公司的领导人 今天有三个优先事项 他们需要什么 |
| 50 | do in the next 12 months, what are they? | 接下来的12个月,他们是什么? |
| 51 | Well, first is to fully recognize and understand | 首先要充分认识和理解 |
| 52 | this is not another piece of software or tool. | 这不是另一个软件或工具。 |
| 53 | This is about your future of your company and | 这是关于你的未来 你的公司和 |
| 54 | AI transformation is a required mandate, that understanding. | 人工智能转换是一项必要任务,即这种理解。 |
| 55 | Second is to read the book, | 第二是读这本书, |
| 56 | AI International,它给出了精确的一,二,三,四步路线图. | |
| 57 | Can't guarantee step two, | 不能保证第二步, |
| 58 | three, four are perfect, but one is absolutely right. | 三,四是完美的,但一个是绝对正确的。 |
| 59 | And that is using a product that we call | 这就是使用一种产品 我们称之为 |
| 60 | Boss AI to really drive, own all the data in your company, know all that's going on and make the | AI老大负责开车 拥有你公司的所有数据 了解所有情况 |
| 61 | decisions. | 决定。 |
| 62 | The last advice really is if that's overwhelming to a CEO, then hire a chief AI officer, | 最后的忠告是 如果这对首席执行官来说是压倒一切的 那么就雇个AI主管 |
| 63 | someone who understands AI and a bit about business, does not need to be a researcher or a | 一个懂AI和一点生意的人 不需要做研究员或 |
| 64 | super expert. | 超级专家。 |
| 65 | Why not the CIO? | 为什么不是CIO? |
| 66 | Some CIOs will also be fine, but CIOs are usually maintaining status | 一些CIO也会很好,但CIO通常保持状态 |
| 67 | quo. | 事实。 |
| 68 | CAIO and the CEO have the aligned goal of transforming the company. | CAIO和CEO有着改造公司的统一目标. |
| 69 | You must get asked all | 你一定要被问遍 |
| 70 | the time, if AI is being commoditized, then how do you keep the competitive edge for your business? | 如果AI被商品化 那你如何保持竞争优势? |
| 71 | For my company or for the CEO we speak to? | 为了我的公司还是首席执行官? |
| 72 | For a CEO that you're speaking to. | 对于一个CEO 你说话。 |
| 73 | Yeah, so the CEO | 是啊,所以首席执行官 |
| 74 | must adopt AI early. | 必须及早通过大赦国际。 |
| 75 | That will give you an early mover advantage and that may help you gain market | 这会给你一个早期的移动者优势, 这可能有助于你获得市场 |
| 76 | share, brand awareness, but more importantly, it starts the AI engine being built and growing | 分享,品牌意识,但更重要的是, 它开始AI引擎 正在建设和成长 |
| 77 | because ultimately the only significant moat any company can have is adopting AI early and letting | 因为最终任何公司所能拥有的唯一重要的护城河 就是提早采用人工智能并让 |
| 78 | the data accumulate and letting that data train your AI to be better than other people's AI. | 数据积累,让数据训练 你的AI比别人的AI更好。 |
| 79 | But this is what for distribution, intellectual property, because a lot of the companies always | 但是,这是分配,知识产权, 因为很多公司总是 |
| 80 | put the customer first no matter what they do. | 不管他们做什么 都把顾客放在第一位 |
| 81 | Does that work in an AI world? | 这在人工智能世界有用吗? |
| 82 | That a commercial company? | 那是商业公司吗? |
| 83 | Yeah. | 对 |
| 84 | Yes, well the AI is a flexible device that it can help you train | 是的 人工智能是一种灵活的设备 它能帮助你训练 |
| 85 | to your culture and your business needs. | 你的文化和商业需要。 |
| 86 | If you want to really make customer happy, that's number | 如果你真的想让顾客开心 那就是数字 |
| 87 | one goal, then that would be measured in the decision support and the data that you get. | 一个目标,然后在决定支持和数据中衡量。 |
| 88 | If you want to go profit first, market share first, it will adapt to what your company goals are. | 如果你想先赢利,先得市场份额, 它会适应你公司的目标。 |
| 89 | I know you wrote the book because a lot of your executives come to you to ask questions. | 我知道书是你写的 因为你的很多高管来问你问题 |
| 90 | What's the question that comes across the most often that they ask you? | 他们最常问你的问题是什么? |
| 91 | Well, they used to ask which model, now they realize it doesn't matter. | 嗯,他们曾经问 哪个模式, 现在他们意识到 这不重要。 |
| 92 | They often still, | 他们经常仍然 |
| 93 | the top models are all roughly equal, even open source models, and they now know that. | 顶级模型都大致相同, 甚至开源模型, 他们现在知道这一点。 |
| 94 | The second question they ask is, how do I avoid the large language model companies from taking | 他们问的第二个问题是 我如何避免大语言模型公司采取 |
| 95 | all my data and we don't know what they'll do with it. | 所有的数据,我们不知道 他们会怎么做。 |
| 96 | We know they won't steal it, | 我们知道他们不会偷的 |
| 97 | but what if they, by looking at it, automatically makes their AI model understand what we do. | 但是,如果他们,通过看它, 自动让他们的AI模型理解我们做什么。 |
| 98 | Even if our company secrets are not released, our competitors will gain the advantage of our | 即使我们公司的秘密不被泄露 竞争者也会从中得利 |
| 99 | playbook. | 游戏本。 |
| 100 | That is a huge worry. | 这是一个巨大的担忧。 |
| 101 | What do you tell them? | 你告诉他们什么? |
| 102 | The answer is on-premise deployment. | 答案是实际部署。 |
| 103 | So basically, get your own server and then get your free open source model, | 所以基本上,得到自己的服务器 然后得到你的免费开源模型, |
| 104 | which will save you a bunch of money. | 这会节省你一大笔钱。 |
| 105 | The money will be a one-time expense to set up the GPUs, | 这笔钱是建立GPU的一次性开支, |
| 106 | 但此后不再向Anthropic或OpenAI支付象征性费用. | |
| 107 | That would be one way. | 这将是一种方式。 |
| 108 | The second safest way would be to deploy it through an Amazon Bedrock or Microsoft Azure, | 第二个最安全的方法就是通过亚马逊的贝德洛克或微软的阿祖尔来部署它, |
| 109 | so the hyperscalers have a history and I think has a lot more trustworthiness. | 因此,超规模的有历史 我想有 更多的可信度。 |
| 110 | If they say they won't give the data to the large language model, | 如果他们说不会把数据给大语言模型 |
| 111 | at least I personally would believe that. | 至少我个人会相信 |
| 112 | How are chief executives asking you about | 行政主管问你怎么样? |
| 113 | Chinese data models and Chinese AI? | 中国数据模型与中国AI?. |
| 114 | They're not. | 他们不是。 |
| 115 | They're already installing it. | 他们已经安装好了 |
| 116 | Because it's cheaper or because it's better availability? | 因为它更便宜 还是因为它更可用? |
| 117 | It's cheaper and good enough. | 更便宜又够好 |
| 118 | 好吧,每个人都知道 Anthropical OpenAI 也许甚至Google都比 | |
| 119 | best Chinese models, but the best Chinese models, if you're willing to use them online, | 但最优秀的中国模特 如果你愿意在网上使用 |
| 120 | are incredibly cheap. | 实在太便宜了 |
| 121 | 它们相当于顶级Anthropical OpenAI模型成本的5%. | |
| 122 | Or if you want to install it on-premise, that's also possible, but there is an | 或者如果你想把它安装在前提上, 这也是可能的,但有一个 |
| 123 | upfront one-time expense to buy the GPUs. | 购买GPU的前期费用 |
| 124 | The talk this week, of course, is what's happening | 这周的谈话,当然,是发生了什么 |
| 125 | with the Nvidia chief executive. | 与恩维迪亚首席执行官。 |
| 126 | How do you feel about Jensen Huang's relationship with China? | 你对黄詹森与中国的关系有何看法? |
| 127 | I think Jensen Huang is a phenomenal global CEO. | 我觉得黄詹森是全球首席执行官 |
| 128 | I think he treats China respectfully, | 我觉得他很尊重中国 |
| 129 | as he does all the countries that I'm aware of. | 就像他做所有我所知的国家一样 |
| 130 | Do you see this as China trying to get more access to frontier chips? | 你认为这是中国试图获得更多的边境芯片吗? |
| 131 | There's a lot of noise out there. | 声闻众生. |
| 132 | How do you see the repercussions from what we've learned? | 你如何看待我们学到的东西的影响? |
| 133 | Well, we haven't seen the outcome of the investigation, so I'd be hesitant to speculate. | 嗯,我们还没有看到 调查结果, 所以我会犹豫 猜测。 |
| 134 | When you look at Chinese AI and also some of the stocks, | 当你看中国AI和一些股票时 |
| 135 | do you worry that they're in as much of a bubble as in the US? | 你担心他们像美国一样 陷入泡沫吗? |
| 136 | I think it's a different kind of bubble. | 我觉得它是一种不同的泡沫。 |
| 137 | I think in the US, | 我想在美国, |
| 138 | in the world, first of all, AI is a phenomenal technology revolution. | 在世界上,首先,AI是一个惊人的技术革命. |
| 139 | It's absolutely not a bubble. | 这绝对不是泡泡 |
| 140 | In the US, what worries me as an investor is whether this cyclical funding—company A gets | 在美国,我作为一个投资者最担心的是 这种周期性融资——A公司是否得到 |
| 141 | to B, to C, back to A—that cyclical funding is the valuation created by that kind of an | 周期性供资是这种基金所产生的估值。 |
| 142 | unnatural and somewhat inappropriate mechanism justifiable. | 不自然和不适当的机制是有理由的。 |
| 143 | We don't know. | 我们不知道。 |
| 144 | The second question is, when there are models almost as good at 95% off, | 第二个问题是,当有一样好的模型 95%的折扣, |
| 145 | should people consider that? | 人们应该考虑一下吗? |
| 146 | And if they do, will that put a price in pressure? | 如果他们这样做,那会给代价带来压力吗? |
| 147 | That would be the sort of—the individual stock might be a little | 这样的话 个人股票可能有点 |
| 148 | bubbly, although they do have great products in the US. | 泡泡,虽然他们在美国有很棒的产品。 |
| 149 | I think the China side, the question is, the business models are not as obvious as in the US. | 我认为中国方面的问题是,商业模式不如美国那么明显。 |
| 150 | You can't just pump out tokens and make that much money. | 你不能只是抽出 令牌和赚这么多钱。 |
| 151 | You can pump out some tokens and make some money. | 你可以抽出一些代币 赚一些钱。 |
| 152 | So I think the question is, firstly, | 所以,我认为问题首先是, |
| 153 | are the Chinese stocks okay with respect to the American stock valuation? | 中国股票对美国股票估值是否满意? |
| 154 | And that depends on the American stock valuation. | 这取决于美国股票的估值。 |
| 155 | And secondly, it depends on how fast the Chinese | 其次,这要看中国人的速度 |
| 156 | companies can find the business model that might at least point to breaking even in two or three | 公司可以找到商业模式,它至少可以指向两到三个 |
| 157 | years. | 岁月 |
| 158 | Obviously not immediately. | 显然不是立即的。 |
| 159 | It's hard for all of them. | 诸众生苦. |
| 160 | You mentioned open source. | 你提到开源。 |
| 161 | You've always been an advocate of open source AI. | 你一直都是开源AI的倡导者. |
| 162 | That hasn't changed. | 没有改变。 |
| 163 | Is there something that would make you change that, or you just think it's the right model | 有什么东西会让你改变, 或者你只是认为这是正确的模型 |
| 164 | for everyone? | 给大家看? |
| 165 | Well, I think to be balanced, | 嗯,我认为平衡, |
| 166 | there are two different arguments about in face of security. | 面对安全,有两个不同的论点。 |
| 167 | Outside security, open source is | 外部安全,开源是 |
| 168 | great for everything. | 伟大的一切。 |
| 169 | Information sharing, transparency, safer, cheap, cheaper, good price | 信息共享、透明度、更安全、便宜、便宜、价格好 |
| 170 | for performance, all that. | 为了表演,这一切。 |
| 171 | But in light of safety, there are two reasonable arguments. | 但鉴于安全性,有两个合理的论点. |
| 172 | One is closed source is terrible because it's in the hands of one company and we don't know | 一个是封闭的源头是可怕的,因为它在一个公司手中 而我们不知道 |
| 173 | what they're doing, how much effort they've put in. | 他们在做什么 付出了多少努力 |
| 174 | Something breaks. | 有东西坏了。 |
| 175 | Can they fix it? | 他们能修好吗? |
| 176 | We can't bring to the wisdom of the crowd to jump in and help out. | 不能引众生智慧而为跳跃助道. |
| 177 | So open source is better. | 所以开源比较好. |
| 178 | But the other complaint is I think also valid, which is so many open source models and so many | 但另一个抱怨是,我认为也是有效的, 这是这么多的开源模型和这么多 |
| 179 | bad people, will they find one to take advantage of it because they have access? | 坏人,他们会找到一个 利用它,因为他们有权限? |
| 180 | I think both | 我觉得两者都是 |
| 181 | arguments are reasonable and I'm still on the side of open source, but I acknowledge some risk as | 我仍然站在开源的一边 但我承认有些风险 |
| 182 | well. | 不错 |
| 183 | I mean, what are Chinese LLM models doing to make sure that they're not used to actually fuel | 我的意思是,中国的LLM模型在做什么 以确保它们没有被使用 实际燃料 |
| 184 | attacks or pinpoint cyber vulnerabilities? | 攻击还是确定网络弱点? |
| 185 | Well, I think every company, American or Chinese, | 我想每个公司 无论是美国还是中国 |
| 186 | has an obligation to do as much security and guardrails as possible. | 有义务做尽可能多的保安和护卫。 |
| 187 | The Chinese companies | 中国公司 |
| 188 | that I know are putting in no less effort on that compared to the American companies. | 我知道与美国公司相比,我们在这方面付出了同样的努力。 |
| 189 | Anthropic is on its way to this trillion dollar IPO. | 安东尼正在前往这个万亿美元IPO的路上. |
| 190 | What are your thoughts on that? | 你对此有什么想法? |
| 191 | Well, my thoughts were, as I said before, they're truly a phenomenal growth machine | 嗯,我的想法,正如我以前说过的, 他们是真正的 一个惊人的成长机器 |
| 192 | and their product is great, their model is great, people are willing to pay, | 而他们的产品是伟大的, 他们的模型是伟大的, 人们愿意支付, |
| 193 | people are getting benefit. | 人们正在得到好处。 |
| 194 | So it's certainly a great company, but is it worth one to two trillion | 所以,它当然是一个伟大的公司, 但价值一到两万亿美元 |
| 195 | dollars? | 美元吗? |
| 196 | How do you justify it and how do you explain the issues about the massive investment | 如何解释大规模投资的问题? |
| 197 | that expects a huge increase next year to justify this investment on the next model? | 明年预计会大幅度增加投资 以证明在下一个模式上的投资是合理的? |
| 198 | When might that go wrong? | 什么时候会出问题? |
| 199 | Because the investment is larger, perhaps even than the company's, | 因为投资是更大的,也许甚至比公司的, |
| 200 | you know, large part of the cash reserves. | 大部分的现金储备 |
| 201 | And then the pressure from open source models, | 然后来自开源模型的压力, |
| 202 | will that cause businesses to shift along with the benefits of on-premise? | 这是否会使企业随着现成的效益而转变? |
| 203 | So I just gave the | 所以我刚刚给了 |
| 204 | pros and cons and I'm not a stock analyst. | 利弊和我不是股票分析师 |
| 205 | I don't know if I would buy or sell. | 我不知道我会买还是卖 |
| 206 | I'm certainly not | 我当然不是 |
| 207 | compelled to buy and I'm not compelled to sell. | 被迫买,而我不被迫卖。 |
| 208 | What's your, very diplomatic, as always, what's | 什么是你的,非常外交,一如既往,什么是 |
| 209 | your advice to AI developers in London and Europe? | 你对伦敦和欧洲的AI开发者的建议? |
| 210 | Embrace open source and build your European | 拥抱开源并构建您的欧洲 |
| 211 | that's on top of open source and that part will always be owned by you. | 在开源的顶端 这部分将永远为您所有。 |
| 212 | UK may be less of an | 英国可能不是 |
| 213 | issue because of the language. | 由于语言问题。 |
| 214 | In the other countries, one could build different, better | 在其他国家,人们可以建立不同的,更好的 |
| 215 | models that fits their language and it's, I mean, the top American models have decent coverage for | 符合他们语言的模型, 我的意思是,美国顶尖的模型 有体面的覆盖 |
| 216 | European languages, but you can always do better. | 欧洲语言,但你总是可以做得更好。 |
| 217 | We're working in Central Asia and we find that | 我们在中亚工作 我们发现 |
| 218 | while the best American models are still best for coding and tasks like that, but there is a need to | 而美国最好的模型 仍然是最好的编码和任务 这样,但有必要 |
| 219 | additionally train. | 额外火车。 |
| 220 | The nice thing about open source model is you can take it and train it on | 开源模型的优点是 你可以把它 并训练它 |
| 221 | local data, whether it's your language in your country or more business data, medical data, | 本地数据,无论是你在贵国的语言 还是更多的商业数据,医疗数据, |
| 222 | it gets better that way. | 这样会更好 |
| 223 | So I think that is the most economical way to expand this fear of | 所以,我认为这是最经济的方法 扩大这种恐惧 |
| 224 | influence of open source and contributing to it from each country. | 开放源码的影响和每个国家对此的贡献。 |
| 225 | So you're telling me actually | 所以你告诉我 |
| 226 | it's not too late. | 还不算太晚 |
| 227 | There is a fear amongst politicians and actually Europe is just is just | 政客们有恐惧 事实上欧洲只是正义 |
| 228 | late to the game because they don't have a lot of these big shiny LLM models. | 因为他们没有太多的LLM模型 |
| 229 | Well, I think these | 我觉得这些 |
| 230 | models aren't going to be that big shining. | 模特儿不会那么亮 |
| 231 | They're just going to solve problems for countries in | 他们只会解决各国的问题 |
| 232 | which the language is not big enough. | 语言不够大 |
| 233 | You know, like for French, German, Spanish, Italian, | 就像法语,德语,西班牙语,意大利语 |
| 234 | English, UK English, there is enough data. | 英语,英国英语,有足够的数据. |
| 235 | That's not a compelling thing to do. | 这不是一个令人信服的事情做。 |
| 236 | What is compelling | 何谓有说服力 |
| 237 | then for all of Europe is to build applications. | 那么对于整个欧洲来说 就是建立应用程序。 |
| 238 | You look at a couple of the UK unicorns, they're | 你看一对英国独角兽,他们是 |
| 239 | largely application companies. | 主要是应用公司。 |
| 240 | Take 11 Labs, great company, but their technologies are not | 11个实验室,伟大的公司, 但他们的技术不是 |
| 241 | necessarily better than every other company, but they build a good brand and platform and show | 肯定比其他公司都好 但他们造了一个好品牌和平台 |
| 242 | value to customers. | 对客户的价值。 |
| 243 | So I think there can be a lot more unicorns as well as big companies adopting AI | 所以,我认为有很多独角兽 和大公司 采用AI |
| 244 | in Europe. | 在欧洲。 |
| 245 | If you look at the pyramid of any technology stack, Semiconductor is always the | 如果你看看任何技术堆积的金字塔 半导体总是 |
| 246 | to get invested in, you know, Intel as an example. | 来投资英特尔作为例子 |
| 247 | And also the second is the platform layer, | 第二层是平台层 |
| 248 | 微软Windows,iOS,Android,以及现在的LLMs. | |
| 249 | But really the biggest moneymaker is going to be | 但真正的最大钱商是 |
| 250 | in the application space. | 在应用空间。 |
| 251 | If you look at the mobile area, think about evaluation of ByteDance | 如果你看看移动区域, 想想对字节的评价 |
| 252 | and Uber and Pinduoduo and Meituan and you name it, pretty much all the aggregate of all the | 尤伯、平杜杜、美藤 你都说得出来 几乎是全部 |
| 253 | applications is much larger than the platform or the Semiconductor. | 应用比平台或半导体大得多。 |
| 254 | So rather than Envy, | 所以,而不是爱薇, |
| 255 | Anthropix trillion dollar valuation or NVIDIA's multi-trillion dollar valuation in aggregate | Anthropix 万亿美元估值或NVIDIA 的多千亿美元估值总额 |
| 256 | will be easily maybe tens of trillions if not more in the application space. | 如果在应用空间中不增加的话,可能很容易达到数十万亿。 |
| 257 | So I think the | 所以我认为 |
| 258 | European companies should look at the application space as the biggest area to focus on. | 欧洲公司应当将应用空间视为关注的最大领域. |
| 259 | And personally, I think London has a great entrepreneurial ecosystem. | 就个人而言,我认为伦敦有一个伟大的创业生态系统。 |
| 260 | And now most of | 现在大部分 |
| 261 | these small companies are getting smaller, require less funding, and can use AI to do a lot of the | 这些小公司越来越小, 需要更少的资金, 可以使用AI做很多 |
| 262 | work. | 工作时 |
| 263 | And that's hugely in favor of cities like London, where there's good top talent, | 这非常有利于伦敦这样的城市,那里有优秀的顶尖人才, |
| 264 | but maybe not as large a talent base to build a large company. | 但也许不是建立大公司的人才基础 |
| 265 | But no longer do these startup | 但不再做这些开始 |
| 266 | one-person companies, OPCs, have to be big. | 个人公司,OPC,必须大。 |
| 267 | Also, you know, the Silicon Valley has the ability to | 而且,你知道,硅谷 有能力 |
| 268 | invest tens of billions of dollars per deal. | 每笔交易投资数百亿美元。 |
| 269 | Europe does not. | 欧洲没有。 |
| 270 | But now these smart app companies don't | 但现在这些聪明的应用公司不 |
| 271 | require that much investment. | 需要大量投资。 |
| 272 | It's so interesting. | 这很有趣。 |
| 273 | Kai Fuli, thank you so much as always for joining | 富里凯,谢谢你一如既往的加入 |
| 274 | us. | 我们 |
| 275 | Always measured and always interesting. | 总是有分寸,总是很有趣。 |
| 276 | Synnovation的主席和首席执行官 | |
| 277 | O1.ai的首席执行官和AI土著的作者Ventures. |
该视频共有字幕 277 条。解锁更多字幕为会员功能,请移动到 价格
