5月31号,在上海举办的摩根大通第20届全球中国峰会上,阿里巴巴集团主席蔡崇信与摩根大通北亚区主席兼大中华区投资银行副董事长康胜德(Kam Shing Kwang)进行了深入交流,探讨了阿里巴巴的变革、人工智能等议题。
在对话中,蔡崇信详细讨论了多个主题,包括分析纽约自由人队赛季伊始的亮眼表现,与之对照的是布鲁克林篮网在上一赛季不佳表现后面临的挑战。他主张长期的成功比短期的成就更为重要,倡导在组织中培养一种灵活和胜利的企业文化。在提到领导力和公司战略时,他描述了公司为加快决策过程和赋予业务单元CEO更多权力而进行的重组计划,特别强调了像吴泳铭(Eddie Wu)这样年轻有才华的领导人在电子商务和云计算等先进领域带来的积极影响。然后,讨论转向了阿里巴巴战略性地利用人工智能(AI),通过结合专有技术和开放源代码技术来提升用户体验和提高其平台的运营效率。面对包括监管挑战、市场竞争和地缘政治紧张等外部压力,他对公司的创新能力和未来发展保持信心,认为阿里巴巴的实力源自于其不懈地专注于核心技能。
通过此次交流,蔡崇信表示:我们看到了阿里巴巴在面对当今变化多端的商业环境和外部挑战时,如何依靠领导力变革、战略调整和技术创新来保持其行业领先地位的决心。这种针对长期发展和内部文化建设的专注,结合对新兴技术领域如人工智能的深入投资,展现了一家全球技术巨头面对未来挑战和机遇的布局策略。而年轻领导力的提升和权力的下放,更是为阿里巴巴注入了新的活力和创新能力,有望在未来的发展道路上继续引领潮流。
以下是讨论交流的翻译内容及原文第二部分(所有内容由通义听悟及通义千问进行转译):
主持人:
真的吗?很棒。在未来几天的峰会上,人工智能无疑将成为一个热门话题。作为生成式人工智能领域的积极投资者,阿里巴巴在这方面有着深刻的见解。你能和我们分享,你认为未来几年AI将在全球扮演怎样的角色吗?更具体地说,阿里巴巴会如何利用AI技术?
蔡崇信:
首先,我来这里谈论人工智能就像外行人在专家们面前卖弄技艺。毕竟我们的听众中有许多真正在从事日常人工智能项目的专业人士。而对于我们这些外行来说,理解今天的人工智能——特别是那些侧重于大型语言模型的技术——就好比你在培养一个大脑,训练它逐渐达到可以与人类智能接近的机器智能。这个过程就像教育一个孩子,你看着他们成长,从小学到中学,再到大学,最终他们可能会取得博士学位,甚至多个博士学位。
蔡崇信:
这正是大型语言模型领域的竞争所在。当人们相互比较各自的语言模型,竞相宣称自己的模型比对方更优时,其实就像在说自己养育了一个拥有三个博士学位的孩子。这些“孩子”不仅精通生物学、数学、心理学等各个学科,简直就是全才。如果你把人工智能和机器智能的训练理解为对孩子的教育过程,那么你会发现,将一个孩子送入大学并完成学业通常需要22年时间,他们之后可能还会继续攻读研究生学位。而我们只用了三到四年时间,就让我们的大型语言模型在知识掌握和数学计算等方面达到了与人类相仿的水平,他们的聪明程度堪比博士生,这实在是令人震惊。
蔡崇信:
阿里巴巴从三个方面参与人工智能的工作。
首先,作为一家技术领先的公司,我们坚信机器智能会不断提升,机器将变得更加智能。很多人讨论着通用人工智能(AGI)的理想,我确信未来某个时刻,我们会开发出在一定层面上能够体现AGI特征的机器,但这取决于你如何定义AGI。就想象一个小孩,他们可能在物理学上极其天才,这就是AI,然而如果你告诉他们去一个天内交10个朋友,他们可能无法做到。因此,在某些领域里,机器完全有可能变得比人类更加聪明。因此,我们坚信通用人工智能(AGI)这一理念,以及AI的持续发展。在当代,我们认识到了所谓“规模法则”,即你投入的资源越多——无论是数据量还是计算能力——大型语言模型的性能提升并不会见顶,反而会继续增长,并且这种增长往往是以超线性的方式进行的。这其实有些吓人,因为这意味着只要持续供给机器以数据,机器便能变得越来越智能。在教育儿童时,数据既是他们的粮食,也是他们在图书馆中翻阅的各式书籍。这正是我们所追求的第一个目标。阿里巴巴追逐着这一理想,我们已经开发出了自家的大型语言模型,“通义千问”,它已成为中国领先的模型之一。在一些领域里,我们甚至在全球范围内也展现出了竞争力。这就是我们的首要目标。
第二点,我们经营着云计算业务。在我看来,阿里巴巴实际上可能是为数不多的几家既拥有自研人工智能技术能力,又运营着云计算业务的公司之一。
考虑到微软和OpenAI是两个分离的实体,它们目前维持着良好的合作关系,但未来可能会选择分道扬镳。微软并没有自主研发AI的实力,而是选择将这一任务外包给了OpenAI。亚马逊虽然在云计算领域活跃,但并未自己开发专属的大型语言模型。而Facebook虽然开发了他们自己的大型语言模型,Lama,并且选择开源,但它们并不涉足云计算业务。
在美国,谷歌是唯一一家同时具备内部技术研发能力的公司。尽管谷歌在云计算市场排名第三,但在人工智能领域,有观点认为它未能超越OpenAI。然而,转向中国,阿里巴巴则独树一帜,我们不仅在云计算市场占据领先地位,而且在人工智能领域也展现出了强劲的竞争力。
AI与云计算的融合至关重要。为什么这么说?因为任何使用我们AI服务的客户,无论是选择我们的开源版本还是通过API访问的专有版本,都将不可避免地依赖于云计算的强大计算力。
我们还打造了最大的开源AI社区ModelScope,该平台集聚了众多其他开源AI资源。社区成员使用这些开源AI工具时,自然也就需要计算资源。正因如此,我们的云计算业务收入在上一季度实现了显著增长,尤其是云业务中的AI收入实现了三位数的增长,这给我们带来了极大的激情。正是AI和云计算的这种联合,形成了一个强大的组合。因此,这是我们参与AI的第二个方面。
第三点,同样重要的方面是AI在多个垂直领域的应用。以电子商务为例,AI在推荐系统中发挥着重要作用,无论是为朋友的生日挑选合适的礼物,还是在虚拟试衣间中尝试不同的服装以查看它们在你身上的合身度。个人助理和客户服务领域需求庞大,这些领域的服务质量可以通过人工智能得到极大提升。在审视我们在电商领域的应用案例时,AI的可用性和潜力让人充满期待。正是看到了AI在如此多的方面的应用前景,我们才会如此积极地投入到AI的研发和应用中去。
系列其他文章:
主持人:
right? Great, so artificial intelligence is going to be discussed, I think, throughout the summit this couple of days. And Alibaba is a very active investor into generative AI, and so could you share with us what you think the role AI will play in the world in the coming years? And maybe more specifically, how the AI adoption will happen in Alibaba?
蔡崇信:
Yeah. First, Well, for me to sit here and talk about AI is like ban men Fu, you know, like you, we have a lot of people in the audience who, are in actually working on AI projects on a day in and day out basis. So the way as a layperson, if you will, to understand AI is with today's AI, which is very focused on large language models, you're basically trying to train and a brain to achieve machine intelligence that could approximate human intelligence and through that training process that's like educating a child your growth, you know, let's say you have children and you are, you send them to school, you send them to middle school, high school, college, and then eventually they get PhDs, or eventually they get multiple Phd degrees.
蔡崇信:
This is the race, that's what's going on in the large language model race. So when people try to compare large language models and mine is better than yours, they're really saying, well, I have a child that has like 3 Phd degrees. They are well versed in biology, math, and psychology, whatever it is, all subject matters. The thing is, if you can understand AI and the training of machine intelligence in that context, in the context of educating kids, so what's happened is it takes about 22 years to send a kid to college and graduate from college, and maybe they'll pursue postgraduate degrees. It took us, you know, maybe three or four years to get to their large language model to be as smart as human beings in terms of knowledge and certain sort of mathematical computations and things like that to a point where they're just as smart as Phd students and that's pretty scary That's, that's incredible. All right, so that's the context.
蔡崇信:
Alibaba is involved in AI in three different ways. Number one, just fundamentally as a technology company, as a pioneering technology, we believe in continuous advancement of machine intelligence that machines will get smarter and smarter, a lot of people talk about artificial general intelligence, this ideal of reaching AGI, Yeah I'm sure at some point we will have machines that could have some aspect of AGI based on however you define AGI, just like think about a child, they could be super, super smart in physics and you know that could be that's AI, but if you ask them to go and make friends, make 10 friends in a day, they may not be able to do that, right? So in certain respects, machines can become even smarter than human beings, right? So we believe in this ideal of AGI and continuous development and today with the idea of scaling law which means that the more resources whether it's data, data resource is a computing power that you put into it the marginal return in terms of performance of these large language models does not diminish. It continues. And in fact, it actually grows in a superlinear way. That's pretty scary because it just means that the machine can get, you know, smarter and smarter as long as you feed the machine with data and data is the food and also the books in the library when you're educating when you're educating a child, right?
蔡崇信:
You use that analogy. So that's number one. We just as a pure ideal, we're we're going for it and Alibaba has developed our proprietary large language model called Tong Yi Qianwen, which is actually one of the leading models in China. And in certain respects, we are competitive globally as well. That's number one.
蔡崇信:
Number 2 is we have a cloud computing business. I think Alibaba is actually, you know, probably one of the very few companies that both have proprietary in house AI capability and also a cloud business.
蔡崇信 :
If you think about it, Microsoft and OpenAI are two separate companies. They have a very nice partnership right now, but maybe in the future they will, you know, maybe go there separate ways. So Microsoft actually does not have their proprietary development of AI, they basically outsourced it to OpenAI. Amazon is in the cloud business, but they don't have proprietary AI that they develop themselves in terms of a large language model. Facebook has their large language model, which they open source Lama, but they don't have a cloud business.
蔡崇信:
The only American company that has sort of both internal in house or is Google. But Google is number 3 in Cloud. And AI is, you know, some say they're not as good as OpenAI, right? But so you come to China and you look at Alibaba, we're the only company that both run a leading cloud business, and also we're competitive in AI.
蔡崇信:
That combination between AI and cloud is important, why? Because anybody that uses your AI. So we have both open source versions of our AI and also our more proprietary version where people can access through Apis. They are, anybody that uses our AI will need to use cloud computing power, they need compute.
蔡崇信:
Alright, we also developed the largest open source, AI community called ModelScope, which has a lot of open source, other people's open source Ai's in that marketplace. So when they use the open source AI within our community, they need compute. So they need computing power. That's how we can grow our cloud computing revenue in the last quarter, our AI revenues, in the cloud business have grown triple digits. So that's very exciting to us, right? So the combination of AI and cloud is a terrific combination.
蔡崇信:
So that's the second way, the third way we're, we're involved in AI and it's important to us is that we can apply AI in a numerous vertical applications and if you think about e-commerce, the scenarios in e-commerce in terms of recommending what to buy, you know, if you want to buy somewhat something for your friends, you know, recommendations for their birthday, let's say, and you know, you want to walk into a virtual fitting room to see how some clothing fits on you. You need personal assistance, you need customer service, a lot of those things can be drastically enhanced by AI technology. And when we look at our use cases in e-commerce, it's incredible. We just see so much upside that AI could be applied. And so through these ways, we're involved in AI and that's why we're all in.
我们关注国内外最热的创新创业动态,提供一站式的资讯服务,实时传递行业热点新闻、深度评测以及前瞻观点,帮助各位创业者掌握新兴技术趋势及行业变革,洞察未来科技走向。