In addition, we're gaining user mind share in live events. For example, over 190 million viewers watched the 2023 CCTV Spring Festival Gala, via live streaming on Video Accounts.
──用户心智这么强，万万没想到To conclude the strategy section, I would like to share with you our perspective on artificial intelligence, specifically the implications of foundation models for Tencent. The most important takeaway is that we expect AI to be a growth multiplier for us going forward. We have a long-standing experience in developing and adopting AI technologies, which has already benefited many of our businesses such as advertising games, short-form videos and cloud computing.
──AI作为收入倍增器，这一点上的思路，倒是和微软一致的We experienced substantial demand for Video Accounts ads promising us to release more inventory and Video Accounts in-feed ads exceeded RMB1 billion in quarterly revenue.
──视频号信息流广告收入-单季度10亿Within Business Services, we're actively helping automobile manufacturers such as NIO, BMW and GAC Motor, enhance their IT infrastructure and product offerings in areas such as smart cockpit solutions, digital maps and data management.
──然后我们会看到23年Q2财报里，汽车行业广告是增长最慢的类目（指：没有达到2位数增长）So in terms of the overall outlook, we believe the advertisers generally cautiously optimistic on a China consumption recovery this year. In more detail, companies that sell low ticket price items seeing a broad-based recovery already.
──谨慎复苏，低价商品优先复苏We see there's great demand to advertise in the short-form video format to consumers who previously weren't reached but in short form video format. And from that perspective, our Video Accounts are highly appealing because many of the viewers video accounts are not viewers [indiscernible] other existing short-form video services.
──CSO是想表达视频号用户的重合度与抖快相比非常低，也符合自己的直观认知，可能是一些爱看资讯的中老年人？消费水平还较高We see a great desire among certain categories of advertisers, for example, luxury products to advertise in format where they actually own the relationship with the consumer as opposed to being intermediated. And so for those advertisers, the mini program inventory is very attractive. And then also as we see the broader economy turning around, there are certain advertiser categories who are extremely price sensitive.
──奢侈品的情感联系，不过财报、财报电话会内容，挺让人涨知识的，能更加了解中国的商业形态And the reason is because our business is actually primarily in the social and communication and gaming business, which means that it's primarily user-to-user and it's involved very high-quality content. And as a result, these are businesses which can be supplemented by these generative AI technologies and the foundation model technologies, but they are not necessarily that easily disrupted by such technology, not like the user to machine experience like search.
──人与人，人与机器，后者更容易被AI颠覆式创新In terms of the infrastructure, we do have enough and required amount of chips to actually create the model. That's not a problem. And more importantly, we felt we actually have a very strong cloud business and the relevant technology is to really arrange and make use of the chips in a scalable and high-density way so that we can create very large clusters of chips that can really deliver the performance that's needed to train these very large modules, especially when the models get more and more sophisticated over time after a few iterations.
──看起来是类似GPU云资源出租的商业模式In terms of the commercial prospect, I think as we have alluded in our prepared remarks, it's a growth multiple. So that means it will improve our existing businesses along the line of improved monetization. Imagine advertising can be actually improved with the generative content, which is highly targeted to the users that can be very, very effective in conversion. It can actually increase efficiency, imagine a lot of our content platforms and content business can actually make use of these tools to generate content for both the creators as well as for ourselves on a much more cost-efficient basis and the user experience can be much more engaging.
──举例主要是AI生成广告素材、内容平台、创作者（后两者我猜是公众号、视频号的图片生成、资源生成）Now in terms of the generative AI, it's definitely natural, right, for us to incorporate some of these technologies into our flagship products like Weixin and QQ and we can actually improve the efficiency of the user experience. For example, we can actually allow mini program developers to develop mini programs at a much faster rate with these generative AIs.
──这个举例倒是非常适当，让用户可以更简单地通过AI来生成小程序，Make SenseIn terms of live streaming e-commerce, we do believe this is a good opportunity for Video Accounts to develop over time. But I actually emphasize the word over time. So we -- our plan is actually to grow it on a consistent measured and high-quality basis. So it is true that in e-commerce business, you actually need to have heavier life operations. But -- and we do intend to put in a heavier life operation to support this.
──愿意为了视频号的电商交易业务，承受更重的运营投入But at the same time, we emphasize one that it is actually profitable, right? So even if you put in the live operations, the cost is going to be small compared to the potential benefits. But more importantly, the way that we're doing it will be, I think, much more thoughtful, much more efficient in the sense that we will try to focus on the high-quality merchants and high-quality product categories, which means that the lineups associated with these would be actually less because a lot of times, the live ops is actually focused on solving problems. And when we actually focus on high quality, the problems will be fewer.
──可以看到一个基本思路是：仍然强调自动化？更专注于高价值类目？而不是全品类的电商On the AIGC, then, as you're probably aware, it is not the provision of AIGC itself that is costly. If you look at a service like Discord that actually is already profitable from distributing AIGC service mid journey for creating graphics. Rather it's the decision to build the large language model that powers the AIGC that is capital intensive. And the capital is not so Microsoft has stated that their large language model several 100 million jobs to build out. However, the nature of the cost is very different from the nature of costs that China Internet was facing.
One to three years ago, meaning that one to three years ago, most of the cost heavy projects that we and our peers in China Internet were engaged in are actually projects where the costs scale with the number of users we have. So if you decided to get into community group buying, then the more users you had the greater your at cost base was. Large language models are different as a fixed cost or a table stakes for being in the game, which is the CapEx that Microsoft referred to. But then that is the majority of the cost at least in the near term price generating revenue.
So we're absolutely embarked on bearing that fixed cost. And if you compare that fixed cost with our revenue, you can see it's a sizable number. But it's not a number that has very dramatic consequence of strong margins. And every other company that wants to build live-language models, whether they're larger than us or smaller than us, will have a cost that's somewhat similar in absolute fixed terms. So this kind of fixed cost is kind of cost, we think, is a more desirable cost because it is fixed as opposed to something like the subsidies where the more users you have, the cost you bought, which in naturally disadvantage to us as the Company with the most users.
──这段有极高的insight，James Mitchell以社区团购举例，社区团购的成本会跟随UV用户数的增长而扩大，此类生意对腾讯这种高DAU公司非常不利，而AIGC是固定成本投入型，无论公司大小，都承担着类似成本，天然对腾讯更加有利！Martin Lau
And also bear in mind that a lot of the cost is actually a hardware cost, which can be amortized over a few years rather than you're spending for the year.
──天然对高用户型公司、高资本型公司有利！！！And thirdly, in terms of financial products that's tied in -- that can be developed, such as wealth management, such as loans, such as insurance over the longer term, we felt that there's also opportunities if we can actually develop these businesses in a cautious way in a complied way with a very high focus on risk management and at the same time, with more focus on working with the existing licensed financial institutions. So we feel that there is a good prospect for our FinTech business going forward.
James Mitchell
And in terms of the Video Accounts monetization, Alex, so the advertising load factor, it's a very light ad load both compared to other tens of properties and also even more so compared to the big incumbent short video services. The advertiser feedback is very positive. And I say that both from a quantitative perspective, but also quantitatively, that the revenue per 1,000 impressions that we achieved on Video Accounts ads following a substantial increase in the previous months is actually superior to the revenue per 1,000 impressions achieved by Weixin Moments and substantially superior revenue per 1,000 impressions achieved by the two short-form video incumbents.
──视频号的eCPM好于朋友圈，好于抖快Now the reopening dynamic is something we've looked at a great deal because we're in the unusual position of operating very big, somewhat similar games in the Western world as well as in China and other geographies. And so, we can actually look at how reopening plays out of games in those different geographies.
And what we see is that the reason why in Western economies, there has been a roughly one-year hangover period for the game industry post COVID is because in those Western economies, the majority of the population was working for home for a period of many months. And that process of most of the population working from home for multiple months did create a high engagement base and monetization base, which the Western game industry is only now emerging from.
On the other hand, in China, there was no phenomenon of the majority of the population working for many months in their individual cities where people work from home for two weeks or in some cases, two months. But those were sporadic and scattered, and the impact has diminished because it was spread out over 3 years and concentrated over 1.5 years. So at this point in time, we don't see a reopening headwind for our game business. And I think it's possible that we won't see one given the unusual nature of the COVID outbreaks work from home behavior in China.
──游戏层面的中西方的差别Now in terms of Video Accounts, it is still true that the average time spend per user is actually much lower than the incumbents on short video. But the number actually has been growing consistently. And this is sort of part of the driver of why the total time spent on our Video Accounts has been growing quite rapidly, and now it's at 1.2x of Moments.
──22Q4视频号观看时长是朋友圈的1.2倍And in terms of the average time spent, of course, it reduces total time spent. And as a result, it reduces the total amount of advertising dollars. But I don't think the average time spent per person actually would necessarily reduce the monetization per unit time. As a matter of fact, if somebody actually spend 30 minutes on the platform versus if somebody spends 150 minutes on the platform, the first 30 minutes probably would carry higher value per unit time compared to the latter number of minutes.
──很有意思的见解，花费的总时间多，并不一定代表商业化价值高，而要关注单位时间的货币化率And then if you look at short video, I think the regulation is broadly stable, right? If there's one potential issue, it probably may be on the time spent per user is actually very high. But if you look at our platform, the time spent per person is actually way below industry standards. So, that even as an issue is probably not going to be that much of an issue for us.
──哈哈，我家视频号人均观看时长低，所以被锤也是先锤其它两家And then in terms of capital allocation, Will, so on the capital return front, we're very active. We'll be sending out the Meituan distribution in kind in the next two days. We announced today we'll be increasing our regular annual dividend by 50%. And previously, when we were an open trading window, we were buying back shares most days. And so we're very busy. I wouldn't say I'm personally busy mainly on maybe on checks. But collectively, as an organization, we're busy on the capital return front.
In terms of investing capital in other companies, then looking domestically at China, where we're actually quite optimistic given what we see on China consumption. We can't forecast how the market will value companies, but we can see what China consumers are doing, which is becoming more active. And we ourselves have therefore become more active in terms of making investments in small, early-stage privately held companies in interesting growth areas in China. Internationally, the environment is obviously extremely dynamic and we're being very selective, but also seeking to be opportunistic. So that's on turning happen and on investing capital.
──资本配置的策略，making investments in small, early-stage privately held companies in interesting growth areas in China.So, our market share of the international game market is still only in the mid-single-digit percentage. But we think that we have certain structural competitive advantages that we can bring to fair including the sort of the development capabilities from China, including the fact that if we're at the forefront of trends such as multi platformization such as shifting to game as a service rather than a premium game model and therefore, both the China game industry in general and then Tencent within that are well positioned to increase their market share of the global and the international game industry over time.
And our focus is around those areas. It's around developing great games, particularly games that have a long life and those could be more player versus player games that are similar to physical sports or they could be more content-driven player versus environment games that are more in the nature of storytelling, but, we're building up our capabilities in both kinds of games.
And then in operating those games effectively. And operating games in the games as a Service era involves much more continual hands-on support than publishing games in the premium game era. But we're reinvesting to be good at both developing and operating good games. And if we do that, we think we can continue to pre-pub our market share in this very big $150 billion plus international game month.
──key insight: 1) 国际游戏可能整体盘子增长慢，但因为腾讯差不多只占5%，所以有增长空间；2）游戏策略是开发长青的竞技类游戏、努力尝试开发内容/故事驱动的游戏、持续做好游戏运营On the geopolitical front, then everything faces some degree of geopolitical risk. I think that the risks faced by the game industry dissimilar to the risks faced by social platforms that you alluded to for a few reasons. One is that games generally you're not seeking to collect data on individual users and then optimize that users experience based on a very sort of targeted data. You're interested in collecting scale data on a large number of anonymized users, but you're not following the behaviors of individual actors.
And in addition, with media platforms, that there's concern around the content that's distributed, particularly the news content that's distributed to uses with games, you don't have new content that's distributed to users, PEOPLE who don't go into a video game if they want to watch CNN. So while I wouldn't say that games absent from geopolitical risk, I think the nature of concerns around games is different from the nature of the concerns around intermediate platforms or social platforms.
──1）游戏不会像广告一样，疯狂收集个人数据以做个人导向的优化；2）游戏不会涉及价值观、新闻等内容的分发And in terms of take rates, I don't think we want to increase the basic tax rate because we felt if there's any additional gain that we want to make, then we should be making it from value creation. So we'll be very focused on saying, oh, if we're actually serving these merchants and helping these merchants to cover the customers, how can we actually help the merchants to make their transactions easier, right?
──不轻易提高税率，而是专注于价值创造
原文 地址 江湖中历来不缺乏传奇。在建筑软件的这片江湖中，风云变幻，豪杰辈出，有两大世家始终屹立不倒——一个来自欧罗巴，名字低调：“图形软件”（Graphisoft），却继承了一身的艺术家气质，手握长剑白衣胜雪；一个来自美利坚，人称“自动桌子”（Autodesk），性格豪放不羁，七种武器样样精通。本文所说的，就是这两大世家的代表人物：ArchiCAD与Revit之间的较量。 这个论题本是老生常谈了，谈到BIM绕不过的就是Revit与ArchiCAD。两者的对比许多帖子都讨论过，但往往大而化之，原则性的东西多，细节的东西少，因此我想再作一次比较，希望能深入一点，具体一点，力争较为全面地反映两者的真实面貌。但这种对比往往两面都不讨好，你懂的，因此本文也多用戏说的语气，我姑妄说之，列位看官也就姑妄听之吧，有不当之处还请多多包涵！ 先介绍一下本人对这两个软件的熟悉程度。我用ArchiCAD有4年了，出过几套施工图，都已竣工，编过一系列向日葵图库，颇受好评，对ArchiCAD的认识偏重于施工图；用Revit一年半，出过四个工程的土建及MEP模型，也用其API编了若干插件，对Revit的认识偏重于建模（包括结构及MEP建模）。应该说对ArchiCAD与Revit的认识都算深入了。 当然两者的深度比较是一个庞大的工程，而且个人看法难免有偏见，技术上也多有误解之处，因此希望各位能指正与补充。 1 软件的思想、架构对比 从软件的历史来说，无疑是ArchiCAD悠久得多，Revit是Autodesk在2002年才收购回来的，但Revit有一个强有力的东家，马上推出“BIM”这个很炫的口号，一下把ArchiCAD沿用多年的“虚拟建筑”这个老老实实的口号给打败了，于是ArchiCAD也只好宣称自己是个BIM软件，搞得在外人看来，倒像是Revit占了先机。 从软件设计的角度来看，两者也是差别巨大的。ArchiCAD从20多年前就致力于三维建筑设计，在这方面积累了足够多的经验，多年来也是沿用其架构做一些小更新、小完善、小整合。从我接触的7.0到最新的14.0，感觉比较大的变动就是10.0版整合PlotMaker、12版支持多核计算提升速度、12版新增幕墙工具、13版团队工作大幅改进。在界面上、使用习惯上一直差别不大，这也在一个侧面反映了ArchiCAD软件设计的一个“精英思路”——我本来就...