Hello Interconnected Readers:
Last week -- American Thanksgiving plus my birthday week -- I re-read a few books. One of them was “Against the Gods: The Remarkable Story of Risk.” One small part of the book talks about Kenneth Arrow, a Nobel Laureate in Economics, and his research on risk management and insurance. According to the book:
Arrow won his Nobel Prize in part as a result of his speculations about an imaginary insurance company or other risk-sharing institution that would insure against any loss of any kind and of any magnitude, in what he describes as a “complete market."
[If you’d like to skip ahead to the six news stories and commentaries – three from English language sources, three from Chinese language sources -- here are this issue’s items:
- “Facebook Struggles to Balance Civility and Growth” (English Source: New York Times)
- “Tech CEOs Are Saying Goodbye to San Francisco” (English Source: The Information)
- “How to Think for Yourself” (English Source: Paul Graham’s Blog)
- “Ren Zhengfei’s speech at Honor’s farewell event” (Chinese Source: Huawei)
- “Alibaba and Baidu Temporarily Suspend Talks to Acquire iQiyi, ByteDance Jumping in with Interest” (Chinese Source: 36Kr)
- “Alibaba’s Cloud-Native Real-Time Data Warehouse Core Technology Revealed” (Chinese Source: CSDN)]
Arrow believes that if this imaginary insurance existed, it’d be good for the world, because more people would be more willing to engage in risk-taking, which progress depends on.
This observation got me thinking. Can such an insurance product exist?
As much as startup and entrepreneurial acts are becoming more common, there still isn’t enough risk-taking to increase the probability of something that’s “good for the world” coming to fruition, or P(Good for the World), the mathematical notation for expressing any probability. Later in this issue, you will see a story about Facebook, where the probability of Facebook posts that are “bad for the world” -- P(Bad for the World) -- was used by engineers and data scientists as an experimental metric, leading up to the 2020 presidential election, to train a machine learning system that detects and demotes posts deemed “bad for the world.” The experiment mostly worked, but the tradeoff was that less people logged into Facebook, which would directly impact its ad business. Thus, only a watered-down version of the experiment was implemented.
Despite more money being used to do early-stage venture capital and angel investing than ever before, which theoretically should support more risk-taking, most of the money is poured into ex-engineers or product managers from Facebook and other big tech companies -- the least risky kind of entrepreneurs to bet on. A parallel exists in China’s tech ecosystem too if you sub-out Facebook for Tencent or ByteDance. Why would this be considered “least risky”? Because that big tech pedigree alone may help recruit a decent team, the team may already have a working product that was an internal project before they “took the plunge”, and this kind of “less-risky risks” will more likely receive future funding that marks up valuation of the investment -- what almost every early-stage investors crave to help raise their own next fund. For these entrepreneurs, the proposition isn’t that risky either. They will likely be able to write their own salary similar to or a bit less than their previous big tech jobs with significantly more equity ownership. They will become their own bosses, while keeping their big tech stocks and stand to benefit as those stock prices rise. I’m not saying this to criticize these entrepreneurs; it’s a great deal and they’d be silly not to take it. The point here is just that there’s not as much risk-taking than meets the eye.
Meanwhile, any social media idea aimed to achieve a higher P(Good for the World) than Facebook rarely gets funded these days. Facebook is the 800-pound gorilla in the industry. If a new idea gets traction, Facebook will either copy it or buy it or, in the case of TikTok, lobby the government to kill it before copying it.
Arrow’s vision of a “complete market” remains imaginary, and the world is worse off because of it. The cost of risk-taking is still too high. The cost of working on something that increases P(Good for the World) is still too high.
P.S. I want to take a moment to celebrate the life of Tony Hsieh, who passed away last week. He is perhaps one of the best models of risk-taking that increased P(Good for the World). His book, “Delivering Happiness”, is arguably one of the best books about business, entrepreneurship, and the search for purpose. I will definitely be re-reading it. RIP Tony.
The news summary below covers the period between November 23 - 29, 2020 of six news stories – three from English language sources, three from Chinese language sources. Disclaimer: all translated article titles are done by me, not official translations from the media outlets.
Before you go on, please check out one of my most viewed posts: "Why Is Facebook Not in the Cloud Business?"
“Facebook Struggles to Balance Civility and Growth” (English Source: New York Times)
My Thoughts: As much as we, as a society, talk about how Facebook has made us addicted to its products, Facebook itself is addicted to the ad-driven business model. That’s Facebook’s “struggle”. Here’s the relevant passage from this Times article that describes this struggle and the path that Facebook’s leadership chose to take:
So the team trained a machine-learning algorithm to predict posts that users would consider “bad for the world” and demote them in news feeds. In early tests, the new algorithm successfully reduced the visibility of objectionable content. But it also lowered the number of times users opened Facebook, an internal metric known as “sessions” that executives monitor closely.
The team then ran a second experiment, tweaking the algorithm so that a larger set of “bad for the world” content would be demoted less strongly. While that left more objectionable posts in users’ feeds, it did not reduce their sessions or time spent.
That change was ultimately approved. But other features employees developed before the election never were.
Civility may be nice, but it simply isn’t as profitable in an ad-driven, attention-driven business model -- an addiction that Facebook, in my opinion, has a low chance of reducing, just like most addictions.
“Tech CEOs Are Saying Goodbye to San Francisco” (English Source: The Information)
My Thoughts: There are some interesting second-order effects worth observing in the next 2-3 years as the first-order effects of tech companies leaving San Francisco (lower real estate cost, less income inequality, smaller tax base) take shape. One second-order effect is: in an increasingly remote-friendly work environment, how much does “proximity to power” matter in realizing one’s professional ambition. It wasn’t that long ago that face time with the CEO is a factor in a person’s professional advancement. Now that the CEOs, in the case of Dropbox, Splunk and Brex as profiled in this article, are moving to a different city, mostly for quality of life and tax-saving reasons it seems, do you move there too in order to advance? Another second-order effect that’s especially interesting to me, as someone who looks at all startups on a global scale, is where tech startups based outside of the US will first enter the US market. It used to be a no-brainer. Nine out of 10 times, the first US hire is defaulted to be the SF/Silicon Valley area. How long will this default last?
“How to Think for Yourself” (English Source: Paul Graham’s Blog)
My Thoughts: I find most of Paul Graham’s essays immensely thought-provoking. This one particularly stood out, because he articulated a clear framework of independent-mindedness that few people can. This framework has three components: fastidiousness about truth, resistance to being told what to think, and curiosity. The degree in which a person either possesses these three components preternaturally or grows them as she matures can determine how independent-minded that person really can be. Also, the footnotes in this essay is worth reading too; don’t ignore them.
“Ren Zhengfei’s speech at Honor’s farewell event” (Chinese Source: Huawei)
My thoughts: The full text of Ren’s speech was released in the form of an all-staff email. The link here is a picture of that email. If you are interested in a English translation of it, Jordan Schneider did one in the ChinaTalk newsletter. Two things from this speech stood out to me. One is Ren’s rationale for selling Honor to a consortium of partners and distributors as a way to help them remain in business. A less-known fact about Huawei is the vast ecosystem of partners and distributors its smartphones and other consumer products have fueled. By selling Honor -- the cheaper smartphone line that requires less high-end semiconductor chips and thus should be less affected by US sanctions -- the “guilty by association” problem is removed. The other point is Ren’s use of the “divorce” analogy, characterizing the sale as a clean break, thus putting an end to speculations that Huawei may still retain a stake in Honor and buy it back in the future. That being said, there’s reporting that Huawei plans to send 6,000 engineers to Honor, so the break may not be as clean as Ren would’ve liked. One more thread worth paying attention to: while Huawei parts with Honor and licks its wound, Xiaomi, the smartphone brand that most directly competes with Honor, has been surging. Its stock price more than doubled since the start of 2020.
“Alibaba and Baidu Temporarily Suspend Talks to Acquire iQiyi, ByteDance Jumping in with Interest” (Chinese Source: 36Kr)
My thoughts: The future of iQiyi, a big longform video platform in China, could be a prelude of where China’s big tech companies plan to position themselves strategically in video streaming. Right now, the industry is neatly partitioned along BAT boundaries: iQiyi is owned by Baidu, while the other two competitors, Youku and Tencent Video are owned by Alibaba and...well Tencent, respectively. This article revealed that Tencent wants to be a majority owner of iQiyi one day, but the price it’s offering is too low for Baidu’s taste. That low price, whatever it is, may very well be justified, because iQiyi is accused of fraud by Wolfpack and Muddy Water, who successfully exposed all the shenanigans at Luckin Coffee. Alibaba has also shown interest, though the negotiation is on hold. Baidu is, in my opinion, the weakest of the BAT. The “B” will soon be replaced by ByteDance. How soon may very well depend on if and how much ByteDance can get in iQiyi.
“Alibaba’s Cloud-Native Real-Time Data Warehouse Core Technology Revealed” (Chinese Source: CSDN)
My thoughts: I track the database industry pretty closely. It has a special place in my heart, because of my previous work at PingCAP, which commercializes the open source distributed HTAP database TiDB. HTAP is a trendy database industry category that stands for: Hybrid Transactional/Analytical Processing. This article introduces both a new technology out of Alibaba Cloud, Hologres, and a new category that I’m sure Alibaba would like to make trendy: HSAP or Hybrid Serving/Analytical Processing. Hologres is apparently one of the core technologies that helped Alibaba Cloud withstand the massive peak traffic of 583,000 transactions/second generated on Singles Day, a topic we covered on last week’s Interconnected Weekly. I won’t go into too much detail on the technical difference between a “transaction” and “serving”, since not everyone is a database nerd like me. However, hybrid databases is definitely an industry trend that will take shape in the coming years. (And if you do happen to be a database nerd, I’d recommend reading the academic paper on Hologres published at the international database conference, VLDB, written by Alibaba engineers.)
If you like what you've read, please SUBSCRIBE to the Interconnected email list. To read all previous posts, please check out the Archive section. New content will be delivered to your inbox (twice per week). Follow and interact with me on: Twitter, LinkedIn.
上周，也就是美国感恩节加我的生日周，我重读了几本书。其中一本是《Against the Gods: The Remarkable Story of Risk》。书中有一小部分谈到了诺贝尔经济学奖得主Kenneth Arrow，以及他对风险管理和保险的研究。据书中介绍：
- “Facebook在掌握文明社交与业务增长之间的平衡的挣扎”（英文来源: 纽约时报）
- “科技公司CEO们在向旧金山说再见”（英文来源: The Information）
- “怎样为自己思考”（英文来源: Paul Graham的个人博客)
- “任总在荣耀送别会上的讲话” (中文来源: 华为）
- “传阿里巴巴与百度暂缓收购爱奇艺谈判，字节跳动有意入股” (中文来源: 36氪）
- “首次公开！阿里巴巴云原生实时数仓核心技术揭秘” (中文来源: CSDN）】
尽管创业越来越普遍，但仍然没有足够的风险承担来增加 "对世界有益 "的东西实现概率，即P(Good for the World)，表达任何概率的数学写法。在本期后面，您会看到一条关于Facebook的新闻，在2020年总统大选之前，一批工程师和数据科学家创造了一个Facebook上"对世界有害 "的帖子的概率 -- P(Bad for the World) -- 作为实验指标，用来训练一个机器学习系统，检测并降低被认为是"对世界有害"的帖子。实验大部分是成功的，但代价是登录Facebook的人少了，直接影响到广告业务。因此，最终Facebook高层只实施了这个实验的淡化版。
尽管用于做早期风险投资和天使投资的钱比以往任何时候都要多，理论上应该支持更多人去冒风险，但大部分资金都投给了来自Facebook和其他科技巨头的前工程师或产品经理 -- 这其实是很低风险的一赌。如果您把Facebook换成腾讯或字节跳动，中国的科技行业里也存在类似的情况。为什么说是低风险呢？因为光是打“前大公司”的招牌就有助于招到一个像样的团队，团队在创业之前可能已经有了一个可用的小产品，很有可能就是以前的一个内部项目。投这种 "风险较小的风险” 也更有可能以后融到更多的资金，有更高的估值。这几乎是每一个早期投资人所渴望的，以帮助自己募集下一支基金。对于这些创业者来说，内在风险也不是很大。他们在新创的公司里有更多股权，给自己开出的工资可能和以前一样或者也就低一点。自己当老板，同时还保留以前在大厂的股票，也期望股票大涨，从中获益。我这么说并不是批评这些创业者，这是个好deal，他们不接受就太傻了。重点是想说，其中承担的风险并没有想象的那么多。
同时，任何想实现比Facebook更高的P(Good for the World)的社交媒体创意现在都很少能获得投资。Facebook是行业中的磅重巨头。如果一个新的社交媒体产品冒出来的话，Facebook要么抄，要么买，要么像对TikTok那样先游说政府把它消灭掉，然后再抄。
Arrow教授对 "完整市场" 的设想仍然只是个设想，世界也因此变得更糟。承担风险的成本还是太高，从事能增加 P(Good for the World) 的事情的成本也还是太高了。
P.S. 我想借此机会庆祝上周去世的Tony Hsieh的一生。他也许是为增加 P(Good for the World)而冒风险的企业家里的最佳典范之一。他的书《Delivering Happiness》，是关于商业、创业和寻找个人目标的最佳书籍之一，我一定会重读它。RIP Tony.
我的想法: 当我们整个社会都在谈论Facebook如何让我们对它的产品上瘾的时候，Facebook本身也对广告驱动的商业模式上瘾了。这就是Facebook的 "挣扎"。以下是《纽约时报》这篇文章中最相关的段落，描述了这种挣扎和Facebook领导层最终的选择：
于是，团队训练了一个机器学习算法，来预测用户会认为 "对世界有害" 的帖子，并在news feed中对其降级。在早期测试中，新算法成功降低了令人反感的内容的可见度，但它也降低了用户打开Facebook的次数，这是一个高管们很在乎的被称为 "sessions"的内部指标。
该团队随后进行了第二次实验，对算法进行了调整，把对更多 "对世界有害 "的内容被降级的力度减小了。虽然这让用户的feed中出现了更多的不良帖子，但并没有减少他们的“sessions”或花在Facebook上的时间。
“科技公司CEO们在向旧金山说再见”（英文来源: The Information）
我的想法: 未来2-3年，随着科技公司离开旧金山的一级效应（房产价格降低、收入不平等减少、纳税基缩小）逐渐变成现实，有些很有意思的二级效应值得关注。其中一个二级效应是：在一个远程办公越来越普遍的环境中，"接近权力" 对实现个人的职业野心到底还有多重要。在不久前，在CEO面前刷脸还是对个人职业发展的重要因素。现在，以此文章所介绍的Dropbox、Splunk和Brex为例，这些公司的CEO们都搬到了不同的城市，似乎大多是出于提高生活质量和少纳税的缘故。那作为一名高官或员工也会为了晋升而搬到那里去吗？作为一个对创业“全球化”极有兴趣的人，另一个我很感兴趣的二级效应就是，发家在美国境外的科技创业公司会在哪里首先进入美国市场。这曾经不是个问题，一般默认在美国的第一名员工肯定要在SF/硅谷地区。这种默认会持续多久呢？
“怎样为自己思考”（英文来源: Paul Graham的个人博客)
“任总在荣耀送别会上的讲话” (中文来源: 华为）
我的想法: 任正非的讲话全文以发给全员工邮件的形式发布，这里的链接是那封邮件的图片。如果您对英文翻译版感兴趣，司马乔丹（Jordan Schneider）在ChinaTalk博客中做了一版翻译。这番演讲中有两点让我印象深刻。一是任正非将Honor卖给合作伙伴和分销商组成的财团的理由是为了帮助它们继续营业。华为的一个鲜为人知的事情是，其智能手机和其他大众产品助推了一个庞大的合作伙伴和分销商生态。通过卖Honor -- 其更廉价的手机产品系列，它所需的高端半导体芯片较少，因此受美国制裁的影响也应该较小 -- 会消除 “被华为牵连” 的问题。另一点是任正非用了 "离婚" 的比喻，将这次出售定为净身出户，从而终结了华为可能仍保留Honor股权并以后再买回来的种种猜测。但又有报道称，华为计划分给Honor 6000名工程师，所以这次“离婚”并不一定像任正非想的那么干净。还有一点值得关注：在华为与Honor分道扬镳、舔舐伤口的同时，Honor最直接的竞争对手，小米，一直在增长。自2020年开始，其股价已经翻了一倍多。
“传阿里巴巴与百度暂缓收购爱奇艺谈判，字节跳动有意入股” (中文来源: 36氪）
我的想法: 爱奇艺的未来，可能是中国科技巨头未来计划在长视频媒体领域战略定位的前奏。目前这个行业正沿着BAT的分界线划分着：爱奇艺是百度的，优酷是阿里的，腾讯视频就不用说了。据这篇文章透露，腾讯希望有朝一日能成为爱奇艺的大股东，但它开出的价格太低，百度不接受。这个价，不管到底是多低，可能都是合理的，因为爱奇艺正被Wolfpack和Muddy Water 指控欺诈，他们曾经成功地揭露了瑞辛咖啡的各种猫腻。阿里对爱奇艺也有兴趣，虽然谈判暂缓。在我看来，百度是BAT中最弱的一个。"B" 很快就会被字节跳动取代。多快很可能取决于字节是否能入股爱奇艺，以及能入股多少。
“首次公开！阿里巴巴云原生实时数仓核心技术揭秘” (中文来源: CSDN）
我的想法: 我对数据库行业跟得很紧，主要是因为我之前在PingCAP的工作，把开源分布式HTAP数据库TiDB做商业化。HTAP是在数据库行业里一个时髦的类别，是Hybrid Transactional/Analytical Processing的缩写。本文既介绍了阿里云出的一项新技术Hologres，同时也介绍了一个新的，我相信阿里很想变时髦的类别：HSAP或Hybrid Serving/Analytical Processing。Hologres是帮助阿里云撑住双十一时每秒58.3万笔交易的巨大流量洪峰的核心技术之一，这个话题我们在上周的《互联周刊》中也有聊过。关于 "交易 "和 "服务 "在技术上的区别，我就不在这里多说了，因为不是每个读者都像我对数据库那么感兴趣。不过，混合数据库（Hybrid）绝对是未来几年会形成的一个行业趋势。(如果您碰巧也对数据库感兴趣，我建议读阿里工程师在国际数据库学术大会 VLDB 上发表的关于Hologres的论文)。