ChinAI #59: Is Winter Coming for Hikvision?
Plus, what are the risks in the rise of newsletters, Substack's big announcement, and the need for editors?
Welcome to the ChinAI Newsletter!
1) The goal: every ChinAI issue is a joint effort where my collaborator helps with translation and/or commentary and (most importantly) plays a quasi-editorial role (*more on why I think newsletters may need editors later on). Toward that end, part of the subscription fees will go toward compensating contributors for their work ($100-200 each issue depending on the length and quality of the translation + analysis + editing). If you’ve got a translation/topic to pitch, or you’re interested in learning more about how to contribute, just reply to this email!
2) I want the main incentive for readers to subscribe to be to support access to good content for all (in the mold of The Guardian or Wikipedia), but folks have also pointed out that it would be nice for subscribers to have certain perks (some have suggested making the archive only available to subscribers). A middle ground I’m exploring that still guarantees access to each issue for everyone but gives a little extra to subscribers is to put together a master library of all the Google doc translations, curated by category and shared with subscribers.
The good news is we’re at 40 subscribers & 5400+ free email list readers. I rely completely on word-of-Tweet/email for promotion, so thanks for spreading the word. I’m playing around with options (right now it’s $9 for a monthly subscription and $50 for a yearly subscription) and would appreciate feedback.
These are Jeff Ding's (sometimes) weekly translations of Chinese-language musings on AI and related topics. Jeff is a Rhodes Scholar at Oxford, PhD candidate in International Relations, Researcher at GovAI/Future of Humanity Institute, and Research Fellow at the Center for Security and Emerging Technology.
Feature Translation: is Winter Coming for Hikvision?
Is winter coming for Hikvision, the world’s largest manufacture of video surveillance products and solutions? In this week’s feature translation, one Chinese investor (he holds shares in Hikvision and goes by the handle Blitzbear) argues the Night King is already inside Hikvision’s gates.
I’m very grateful to Charles Rollet, a freelance journalist who also works as a researcher with IPVM (the leading publication covering the video surveillance industry) for adding his comments throughout the Google doc full translation and letting me include some of his analysis in the key takeaways below. Check out his fantastic portfolio — which includes many in-depth stories on Hikvision.
If you’re like me, then Hikvision is probably one of those Chinese companies you hear mentioned in the news — for their role in the repression of Uyghurs in Xinjiang or for being one of two Chinese companies (Dahua is the other) targeted by a Congressional ban on U.S. federal agencies purchasing new equipment — but you don’t really know much about beyond that.
Enter our (relatively knowledgeable but probably biased) guide —Blitzbear, a Chinese investor who used to hold a decent chunk of shares in Hikvision before selling most of them and holding on to a small part to “use as an observation deck.” He spent a weekend reading Hikvision’s half-year report and came away very disappointed with Hikvision’s performance. Key takeaways:
— Blitzbear highlights two main issues: 1. Can Hikvision find commercialization pathways for intelligent security in video surveillance? 2. What’s the deal with Hikvision’s cash flow and all the loans it’s taking on?
— On the first question, Blitzbear argues that Hikvision is not making the transition from selling products and “eyes” (cameras were 47.7% of total revenue this period) to selling cloud intelligence services and “brains” that can connect cameras and hardware to the cloud computing models via 5G networks — especially in projects that can be replicated in different data application scenarios.
— A key structural trend that supports the first point: high-end cameras that support facial recognition and other AI services are becoming cheaper, so Hikvision can’t just rely on these cameras anymore and they need to get on the AI track. As Charles notes, “At the U.S.’ largest security camera trade show, I saw $59 facial recognition cameras for sale with an $8 per month app subscription.”
— On the second question, Blitzbear highlights Hikvision’s massive increase in short-term debt (an increase of 3.4 billion RMB throughout 2018 and an additional 1.1. billion in 2019 so far). Interestingly, some of its long-term loans came through a public-private partnership (PPP) or Design-Build-Finance-Operate-Transfer (DBFOT) scheme. In laypersons terms, Hikvision goes to the Chinese (subnational) government entity who is looking to do a surveillance project and says, “Since you will be paying me later, you might as well first lend me the money, and I will take it to make and build the surveillance infrastructure, and then you can pay me less later.”
— Crucially, in this period, Hikvision signed six projects according to this scheme (around 1 billion RMB in total pledged loans) and all were surveillance projects in Xinjiang. See Ipvm’s excellent coverage of Hikvision’s projects in Xinjiang, which include cameras for re-education camps and hundreds of mosques. Blitzbear highlighted one such project signed this year with Luopu, a sparsely populated rural county of about 280,000 (almost entirely Uyghur); I dug into Hikvision’s interim report to find the other five, which are listed in the full translation.
— Charles also added some critiques of this piece. This was one of the benefits of getting his “editing” because I often get too attached to each week’s piece (from spending time translating and getting into the head of the author). Charles writes, “A huge omission here is that Hikvision faces much more direct threats, in terms of possible US sanctions over its Xinjiang activities. This could put them in a Huawei-like situation.” Another flaw in the piece, according to Charles: “When you hear talk of Hikvision's financial issues, always keep in mind that Hikvision's controlling shareholder is the Chinese government, so it's unlikely it would be allowed to flounder like a normal private firm,” and “Chinese government video surveillance spending is still huge; whether that gravy train ever stops or slows down is an issue unaddressed by this article.”
FULL TRANSLATION: Has Hikvision's Cold Winter Come, I found the Answer in its Semi-Annual Report
This week’s proposition: every newsletter that gets a substantial degree of traction should consider getting an “editor” or explore substitutes for “editing” functions.
I operate ChinAI through Substack, a newsletter publishing platform which recently raised $15.3 million in a round led by Andreessen Horowitz. This awesome news (at least for people who use Substack) got me thinking about the bigger picture of how newsletters fit into the media landscape. It’s obvious that more and more people are writing and reading newsletters (one recent survey found that nearly 60 percent of American adults subscribe to an email newsletter of some sort). As Substack frequently points out, newsletters allow writers to build a direct relationship with an audience that trusts them more than social media and traditional media channels.
But how do we ensure that this trust is deserved or can last? At their core, many of the most influential newsletter are essentially personal blogs delivered to your inbox. While I’ve laid out some of the benefits above, there are also costs to concentrating all the functions of a media source/platform into one individual’s hands. Let’s take a look at two “editing” functions that are often loss when you go from a traditional media platform to a newsletter. I divide these into editing before publishing (EBP) and editing after publishing (EAP).
Let’s consider a case where a newsletter is taken to be the go-to source for all things on a subject (e.g. Bill Bishop’s Sinocism which I’ve criticized in the past and which, fittingly enough, was the first newsletter launched on Substack. All the valuable editing that happens before any article publishes on, say the Washington Post or other traditional media outlets, is lost in newsletter content. I know firsthand that this type of editing would be really helpful for ChinAI, as evidenced by: i) a lot of past typos, ii) a lot of past ass-holeish comments that I wish I could take back, iii) all the dogs that didn’t bark (e.g. personal biases that aren’t questioned & mistakes that haven’t been found yet). Crucially, newsletters also get much less editing after publishing, especially ones that are under a paywall (which is Substack’s existing model). Many traditional media sources have public editors that can critique their posts and, of course, most of their posts are open for everyone to see, comment on, and critique. This is not the case for personal, paid newsletters.
My model for ChinAI is to be very open and upfront about my own fallibility and personal biases. That’s why we make every ChinAI post open for all so that people can go back and call me out for everything I’m writing. That’s why we’re going to incorporate more collaborators to play an “editing” role in the future. Again, I’ve said I plan to redefine what it means to be a gatekeeper in the “China-watching” space. One addendum: let’s also redefine the model of making newsletters along the way.
ChinAI Links “Four to Forward”
Great piece by Olivia Carville of Bloomberg on the immense implementation challenges for federal agencies to remove surveillance cameras made by Hikvision and Dahua. At least 1,700 of these cameras are still operating in places where they’ve been banned. This is a conservative estimate because a) only a small percentage of offices actually know what cameras they’re operating and b) two cameras running identical Hikvision firmware could carry completely different labels as both companies have U.S.-based warehouses that repackage cameras. This is why when you do industrial policy you should probably HAVE A PLAN in place to implement it properly, especially given the complexity of tech supply chains. A recent Reuters article details the widespread misunderstandings and confusions with the NDAA as well as other new regulations that have restricted the role of Chinese companies in the U.S. marketplace.
This week’s must-read is another War on the Rocks article by Peter Mattis and Matt Schrader. It’s a balanced thoughtful take that recognizes the U.S. needs to address China’s systematic theft of IP and exploitation of the openness of Western institutions but needs to do so without alienating the Chinese diaspora many of whom say they have already experienced discrimination amidst the growing U.S.-Sino tensions. Two aspects that I really liked in this article: 1) a reference to America’s problematic history with enforcement actions directed at Asian-Americans and 2) a call for raising the “China literacy” of U.S. officials and American society. Hopefully ChinAI can contribute to #2.
This week’s sobering read was a NYT piece by Ana Swanson. “The Committee on the Present Danger, a long-defunct group that campaigned against the dangers of the Soviet Union in the 1970s and 1980s, has recently been revived with the help of Stephen K. Bannon, the president’s former chief strategist, to warn against the dangers of China. Once dismissed as xenophobes and fringe elements, the group’s members are finding their views increasingly embraced in President Trump’s Washington, where skepticism and mistrust of China have taken hold…An increasing number of people in Washington now view the decoupling of the two economies as inevitable — including many of the members of the Committee on the Present Danger (Bannon, Senator Ted Cruz, and Newt Gingrich were at an inaugural meeting in April).” Again this is why in issues #53, 54, and 55 I came down so hard against these New Tech Cold War Warriors, because they are either intentionally or subconsciously promoting the memes that pave the way for us to repeat the worse McCarthyist excesses of the Cold War.
Whoo okay that was a lot, let’s take a deep Exhalation and read Ted Chiang’s wonderful collection of short stories. One chapter in Exhalation on “digients” is especially meditative for those interested in ethical implications of transformative AI.
Thank you for reading and engaging.
Shout out to everyone who is commenting on the translations - idea is to build up a community of people interested in this stuff. You can contact me at email@example.com or on Twitter at @jjding99