Reading :: The Party’s AI

 The Party’s AI: How China’s new AI Systems are Reshaping Human Rights

By Fergus Ryan, Bethany Allen, Shelly Shih, Stephan Robin, Nathan Attrill, Jared Alpert, Astrid Young, and Tilla Hoja

This research brief by the Australian Strategic Policy Institute (ASPI) makes for depressing reading. In it, the authors methodically examine how Chinese policies have implemented AI in a number of different ways, mainly internally, but with external implications as well.

Internally, China is regulating and applying AI in ways that enforce their official policies. For me, the most evocative example was in justice (Ch.2). Surveillance and policing are, of course, boosted with facial recognition, omnipresent cameras, phone trackers, and biometric databases; the government plans to build in capabilities to time and coordinate government responses. But AI is also used in courts and prosecution — AI is meant to play an auxiliary role in a system that is understaffed. But in practice, “AI designed to aid the prosecution might do so in ways that aren’t consistent with due process and the fair treatment of defendants” (p.38). The 206 System used in Shanghai “is able to make sentencing recommendations, review evidence and keep tabs on ‘deviations’ by prosecutors” — if the prosecutor disagrees with the system’s recommended decisions, the prosecutor must explain the deveiation and send a completed approval form to the court leader (p.40). Additionally,  “Shenzen in 2024 announced the country’s first AI-assisted trial oversight system for judges,” and this system “helps to generate judgments for confirmation by the judge” (p.40). 

Similarly, China is now using AI for surveillance targeting ethnic minorities (Ch.4), censoring politically sensitive images (Ch.1), and censoring online disocourse more generally (Ch.3). Censorship involves not just suppressing information, but also being vague — for instance, if you present an LLM with “images related to the Tiananmen Square massacre” (p.19), Chinese LLMs avoid using key terms such as “crackdown,” “reform,” or even “Beijing,” and “tended to frame the event as a necessary measure to maintain social stability”; US-based LLMs ChatGPT and Gemini were less likely to do this (p.19). The authors found similar results when asking LLMs to describe images of Falun Gong and the Dalai Lama. Additionally, results varied depending on the language the human operator used to query the LLM: English, Chinese (simplified), and Chinese (traditional) (p.27). 

Such censorship doesn’t just affect domestic Chinese audiences, it also affects others using these systems. That’s true when interacting with LLMs outside China’s borders. But the report also describes a specific case in which Chinese LLMs deeply affect those outside China’s borders: AI fishing platforms (Ch.5). “Fleets of Chinese fishing trawlers prowl the world’s oceans and costs, pulling in enormous catches at industrial scale,” the report states (p.55) — their distant-water fleet of fishing vessels account for “around 15% of global marine capture” (p.56). Sometimes these fleets poach from other countries’ waters. In this overfishing, they are aided by ‘AI-enabled fishing forecasting platforms” (p.55) that coordinate AI forecasting and satellite data to increase accuracy and fishing hauls.

In conclusion, the report argues that the Chinese government has the goal of “ensuring that global AI standards benefit Chinese companies and China’s authoritarian political system” (p.62). The authors make several policy recommendations, which seem broadly positive (ex: “Promote transparency around AI vendors”) but unlikely to be pursued by the global community.

Overall, I found the report to be both insightful and depressing. A decade ago, I would have considered the report as cataloguing China’s problems and issues, and in some ways positive for the USA, because AI censorship and conformity lead to inflexibility, caution, and conservativism — preserving Western advantages in innovation by lowering the costs of failure and rewarding flexible thinking. Or so I would have told myself. But the West is also growing more authoritarian. The US system is different, but as Elon Musk’s well-publicized tinkering around Grok suggests, it allows similarly problematic use of AI. 

Reading :: Mastering Your Entrepreneurial Journey

 Mastering Your Entrepreneurial Journey: From Vision to Venture

By Andreas Kuckertz, Thomas Leicht, Maximilian Schieu, Indra Da Silva Wagner, and Bernd Ebersberger

This open-access book was written by five members of the Institution of Marketing & Management at the University of Hohenheim in Stuttgart, Germany. It’s pitched to a general audience who is interested in entrepreneurship. In this relatively short book, they cover:

• What entrepreneurship is
• How to begin an entrepreneurial journey
• How to validate the problem
• How to conduct entrepreneurial prototyping and product development
• How to develop a business model
• How to pitch
• How to develop and leverage a network
• How to deal with entrepreneurial failure

Honestly, the things they cover are eerily similar to those I cover in my classes and workshops. They tell us that this book addresses a gap in the entrepreneurial literature: On one hand, successful entrepreneurs have written lots of books based on their personal experiences; on the other hand, academics have written well-sourced articles with evidence about what works. This book is in the middle: Based in evidence from the academic literature (much of which was written by the authors), but accessible to aspiring entrepreneurs (p.2). The book can be read sequentially or not (p.3). 

The chapters are fairly short, they offer heuristics to help readers conceptualize the advice, and all end with “Three Things to Do Right Now” — specific advice for the readers to take. I liked this structure, although the “Three Things” tended to be a bit vague — I would have preferred them to be more systematically tied to the heuristics or principles on one hand, and a specific example of a venture on the other. In some cases, the advice is vague enough that I think new entrepreneurs may find it hard to follow. For instance, the chapter on business models (the authors do not number their chapters) ends by suggesting readers identify market segments, resources, etc. Although they mention Osterwalder and Pigneur in the chapter, they don’t show the Business Model Canvas or discuss how the different segments dynamically relate, so it’s hard for readers to visualize how these different parts relate. Similarly, the chapter on pitching goes over Sequoia Capital’s outline for a pitch, but they don’t note that the pitch genre looks different for pre-seed vs. seed rounds or angel investors vs. VCs (for instance). 

And that’s okay. New entrepreneurs should be reading a variety of books. This one is a succinct, accessible introduction to the basics of startups and Lewan Startup. I would recommend it as a good 50,000-foot overview, and I could see using it in an early-stage accelerator — with more specific texts to cover the details. For that application, I highly recommend it.

Reading :: Postprocess Postmortem

Postprocess Postmortem

By Kristopher M. Lotier

This book, based on Lotier’s dissertation, really took me back. Lotier examines “postprocess,” the “movement/theory/attitude” (p.3) that took hold of composition studies in the 1990s, then seemingly disappeared. Lotier argues that it — or at least some form of it — has stuck around in contemporary composition theory, but under different names, often under the heading of new materialism.

The book is grounded in two strands of thought that have often been considered together and more or less combined. One is grounded in the pioneering works of Thomas Kent, who characterized this strand of theory as “paralogic rhetoric” or “externalism”; Lotier characterizes this strand as “postprocess” without a hyphen (p.26). Paralogic rhetoric characterizes discourse production and reception as uncodifiable dialogic activities, and thus centers on hermeneutic guessing (pp.27-28). In contrast, he uses “post-process” with a hyphen to characterize a second strand, one that relies on critiques of subjectivity and focuses on discursive conditions (p.22); this strand is based in Trimbur and others.

I mentioned that this review takes me back, and that’s because Kent was at Iowa State University, and he and others on the faculty discussed paralogic rhetoric quite a bit in my PhD classes in 1994-1999. So Lotier’s discussion of those years at Iowa State (in Ch.5), which covered contributions by Kent, Blyler, Ewald, and my dissertation director David R. Russell, brought me back to those days. As Lotier points out, ISU’s program in professional communication adopted postprocess ideas early and brought them into PC scholarship. In contrast, post-process (with the hyphen) was, he says, grounded in Saint Thomas University in New Brunswick (Chapter 4) in the works of Anthony Pare and colleagues Hunt, Reither, and Vipond. (I certainly read a lot of Pare in grad school!) 

So what happened to postprocess/post-process? Lotier argues that many of its key tenets have been taken up by new materialist scholarship. Here, he demonstrates that many postprocess ideas have become mainstream: writing is grounded in specific moments, no generalized theory can completely capture what happens in these moments, readers and writers coconstruct meaning, materal conditions affect what is written (p.191). 

Overall, I thought this was an interesting and enjoyable read — although I suspect a lot of the enjoyment comes from rethinking the articles that I saw my professors writing in the 1990s from a new perspective. I really appreciated the methodical recontextualization that Lotier offers here, as well as his exploration of how these ideas have been taken up in different strands of thought. If you’re interested in comp theory, or if you remember comp in the 1990s, or if you want to see how historiography can explore the evolving thought in a specific field, definitely pick it up.