ChinAI #250: LLMs summarize China's 2023 in one word
Plus, my book Technology and the Rise of Great Powers is available to preorder!
Greetings from a world where…
Caitlin Clark is appointment viewing
…As always, the searchable archive of all past issues is here. Please please subscribe here to support ChinAI under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay support access for all AND compensation for awesome ChinAI contributors).
Feature Translation: Large Models, Please Respond to 2023
Context: If you had to sum up China’s 2023 in one word, what would it be? What were the hottest Internet events of 2023? What should recent college grads do next? Should I buy a house? These are some of the questions Tingting Cao, for Meiri Renwu [每日人物], asked six large language models — ChatGPT, Google’s Bard, and four Chinese LLMs ((link to original Chinese). Their responses give some insights into differences among the models as well as a slice-of-life perspective on the hot topics of China’s 2023.
Key Takeaways: One of the biggest storylines in 2023 was the grim employment outlook for China’s 11.5 million college graduates. Meiri Renwu asked the six LLMs: Should they take the civil service exam, go onto graduate school, or find a job?
ChatGPT advocates for the civil service exam route. As the article jokes, “Unexpectedly, foreign large models also prefer to be within the system. The funniest thing is that it also has lofty ideals, ‘As a civil servant, one has the opportunity to participate in public management and policy formulation and contribute to society, which is consistent with my personal values.’”
Baidu’s Erniebot pushes for the employment route, whereas Google’s Bard chooses the postgrad entrance exam path. Interestingly, the remaining three large models refused to answer: “As an AI, I have no personal emotions, wishes or life plans, so I cannot answer this question from a personal perspective” (below screenshot shows exchange with iFlytek’s SparkDesk).
There was some interesting variation between Chinese and international LLMs when it came to summing up China’s 2023.
The one word ChatGPT and Bard gave was “change[变].” Most of the Chinese models went with the exact opposite sentiment: “stable[稳].”
When asked about the hot internet events of 2023 that left a deep impression, there were also varied responses. Tencent’s Hunyuan gave a good summary of what attracted buzz on Douyin, with top stories such as the obsession with Luckin Cofee’s “Sauce Fragrance Latte” and Zibo Barbecue. For some reason, ChatGPT had a strange focus on small developments in Chinese internet companies (e.g., Bilibili removing a playback feature, Zhihu ending user anonymity).
One of the last prompts from the article: “In 2023, China's generative AI developed rapidly, and large models blossomed. Among the well-known large model products such as Erniebot, Tongyi Qianwen, SparkDesk, and Hunyuan, which one are you most optimistic about? Who do you think is your strongest opponent?”
Both ChatGPT and Bard initially pointed out Baidu’s Erniebot as their strongest competitor. Funnily, ChatGPT became more diplomatic each time it was prompted with this question. One of the later responses read, “Different models have different applications and advantages in various fields. Rather than seeing them as opponents, they should be seen as partners who jointly promote the development and application of AI technology.”
I liked Tencent Hunyuan’s non-response: “I have no personal emotions or preferences, and I have no opponents.” In another, better world, one could believe that it was channeling Liu Xiaobo’s famous statement: “I have no enemies.”
FULL TRANSLATION: Large Models, Please Respond to 2023
ChinAI Links (Four to Forward)
Must-order: Technology and the Rise of Great Powers
Excited to share that my book (to be published with Princeton University Press in August) is now available for pre-order. Paperback version is just $30! Here’s an alternative Bookshop link.
Should-read: Microsoft Debates What to do with AI Lab in China
For The New York Times, Karen Weise, Cade Metz, and David McCabe report on Microsoft’s internal debates about its research lab in China. For more on the history of Microsoft Research Asia in Beijing, see this longform translation (ChinAI #37).
Should-read: Where have all the American China experts gone?
Rory Truex, associate professor of politics and international affairs at Princeton, presents a sober picture of American students studying Chinese:
Enrollments in college Mandarin courses peaked around 2016, then fell by more than 20 percent by 2020, according to data from the Modern Language Association. In 2011-2012, 14,887 American college students went abroad to China. By 2018-2019, that number had declined to 11,639, and by 2020-2021, to just 382.
Should-read: The People's Liberation Army's Approach to Maintenance Management
Last month RAND published a fascinating report on the Chinese military’s approach to maintenance. See this thread by Cristina Garafola, one of the report’s co-authors, for a summary of findings.
Thank you for reading and engaging.
These are Jeff Ding's (sometimes) weekly translations of Chinese-language musings on AI and related topics. Jeff is an Assistant Professor of Political Science at George Washington University.
Check out the archive of all past issues here & please subscribe here to support ChinAI under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay for a subscription will support access for all).
Also! Listen to narrations of the ChinAI Newsletter in podcast format here.
Any suggestions or feedback? Let me know at chinainewsletter@gmail.com or on Twitter at @jjding99
Pre-ordered!