ChinAI #304: Year 7 of ChinAI
Reflections on U.S. AI Strategy in Trump 2.0
Greetings from a world where…
ChinAI is now seven years old, which means it should be developing a solid sense of time, overcoming some fears from its younger days, and becoming more sensitive to the feelings of others (we’re still working on that last one). *A special note of thanks to the paying subscribers who have financially supported this work over the years.
…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).
Damage Control: US AI Strategy in the Trump 2.0 term
About six years ago, I testified in front of the U.S.-China Economic and Security Review Commission, a bipartisan body created by Congress, for the first time. The standard operating procedure is for experts to conclude their testimony with three bite-sized recommendations on pragmatic improvements to U.S. policy on a particular issue. At the end of my testimony, which argued that China was not poised to overtake the U.S. as an AI superpower, I went with a different approach. My conclusion:
“Given the U.S. structural advantages and current lead in AI, maintaining the status quo is a defensible policy option to enhance U.S. competitiveness in AI.”
To put it simply, don’t do stupid stuff.1 In my circles, I get the chance to work with a lot of young people who are dedicated to doing the most good they can do. It’s inspiring, but at times, I wonder if we overlook the importance of doing less harm. In Matthew chapter 19, a young man asks Jesus what one must do to enter the Kingdom of God. Jesus replies, “Go, sell your possessions and give to the poor, and you will have treasure in heaven. Then come, follow me.” In other words, donating to charity and going to church every Sunday does not cleanse you of the ways your material comforts are sustained by the exploitation of others. It is not enough to do good. Can you limit the damage you do to the world around you?
As I’ve been making the rounds to spread the word about my book Technology and the Rise of Great Powers, the most common question I get is not about the history of why the U.S. outpaced Germany in the second industrial revolution (shocking, I know). People want to know what GPT diffusion theory implies for U.S. AI policy going forward. The book’s central finding is that leadership in past technological revolutions rested on which country’s education institutions effectively broadened the base of engineering skills linked to the pertinent general-purpose technology (GPT skill infrastructure).
It’s funny: back in December of last year, a good friend of mine was putting together a memo for David Sacks, the Trump administration’s incoming “AI czar.” My friend asked me to send over suggestions on changes the U.S. should make to be a more competitive leader in AI. My primary recommendation: invest in the institutions that widen the pool of “average AI engineers.” According to my analysis, one of the key sources of the U.S.’s advantage in AI diffusion is its university system: in terms of universities with at least one faculty member who has published one paper in a top AI conference — a rough proxy for GPT skill infrastructure for AI — the U.S. leads the world with 159 such institutions (by comparison, China has just 29). In the conclusion chapter, I also emphasized “investment in community and technical colleges as a way to unlock latent potential in the U.S. AI talent pipeline.”2
Lately, I’ve been engrossed with The Pitt, an ER-like TV series set in the emergency room of a Pittsburgh hospital, so please forgive a slight medical detour. In many of the scenes, the doctors, nurses, and students perform damage control resuscitation. The basic premise: limit blood loss in severely injured patients before seeking definitive repair of injuries. This standard of care traces back to the U.S. Navy concept of damage control, used to describe protocols needed to save a ship during a maritime emergency.
Looking back on my attempts to provide input to Sacks, as the Trump administration continues its campaign to undermine higher education, I can only laugh at my own naiveté. Sure, Trump’s science and technology advisers could call for greater support for community college programs that establish alternative pathways to train AI talent. But that seems cosmetic in a world where the ruling party is slashing the Department of Education, which will likely lead to the closure of many community colleges and make higher education less affordable and accessible.3
The patient was bleeding out, and I was focused on improving his ankle mobility.
I don’t know where we go from here. Remember, ChinAI is only seven years old, so it hasn’t learned about disciplines like relentless optimism. For me, a starting point is to recognize that we are in a time of damage control. Save the bite-sized, pragmatic recommendations for another day. One last thing about damage control procedures: if you can limit the excessive bleeding and heat loss, the patient can eventually return to the operating room. Then, there’s an opportunity for definitive repair.
ChinAI Links (Four to Forward)
My four favorite issues of ChinAI from this past year:
ChinAI #260: Why are so many young Chinese people joining the Momo army? A longform report on the “Momo” army on China’s internet and their quest for online anonymity
ChinAI #262: Expert Draft AI Law Changelog: Saad Siddiqui constructed a detailed changelog for various versions of a draft AI law published by a team of Chinese experts — a window into the policymaking process.
ChinAI #275: What does China's government procurement market tell us about large model diffusion? This issue was the first time I had come across data on China’s public procurement market for AI models.
ChinAI #281: What can China's special project on machine tools tell us about its overall science and tech policy? This is one of those issues that the hardcore readers of ChinAI will appreciate.
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.
This four-word mantra became the unofficial Obama doctrine of foreign policy. Looking back, it received way too much unfounded criticism.
Gelhaus, Diana, and Luke Koslosky. “Training Tomorrow’s AI Workforce: The Latent Potential of Community and Technical Colleges.” Center for Security and Emerging Technology, April 2022.
I should have taken the current Vice President of this country at his word: “We have to honestly and aggressively attack the universities in this country. … The professors are the enemy.”
China has started their NLP since 1991 with Corpus validating... I reckon their strategy is regardless of computation method & model framework, we would always need well-defined and clean data. Consequentially, the AI game, and later, with Quantum computation, is their game
When we scale a product of AI as Oracle, we are doing harm in the process. That's the entire point. The Incentives of American capitalism on the development of AI produces unintended consequences, it already has.