Popular science writer dispels rumors about China breaking through the lithography technology chokehold
Greetings from a world where…
for the rest of the college football season, this status update will be devoted to tracking the Iowa Hawkeye offense’s march to mediocrity
…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: Can China independently develop advanced lithography machines?
Context: This past week, the tech-obsessed slice of China’s internet was abuzz with news that Tsinghua University had developed a new approach to producing extreme ultraviolet (EUV) light sources. The ability to generate wavelengths of EUV light is important for advanced lithography machines, key equipment used to produce the chips that train fancy AI models and power our mobile phones. Chinese netizens got very excited about this breakthrough in a “chokehold technology” [卡脖子技术], even speculating that a lithography factory was already under construction in Xiongan. Since 2019, export controls have limited the Dutch company ASML from exporting EUV lithography machines to China.
In this week’s feature translation, Jie Wang, an influential science writer, pours some cold water on these rumors. The host of the influential “Science has a Story” [科学有故事] podcast, Wang’s work aims to popularize scientific knowledge. This piece (link to original Chinese) was published in The Intellectual [知识分子], a platform that covers the state of science in China, founded by Chinese and Chinese-American scientists. The Intellectual often expresses more views that are more “technoglobalist,” as opposed to technonationalist.
Key Takeaways: Interestingly enough, the Tsinghua University paper that started all the rumors about China breaking through the lithography chokehold was published in early 2021.
Wang hasn’t figured out why it has just recently surfaced and hyped up — to the extent that you have Big V Weibo influencers (> 500k followers) posting images of a lithography factory in Xiongan that doesn’t exist (see screenshot below).
This 2021 Nature paper, on the steady-state microbunching (SSMB) mechanism for generating EUV light, was co-authored by researchers from Tsinghua University and Helmholtz-Zentrum Berlin (HZB). Wang states that the paper’s principles are still in the verification stage and are 15-20 years from being truly practical.
Why is it so hard to build advanced lithography machines? These systems are made up of three key parts, and “the technical challenges of each part are comparable to landing on the moon.”
In terms of the light source, which the Tsinghua paper and buzz speaks to, only one German company (TRUMPF) can make the laser that continuously produces EUV light. This laser has more than 45,700 parts, and even TRUMPF relies on a Lithuanian company to produce key equipment.
Second, let’s talk optical systems. Only one German company (ZEISS) can develop the optical systems for EUV light. To illustrate how smooth and how flat these lenses need to be, Wang states that if a virus were dropped on the lens, it would be like a 100-meter high hill.
Third, the high-precision console that is used to carve out transistors. It’s composed of 55,000 components, dependent on patented technologies provided by the following countries (at least): Japan, South Korea, Taiwan, the United States, Germany, and the Netherlands.
Wang’s powerful conclusion: “If China wants to break through the technological blockade and independently produce lithography machines, it needs to achieve complete independent innovation in all three key parts. All we can say now is that in the first light source section, we see a little bit of hope.”
Wang’s take on the future of EUV lithography machines:
Tsinghua Professor Chuanixang Tang, a corresponding author on the aforementioned Nature article, applied to list the SSMB experimental device as a major scientific research project in the 14th Five-Year Plan. However, Wang couldn’t find any news about this project. “Note that this kind of civilian scientific research project is not a military project and does not need to be kept confidential. All project approvals need to be publicized. In other words, at least so far, this project has not even been approved,” he writes.
Based on this, he sketches out a timeline. Let’s say, optimistically, Professor Tang’s project gets launched next year. Building the required scientific research facility might take another five years; add on another three years for testing; then, five more years for building a commercial laser. So, that’s 13 years for making a breakthrough in one of the three key parts of a lithography machine. “But can the other two key parts of the lithography machine be completed in these 13 years? There is not even a shadow (of evidence) of this yet,” Wang states.
Wang closes out the article with this: “Within 20 years, it is impossible for any country in the world to be able to completely independently build a lithography machine that represents the most advanced level in the world, and the United States is no exception…Being practical and realistic is the right way to develop science and technology. For ultra-precise and complex machines like lithography machines, seeking the greatest range of international cooperation is the best solution.”
ChinAI Links (Four to Forward)
CSET’s Emerging Technology Observatory “compiles, tags, and summarizes news and commentary from selected Chinese sources, helping English-speaking users keep up to date, skim the latest news, and discover new perspectives.”
Cheers to Zachary Arnold, analytic lead for ETO, and other team members on putting this together. This brings ChinAI one step closer to retirement!
Two Links from my testimony at the Senate Intelligence Committee
1) Full Hearing Video. It was an honor to testify before the Senate Intelligence Committee last week, on the subject of AI and national security. A few highlights:
41:50: my opening statement.
1:02:30 to 103:30, followed by 106:00 (discussion about safety and security issues in military AI systems): I think Senator Rubio asked a great question about whether authoritarian countries can put in place the necessary guardrails for military AI systems. This is a research direction that I’m currently working, and I tried to share some initial insights.
2:12:00: Senator Warner, chair of the Senate Intelligence Committee, explains faulty reward functions using the OpenAI CoastRunner game example. It’s pretty impressive to have a sitting member of Congress who has this depth of knowledge about AI. And he’s a GW alum!
This is the article that undergirded my testimony and discussion with the Senators. Thanks to everyone for sharing it, as it’s now the most-read article in the Review of International Political Economy journal published within the last 12 months.
For Associated Press, Matt O’Brien and Hannah Fingerhut report on the advanced AI super computing data center in Iowa, used exclusively to enable OpenAI to train its GPT-4 model. They write, “For much of the year, Iowa's weather is cool enough for Microsoft to use outside air to keep the supercomputer running properly and vent heat out of the building.” Call it the “Coastal Data, Midwestern Computing” project! (If you got that last reference, we can be friends)
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).
Any suggestions or feedback? Let me know at email@example.com or on Twitter at @jjding99