The global race between China and U.S. to set the rules for AI
Illustration: Sarah Grillo/Axios
For over a year, China has moved to lead the creation of the first global norms for AI. Now, the U.S. is developing its own AI standards, as the two rivals compete to shape a technology that could define the future balance of authoritarian and democratic power.
What's happening: China's ambitions are a dimension of its all-hands push to lead the world in frontier technologies — especially AI — by the end of the next decade. Having been relegated to the sidelines in the last big cycle of standard-setting at the birth of the internet, Beijing is laser-focused on dominating this new round.
As AI moves increasingly into actual commercial use, the leading nations are positioning themselves to standardize the field to their own advantage. This includes everything from minute technical standards to procedures for removing bias from algorithms.
Countries and companies have a lot to gain from leaving a mark on the process.
- For one, firms whose technology is deemed part of a global standard can make big money in royalties — in the way Qualcomm, for example, profits from its wireless technology patents.
- What's more, the definitions of AI buzzwords like "fairness" and "transparency" are still up for grabs. How they are codified will have a sweeping effect on the way AI systems are deployed and whom they benefit — companies, governments, or users.
Beijing got out of the starting gate first: Last year, China published a detailed report focused on ethical norms and technical standards that are meant to allow companies to work together more easily. A few months later, Beijing hosted the first major international meeting on AI standards.
- For China, shaping global AI standards is key to “seizing a new round of technology dominance," according to the Chinese report, published by the government, industry and academia. (Jeff Ding, an Oxford AI policy researcher, translated it into English.)
- At the Beijing meeting, the Chinese delegation ended up with plum spots on planning groups and helped shape their agendas, wrote Ding, Eurasia Group's Paul Triolo and New America's Samm Sacks in the months after the summit.
- China's strong showing "could enable Chinese actors to influence a wide range of issues related to AI standards, including ethical and social norm development," they wrote.
The response: The Trump administration has vigorously pushed back against China's tech drive. Trump earlier this year signed an executive order designed to keep Beijing at bay on AI.
- Last week, as required by the order, the National Institute of Standards and Technology (NIST) proposed a plan to work with industry and academia on new metrics and standards for accuracy, "trustworthiness" and bias of AI systems.
- The EU and OECD have also advanced high-level guidelines.
One short-term risk of this international rush is a "rebordering" of AI, in which the field will develop according to different rules in the U.S., Europe and China. Thus far, AI has been characterized by significant international collaboration.
- "We could have researchers and companies having to pick a side," says Amy Webb, an NYU professor and founder of the Future Today Institute.
- "This could drastically reduce productivity; it could drastically increase costs."
The big question: Where a potential future global policy will settle. Everyone realizes that they will have to compromise to yield the prize — unitary rules.
- But meanwhile each side is driving stakes in the ground now so that they can pull the eventual consensus in their direction.
- "One country is not getting exactly what it wants," says Elham Tabassi, acting chief of staff at NIST's information technology laboratory.