In this week’s episode of Power and Consequences, we (Gary Gensler and Simon Johnson) discuss the current global landscape of the artificial intelligence business, with an eye to who may win this high-stakes race. This is first of a two part series - next week we’ll discuss the global economic consequences of such competition.
Currently, it’s advantage USA for large language models, though China also is doing very well. The US leads in computational power dedicated to AI due in large part to its advances in chip design, export controls, financial markets, and significant investment in data centers. The US model companies generally emphasize so-called ‘close weight’ models keeping proprietary the parameters to their algorithms. Chinese companies have led on ‘open-weight’ models which may advantage adoption diffusion.
China has demonstrated significant capabilities in AI computer science and has shown impressive advances in closing the AI performance gap with the US. With its size, political system, sensor-based economy, and social credit system, China also has greater access to personal data than any other country.
There are also many other important components of the AI technology stack and global competition beyond the LLM models themselves. Other niches — very profitable niches — are occupied by others — for instance for high bandwidth memory, lithography, and semiconductor fabrication labs.
Surely each country will wish to establish some scope of AI sovereignty, as well as adopting AI to fit their strategic needs, exerting control over their own economy and mitigating trade-offs between efficiency and AI supply chain resiliency. There also will be desires to develop applications that fit local conditions, particularly in countries very different (in income, culture, other features) from the US or China.
For further information listen to Gary’s fireside chat ‘AI & The Future of Sovereignty’ with former US National Security Advisor Jake Sullivan or read ‘AI Sovereignty Is About Options, Not Ownership’ by Ren Ito or ‘Is AI sovereignty possible?’ by a Brookings team earlier this year.
We also recommend the recently published Priority Technologies, edited by Elisabeth Reynolds.
