AI: US-China Needle Un-Threads bit more

...another political tit for tat

The Washington Post headlines the latest step to continue to curb China’s access to the bountiful exports of US technology companies, impacting the businesses of US tech companies like Nvidia, AMD, Intel, Apple and others, with  “Commerce Department moves to cut key supply lines to China’s AI industry”:

“The Commerce Department announced Tuesday a raft of new export controls aimed at slowing China’s development of advanced AI technologies, a key plank of the Biden administration’s China policy.”

“Commerce Secretary Gina Raimondo said in a late Monday briefing that the rules were aimed at hobbling advances in China’s military, even as she acknowledged they would have broader commercial impact. AI is a dual-use technology that promises to be critical for next-generation military systems — but also for key civilian sectors in a modern economy, including consumer electronics, health care and education.”

As I’ve highlighted in previous posts on US-China tussles over AI and related technology exports, it’s really important for US interests in the long run, both economic and national security, to ‘thread the needle’ with a balanced approach vis a vis China.

Both countries do realize this for their own longer term economic interests, with encouraging senior level visits between the two countries in the past few months, and the months ahead. And there is an effort even in these latest restrictions, to allow US chip vendors like Nvidia, AMD and others to sell their wares to sell in China, as Reuters reports:

“Buried deep in more than 400 pages of rules issued on Tuesday, officials at the U.S. Bureau of Industry and Security (BIS) said they are open to the semiconductor industry's input for finding ways to keep sending AI chips to China for small and medium-sized systems.”

“The rules were designed to curtail China's ability to exploit American chips to build massive supercomputers that can be used to create technologies similar to OpenAI's ChatGPT and could also be used for military purposes, officials said.”

“Thomas Krueger, a former U.S. National Security Council export control official, said "the organizing principle for all these rules is to keep them focused on those capabilities that can enable Chinese military systems. They're not interested in going after broad consumer applications. They're really trying to thread that needle."

But the imperatives of contemporary day-to-day politics continues to present headwinds. In the meantime, these moves will continue to hobble US tech companies globally, a point I’ve also discussed, and highlighted in the Post piece as follows:

“The new rules are notably expected to impact Santa Clara, Calif.-based Nvidia, the leading supplier of chips for training advanced AI models and the sixth-most-valuable company in the world. Nvidia had previously been allowed to ship slower lines of AI chips to China, but these are also expected to be blocked from sale to China without an exemption.”

This likely refers to Nvidia being restricted in exporting their modified H800 GPU chips for AI applications to China, a lesser version of their highly sought H100 GPUs and AI infrastructure. Other US chip companies will also be impacted by other Commerce Department moves:

“One of the new rules “prevents the workaround of simply purchasing a larger number of smaller datacenter AI chips which, if combined, would be equally powerful as restricted chips,” the Commerce Department’s Bureau of Industry and Security said in its update.”

And this week saw one of their leading technology companies Baidu (China’s Google), announce Ernie 4, their next generation LLM AI that goes head to head with OpenAI’s GPT 4/ChatGPT, and Google’s upcoming Gemini LLM AI. As the AP reports:

“Chinese search engine and artificial intelligence firm Baidu on Tuesday unveiled a new version of its artificial intelligence model, Ernie 4.0, claiming that it rivals models such as GPT-4 in the U.S.”

“Baidu’s CEO Robin Li demonstrated Ernie 4.0 at the company’s annual Baidu World conference in Beijing. He said the model has achieved comprehension, reasoning, memory and generation, which uses algorithms to produce and create new content.”

“Li said that Ernie 4.0 was able to understand complex questions and instructions and apply reasoning and logic to generate answers to questions.”

The system is also designed to generate images with multimodal capabilities, again matching the efforts of US companies OpenAI, Microsoft, Google, Meta, and others. They’re forging ahead.

In the meantime, as I’ve also outlined in earlier posts, China will find a way around US moves on chips, likely accelerating their efforts to build their tech industry both top down and bottom up. They’ve already taken small steps hobbling the leading US tech company in China, Apple, with restraints of iPhones used in government offices. And Chinese companies continue to invest aggressively in building their AI related technologies and services, as we saw with the latest Huawei smartphones a few weeks ago.

As I’ve said before, not threading this needle with China for the US, is collectively like smacking ourselves in the head and/or shooting ourselves in the foot. For short-term political imperatives. Alternatives like India and other places will take years if not decades to build their chip and tech supply infrastructure.

The US continues to need to thread the needle with China on tech, AI, and beyond. For both the near and long term US AI, technology and broader economic interests. It’s the only way the long-term Math works, despite the near-term political tit for tat. Stay tuned.

(NOTE: The discussions here are for information purposes only, and not meant as investment advice at any time. Thanks for joining us here)