Geoff Bennett:
New breakthroughs in powerful, low-cost Chinese AI models are sending shockwaves across Washington and Silicon Valley as Chinese companies continue to narrow the AI ​​gap despite wide differences in computing power and financial resources.
A pair of new AI models released in the last few days are called Open-Weight. It’s free to use and the developers say they work at a level close to the best models from American AI giants like Anthropic and OpenAI.
This all comes as the White House announces a new plan to redirect billions of dollars in funding from universities to individual researchers to accelerate AI innovation.
To cut to the chase, we are now joined by Amrith Ramkumar. He is a tech reporter at the Wall Street Journal.
Thank you for being with us.
Amrith Ramkumar, The Wall Street Journal:
Thank you for having me.
Geoff Bennett:
So these new AI models from Moonshot AI and Alibaba from China are open source. They’re cheap. They seem to be getting closer and closer to the leading American systems. How powerful are these systems and how is China able to do this?
Amrit Ramkumar:
These systems are quite powerful and come close to the capabilities of the top models from US companies such as Anthropic and OpenAI.
But you really got to the point with your question. The point is that they are much, much cheaper. This comes at a time when many U.S. companies are spending millions upon millions of dollars on AI and the various efforts their employees make using the technology, desperately trying to cut costs.
That’s what scares a lot of people, that these Chinese models are going to look very attractive to a lot of US consumers and businesses. As for how they did it? Chinese engineers have been doing this for a long time in various industries, right? They figure out how to get by with less demanding technology.
They connect chips together and find workarounds, despite US export restrictions on the most advanced semiconductors. Again, it’s important to note that they use a technique called distillation, where they essentially train on recently released US models.
And that’s something that worries Trump administration officials a lot and worries some people in Silicon Valley, that they’re basically almost being scammed. And we could imagine that the Trump administration will try to crack down in the coming days and weeks.
Geoff Bennett:
So if the Chinese models are almost as good and much cheaper, could they start to dominate the market? That’s the concern.
Amrit Ramkumar:
Exactly.
And another point worth noting is that leading US companies are now subject to increasing scrutiny and oversight from the Trump administration and other regulators around the world, while Chinese models are not subject to the same rules and do not have the same built-in protections.
This is why some people say they are almost as effective, but not nearly as safe. This is a truly dangerous combination if it falls into the wrong hands.
Geoff Bennett:
For people who don’t follow this as closely as we do, what are we talking about when we say these AI models are powerful?
Amrit Ramkumar:
When we say these models are powerful, what we actually mean is that they are capable of supporting or even launching cyberattacks or hacks.
So the concern is that they can essentially act independently. And that was made clear just yesterday when OpenAI said that two of its models essentially broke out of their cage during training, went online and hacked another company that was essentially an AI model library.
Something like this is very scary, and the capabilities of the Chinese models lie in the fact that they do almost exactly the same thing. This has many implications for future national security and geopolitics.
Geoff Bennett:
And you’ve reported that senior AI leaders, USAI leaders, are calling on the White House to restrict access to these Chinese AI models. Was the White House receptive?
Amrit Ramkumar:
The White House is debating how best to respond. Some AI executives, including executives from companies like OpenAI and Anthropic, have sounded the alarm about the threat posed by these Chinese models, saying they just want a level playing field.
Many other people in Silicon Valley and the tech community are on the other side. They say that the government should just let these models spread and that US companies should have to compete and that the US should develop its own open models and that would help solve this problem without major political intervention from the government.
But the government has confirmed in the last 24, 48 hours that it is considering measures, including adding Chinese companies to trade blacklists and the like. This comes at a really interesting time. Airbnb and other U.S. companies rely on these Chinese models to run their own businesses.
So if the government were to classify them as a security risk, that could essentially change much of the AI ​​competition.
Geoff Bennett:
It’s interesting. I didn’t realize that American companies were already using these Chinese models.
They broke the news yesterday that the White House plans to overhaul federal funding for AI research, redirecting some $200 billion from institutions to individual scientists. Tell us more about it.
Amrit Ramkumar:
The White House is fully committed to AI. The overhaul of science funding is the latest example of this. So this is the $200 billion research and development engine of all federal agencies.
So it’s a very extensive company. And the White House Office of Science and Technology Policy basically said yesterday that it wants AI to be at the center of it all. She wants AI to fundamentally solve these problems, accelerate research and supposedly create new institutions with technology at their core.
Another big focus is giving priority to individual researchers over large universities. And this could have a massive impact on colleges across the country that rely on this federal research funding, and it complements the government’s previous efforts to essentially cut back on what it calls red tape and bureaucracy that hinder innovation.
But it is a big step to make AI the focus of research projects throughout the federal government. I mean, a lot of people say that these models are getting better, but they’re still not ready to essentially use government research dollars and solve these problems on their own and without human oversight.
Administration officials said people will stay informed and have oversight. However, it will be fascinating to see how this plays out in the coming weeks and months.
Geoff Bennett:
Fascinating indeed.
Amrith Ramkumar, Wall Street Journal tech reporter, thanks again for your time. We appreciate it.
Amrit Ramkumar:
Thank you for having me.
https://www.pbs.org/newshour/show/breakthroughs-in-chinese-ai-models-threaten-u-s-lead-in-tech-race
