More Signal.
Less Noise.

‘This is a game changer’: Janice Gross Stein on DeepSeek and China’s growing influence on AI

Video

A visitor at the 2024 World Artificial Intelligence Conference in Shanghai China, July 4, 2024. Andy Wong/AP Photo.

Janice Stein, Belzberg Professor of Conflict Management and the founding director of the Munk School of Global Affairs and Public Policy at the University of Toronto, joins Hub publisher Rudyard Griffiths for a breakdown of DeepSeek, the new Chinese machine learning model everyone in technology is buzzing about. They also discuss China’s growing influence in AI and the technological race that is now underway between China and the U.S.

You can read Rudyard Griffiths’ breakdown of DeepSeek here.

The following is an automated transcript. If you are quoting from or referencing this episode, please refer to the audio to verify.

RUDYARD GRIFFITHS: Rudyard Griffiths here, the publisher of The Hub. Welcome to this our regular up to speed short video series where we talk with experts writing and commenting in the hub about big stories and ideas breaking in the news. Extremely fortunate to have on the program today. Janice Gross Stein, the founding director of the Munk School of Global Affairs, to talk with us about some coverage and commentary that we have in the hub today on the big story that shook the tech world and is shaking stock markets this morning, deep seek the new Chinese AI Janice, thanks for coming on the program.

JANICE GROSS STEIN: It’s a pleasure to be with you, Rudyard, and what a big story that is.

RUDYARD GRIFFITHS: Yeah. Well, tell us right off the bat, what do you think is, is the innovation that deep sea has clearly pioneered here reflected in the shock in Silicon Valley in reaction to the arrival of this new Chinese AI model, and the stock market reaction today, big sell offs and chip makers in large American tech companies.

JANICE GROSS STEIN: Rudyard, the innovation is not in the Quality of the AI. It is in two big areas, one, cost, cost, these new AI models are so expensive, even the price that these hyperscalers are now charging businesses to use them is growing. The Chinese model does it for 90% less in cost. That’s why you’re seeing the market sell off. The second story, which I don’t think has yet been fully baked in, is energy use, right, right? And we haven’t seen a lot of commentary about that. Why does this matter? The models were forecasting energy consumption that, frankly, no country was able to meet. We had to grow the grid. We had to build data centers. You know the announcement that Donald Trump made over the weekend. We are talking about billions of dollars that are going in to power the compute required. Well, what did deep seek Do you did it for one six. In other words, this is about the chips that they use to build their model, and the numbers are, frankly, absolutely mind blowing. 16,000 for to the best of our knowledge, for open AI? 2000 at the maximum for deep seek, that is an enormous difference in future power consumption.

RUDYARD GRIFFITHS: Yeah. So you know it matters when you slash costs by 10x that is the Silicon Valley definition of a disruptive, transformational change. Janice, the way I’ve thought about it in a piece that I’ve written today in the hub is, you know, we were assuming the last two years that AI would be a Manhattan style project, that you would need to assemble massive amounts of capital, material, server farms, nuclear reactors, to eke out at the end of it, this synthesized, synthetic machine intelligence. And I think the big disruption here, the big win, in a way, is for open source that now this technology can be delivered much more cheaply per unit of machine intelligence. It can be distributed widely through through the through the internet, through people simply downloading the source code of deep seek, the other models open. Ai, which powers the entire Microsoft universe, is proprietary. It’s a black box we we Sam Altman has not let us peek inside it. I think that’s a big change, too. Janice, this is, this is the genie out of the bottle, and the whole Manhattan style, top down, big tech, big government. Assumption that we had that that was the only way that AI would out seems to really have been rocked the last seven days.

JANICE GROSS STEIN: People are completely right about that, Rudyard, but let’s talk now about you’ve talked all about the upsides of open source. Let’s talk now about the dark side. Deep seek used open source to build it is now itself going to be open source. Think about the security concerns, right? You can download, download this one iPhone or a laptop, a powerful version of this. You can run denial of service attacks. You can you can scale up hacking right off, you know, your desktop, using state of the art. Ai, that’s right, and this is a game changer for national security. The United States is now having last generation discussions about Tiktok. That will be nothing in comparison to what people will be able to do downloading it, as you said, on their phones, you will have Chinese generated source code on your phone. Rudyard that even if you’re not using it from a level of purposes, becomes much easier to hack and to control. I can tell you that as much as the markets are royal today. There isn’t a National Security Agency insight that isn’t saying, Oh my God, what do we do about this?

RUDYARD GRIFFITHS: It’s the number one app right now, downloaded on Apple’s all present, all powerful app store, Janice. Just to wrap up this conversation, this seems like a big win for China, it seems like China has been able to innovate. You know, under pressure, you get diamonds. And boy, did they get diamonds with deep seek in terms of a model that allows them to to fuse, or certainly blunt the effects of a lot of the high end microchip sanctions and embargoes that the US had imposed upon them, and they now, as you said, get to share a code with the world and all the soft power benefits that come with that ability. Janice, it’s just to me, amazing that a top down regime in the form of the Communist Party of China is using open source the West’s own kind of one of the West premier systems for technological innovation and collaboration to undo the West and particularly America’s dominance in an AI. The ironies here are so rich.

JANICE GROSS STEIN: you know, again, absolutely right. And this is the story of Chinese innovation. They did not innovate in large language models, right? That breakthrough came right here in the city of Toronto. But what they’re brilliant at is taking a very high value expensive innovation and re engineering down. They still used Nvidia chips Rudyard, but from everything we’re able to discern, and we will know more but from everything we’re able to discern, they used Nvidia chips legally. So it’s not easy. Export controls were irrelevant. They re engineered it down so that it becomes affordable and mass produced and scalable. That’s what the Chinese are brilliant.

RUDYARD GRIFFITHS: Yeah, they can innovate, and we should never underestimate their ability to do so. Janice Gross Stein, founding director of the Munk School of Global Affairs. Thank you so much for coming on up to speed today to talk about this fast breaking news, check out, ladies and gentlemen, Deep seek.

The Hub Staff

The Hub’s mission is to create and curate news, analysis, and insights about a dynamic and better future for Canada in a single online information source.

Go to article
00:00:00
00:00:00