In the Global AI Race, China Is Gaining on the U.S.
14:08 Share

No problem.

Welcome to Tech News Briefing. It's Thursday, July 3rd. I'm Katie Dayton for The Wall Street Journal.

A hybrid electric airplane startup promises to solve a host of issues with flying, including long runways and loud takeoffs. We speak to its CEO about how he plans to do it. Then we're checking in on the international AI race, where Chinese companies are hot on the heels of their American counterparts.

But first, if you've ever dreamed of owning a jetpack, a Virginia-based startup called Electra is engineering what it thinks could be the next best thing. A small plane that can take off and land more like a helicopter, but without the noise or downdraft. Electra has backing from Lockheed Martin and expects to have its EL9 planes flying in the sky in 2029. We're here with the company's CEO, Mark Allen, to learn about how it's going to work. Mark, so...

In our story, we have these wonderful visuals of your plane taking off and landing. Unfortunately, we don't have that visual medium here. Mark, could you explain to listeners what a takeoff and landing looks like in an electroplane? When you watch an airplane take off, you expect to see it

make its way down the runway, build speed, and then jump up in the air. This airplane just skips those first two steps. It just jumps straight up in the air just as quickly as it starts moving. I mean, literally less than three seconds. And I think most people would imagine that that looks like a helicopter in that case. What's the difference here? It ends up operating in the exact same profile as a helicopter, but it's just a plane. You can tell because from the minute it leaps in the air, it's wing-borne.

It's not fighting to go straight up with rotor blades spinning hard and beating all that air down, creating downwash, creating noise. It just whisper quiet. It just happens to land like a helicopter. And I wondered, what kind of pitfalls did you stumble upon in development of the plane? Was there anything major that you had to readjust once you started building and flying these things? It's a super difficult problem to solve, to fly slowly.

with total stability and control. It's been solved through the flight physics work around the aerostructure. It's been solved through the algorithms that are part of the flight control computer and the flight management system.

And so the computer is doing a lot of hard work to make it super simple for the pilot to fly. So it's more like a Tesla kind of screen that we're seeing ahead of the pilot as opposed to all of those buttons and dials that none of us understand if we don't fly planes. That's right, Katie. It presents very differently. In aviation, we call it the six pack. There are these six principal dials that have been the very core components.

of the airplane cockpit interface for the pilot for generations. If this plane entered the market right now, what would the commercial proposition be for consumers and for airlines too? We have a $2,200 plus backlog of provisional orders, which is some $13 billion in sales value.

And the customers that are buying it are going to use it for a lot of different things. So for example, one right now flies nine seaters on short routes across Northeast quarter, but it has to wait on the runway for 15, 20 minutes before it can take off. And it burns fuel the whole time. Our airplane will use the helicopter departure and then it'll fly those short 20, 30 minute routes. And it'll thereby reduce the operating costs by about 40% for that customer. We have another customer who flies very small jets and,

And they're starting to get prohibited from flying into certain airports that are shutting down to noise, places like Santa Monica. And there are a lot of other communities out there that are also having fights about airport noise. They're buying our airplane so that they can offer the community a way to stay connected to move passengers and cargo, but to do it totally quietly.

And then a third customer lives and operates down in the islands, and they move goods and people between different islands. There are some they simply can't fly into because the runways are too short. So you're solving for a lot of things here, having to travel to the airport, having to get through the airport, noisy, sometimes scary takeoffs. Why hasn't this been done before? There were a number of small light jets that tried to do this 25, 30 years ago.

But we've finally brought a number of overlapping, intersecting aviation technologies together to do something that you just couldn't have done 10 years ago. It's hard. It's really hard to be able to have the level of stability and control at super slow speeds that we

that we've achieved. But now with all the advances in electric vehicles that are out there around the world, there's a ton of great learning. Batteries, they continue to advance. The ability to use sophisticated algorithms for the flight control system I described earlier, those also are unlocked by the ways in which compute is so radically advancing these days. The whole tech stack has grown together and the best innovation happens when you intersect different parts of that advancing tech stack to do something that's just unheard of.

That was Mark Allen, the CEO of aerospace startup Electra. And you can see a demonstration of Electra's plane on WSJ.com. We've added a link in the show notes. Coming up, you've heard of OpenAI and Anthropic. But what about AI from Baidu and DeepSeek? Chinese companies are gaining ground on their American rivals. Find out how they're competing after the break.

This episode is sponsored by Morgan Stanley's Thoughts on the Market. Today's financial markets move fast. Morgan Stanley moves faster with their daily podcast, Thoughts on the Market. Thoughts on the Market covers daily trends across the global investment landscape with actionable insights from Morgan Stanley's leading economists and strategists. And with most episodes under five minutes long, staying informed has never been easier. Listen and subscribe to Thoughts on the Market wherever you get your podcasts.

If you live in the US, it can seem like our homegrown companies, Google, Meta, Microsoft, are the big names on campus when it comes to generative AI. And they were, until Chinese companies entered the picture. In Europe, the Middle East, Africa and Asia, users ranging from multinational banks to public universities are turning to large language models from Chinese companies, putting American superiority at risk.

WSJ reporter and editor Lisa Lin has been following the competition. So Lisa, if we were looking at China and America as two cars in a race over the last few years, who has been in the lead and when did that gap start closing up?

With generative AI, you just have to go back to the end of 2022 when ChatGPT came out. And it really changed the game. At that point, OpenAI and ChatGPT in the realm of generative AI was way ahead of the Chinese. After ChatGPT came up, a Chinese rival called Baidu was quickly out of the door.

with a new generative AI model. But at that point, the Chinese lagged very firmly and very much behind the US. However, fast forward to the start of this year in January, when a Chinese company called DeepSeek unveiled its latest generative AI model, you've really noticed like a huge improvement. And the Chinese models have just since then been going from strength to strength.

Some of them are leading benchmarks and others are just displaying performances just not very far from their Western peers. You mentioned in your reporting that some big household names like HSBC, Standard Chartered, they're testing these Chinese models. What is it that Chinese AI firms are doing differently to American counterparts that's been so compelling for these companies?

So first and foremost, I have to say that the US products are still in the lead. If you look at the use of ChatGPT, for example, it's way and by far the most used generative AI app in the world. However, the one big realization from the introduction of DeepSeq at the start of the year was that DeepSeq was able to match the benchmarks of its Western rivals and

or come close at least, and do it at a fraction of the computing power and the cost that was required. So this is actually super attractive to marketers

countries in, for example, Latin America or countries in Asia, Southeast Asia in particular, who might not have the computing resources or the financial resources to deploy on ChatGPT or Gemini or the US equivalents. Because ChatGPT and Gemini, a lot of these models are closed source, and that means they charge a premium for the product.

What the Chinese have done is not only have they been able to train their models more efficiently and on less compute, meaning less cost, they've also made their models open source. And when they make the models open source, it's very easy for a developer in any part of the world to create their own AI agents on it. And I wonder...

What's been going on behind the scenes geopolitically that's led to this outcome that we're looking at today? How much of a hand have the U.S. and Chinese governments had? So the reason why this U.S.-China AI rivalry is watched so closely is because the backdrop to all of this is

was in late 2022, probably about the same time when CHAP GPT was released, the US government imposed export controls on China. And at that point, they blocked the sale of advanced chips and chip making technology to the country. And all this was done in the name of national security.

And since then, the Chinese government has really doubled down on an internal effort to create its own domestic Chinese supply chains, not just for chips, but for AI models, so that they can be independent of the US, the West and its products.

And aside from simply losing market share, what concerns do major American figures have with AI from China gaining dominance and Chinese AI maybe getting more integrated into our lives in the West? The biggest concern is censorship and the idea that the tech we use isn't agnostic.

You know, the tech we use actually carries the values of their founders or their developers. And in this case, Chinese generative AI might carry the viewpoints of the Chinese Communist Party. We've tested DeepSeek's app at the journal, and it's shown that the answers do carry a certain bias. This, however, doesn't really apply to open weights. Open weights, meaning basically when the parameters used by AI are open source. How does that factor in here?

So when you open source your models, just the way that DeepSeek, Alibaba and Baidu have, and what developers are taking are the open weights and not just using the consumer product developed by these companies, then the censorship in theory should not exist. Lisa, you write that Chinese companies have started to snap up customers by offering much lower prices.

How could that change the AI race for US companies? There is a view, and this view is gaining in popularity, that because the Chinese are making their models open sourced, at some point it's going to put some pressure on Western rivals such as Gemini or OpenAI and Tropic, for example, to justify why they're charging consumers high premiums for using their products.

The decision for these Chinese companies to have an open source model, was that because they knew it would undercut the prices? Or was there other reasons why they wanted to go down that route? The decision to open source the models wasn't purely because they wanted to undercut Western rivals. They do want to compete globally, that we know. But the biggest reason driving the move to open source the models is actually because when

When you have a closed source model, you only have the developers working for your company tweaking the model, making iterations. But when you open source your model, you have developers from the whole world working on making iterations and making your model better. That was WSJ reporter and editor Lisa Lin. And that's it for Tech News Briefing. Today's show was produced by Julie Chang. I'm your host, Katie Dayton.

Additional support this week from Melanie Roy, Jessica Fenton and Michael LaValle wrote our theme music. Our development producer is Aisha Al-Muzlim. Scott Salloway and Chris Tinsley are the deputy editors and Falana Patterson is the Wall Street Journal's head of news audio. We'll be back this afternoon with TMB Tech Minute. Thanks for listening.

How businesses connect with customers defines their brand. Sierra is the AI platform for building better, more human customer experiences. Fast answers, no canned responses, no hold music, no frustration. Visit sierra.ai to learn more.