Anthropic CEO Urges Trump to Win AI vs. China
14:28 Share

Hello, friends. Today we have a very interesting Long Reads episode for you.

I've got two pieces that I'm going to share, both about the U.S. and China and what the Trump administration should do next about AI. The first piece is notable not only for its opinion, but from its provenance. Appearing in the pages of the Wall Street Journal, the piece published this week was called Trump Can Keep America's AI Advantage, and it's by Matt Pottinger, chairman of the China Program at the Foundation for Defense of Democracies, but it's also by Dario Amodei, the CEO of Anthropic.

Let's listen to that one first. This is, of course, read by an 11 Labs version of me, and then we'll come back here. Trump can keep America's AI advantage. Legislators on both sides of the aisle recognize that the U.S. must lead the world in artificial intelligence to preserve national security. This gives the incoming Trump administration a chance to establish a historic advantage for the U.S. and the free world. AI will likely become the most powerful and strategic technology in history. By 2027, it will be the most powerful and strategic technology in history.

AI developed by Frontier Labs will likely be smarter than Nobel Prize winners across most fields of science and engineering. It will be able to use all the senses and interfaces of a human working virtually — text, audio, video, mouse, keyboard control, and internet access — to complete complex tasks that would take people months or years, such as designing new weapons or curing diseases. Imagine a country of geniuses contained in a data center. The nations that are first to build powerful AI systems will gain a strategic advantage over its development.

Incoming Trump administration officials can take steps to ensure the U.S. and its allies lead in developing this technology. If they succeed, it could deliver breakthroughs in medicine, energy, and economic development. It could also extend American military preeminence. If they fail, another nation, most likely China, could surpass us economically and militarily. It's imperative that free societies with democratic oversight and the rule of law set the norms by which AI is employed. They won't be able to do that if they fail.

They won't be able to do so if totalitarian governments pioneer these technologies. Export controls which ban shipments to China of the high-end chips needed to train advanced AI models have been a valuable tool in slowing China's AI development.

These controls began during the first Trump term and expanded under the Biden administration to cover a wider range of chips and chip manufacturing equipment. The controls appear to have been effective. The CEO of one of China's leading AI firms recently said the main obstacle he faces is the embargo on high-end chips. China is trying to work around U.S. controls, including by using shell companies to set up data centers in countries that can still import advanced U.S. chips.

This enables China to train its AI models on state-of-the-art chips and catch up with U.S. competitors. The Trump administration should shut down this avenue of circumvention. One solution is to ensure that data centers in countries that China might use to skirt export controls are allowed to access U.S.-designed AI chips only if they adhere to verifiable security standards and commit not to help China's AI efforts. AI hardware exports should be tracked.

We should also ensure that frontier AI remains under our security umbrella by keeping the largest and most critical AI data centers within the U.S. and its closest partners. Skeptics of these restrictions argue that the countries and companies to which the rules apply will simply switch to Chinese AI chips. This argument overlooks that U.S. chips are superior, giving countries an incentive to follow U.S. rules. China's best AI chips, the Huawei Ascend series, are substantially less capable than the leading chip made by U.S.-based NVIDIA.

China also may not have the production capacity to keep pace with growing demand.

There's not a single noteworthy cluster of Huawei Ascend chips outside China today, suggesting that China is struggling to meet its domestic needs and is in no position to export chips at a meaningful scale. Because of America's current restrictions on chip manufacturing equipment, it will likely take China years, if not decades, to catch up in chip quality and quantity. The CEO of ASML, the world's largest maker of semiconductor manufacturing equipment, has said that these restrictions will cause China to lag 10 to 15 years behind the West in high-end chip manufacturing.

That could give the U.S. a head start during a critical window. Whoever advances most during the next four years will be in a much stronger position in the decades that follow, given that AI gains will likely compound on one another. The export and security terms that the U.S. sets will define the chip market for producing powerful AI systems.

Countries that want to reap the massive economic benefits will have an incentive to follow the U.S. model rather than use China's inferior chips. Along with implementing export controls, the U.S. will need to adopt other strategies to promote its AI innovation. President-elect Trump campaigned on accelerating AI data center construction by improving energy infrastructure and slashing burdensome regulations.

These would be welcome steps. Additionally, the administration should assess the national security threats of AI systems and how they might be used against Americans. It should deploy AI within the federal government, both to increase government efficiency and to enhance national defense. Mr. Trump has likened AI to a superpower and has underscored the importance of the U.S. staying right at the forefront of its race against China. His administration's actions will help determine whether democracies or autocracies lead the next technological era.

Our shared security, prosperity, and freedoms hang in the balance. All right, back to real NLW here. So this is a pretty hawkish piece. And not that I know anything about Dario's politics in particular, but this is very reminiscent of an earlier piece by Sam Altman saying very similar things. So we now have the leadership of the two big AI lab startups, both sending a very similar message on trying to preserve and extend the US's lead when it comes to AI vis-a-vis China.

Today's episode is brought to you by Vanta. Trust isn't just earned, it's demanded. Whether you're a startup founder navigating your first audit or a seasoned security professional scaling your GRC program, proving your commitment to security has never been more critical or more complex. That's where Vanta comes in. Businesses use Vanta to establish trust by automating compliance needs across over 35 frameworks like SOC 2 and ISO 27001.

Centralized security workflows complete questionnaires up to 5x faster and proactively manage vendor risk. Vanta can help you start or scale up your security program by connecting you with auditors and experts to conduct your audit and set up your security program quickly. Plus, with automation and AI throughout the platform, Vanta gives you time back so you can focus on building your company. Join over 9,000 global companies like Atlassian, Quora, and Factory who use Vanta to manage risk and prove security in real time.

For a limited time, this audience gets $1,000 off Vanta at vanta.com slash NLW. That's V-A-N-T-A dot com slash NLW for $1,000 off.

If there is one thing that's clear about AI in 2025, it's that the agents are coming. Vertical agents by industry, horizontal agent platforms, agents per function. If you are running a large enterprise, you will be experimenting with agents next year. And given how new this is, all of us are going to be back in pilot mode.

That's why Superintelligent is offering a new product for the beginning of this year. It's an agent readiness and opportunity audit. Over the course of a couple quick weeks, we dig in with your team to understand what type of agents make sense for you to test, what type of infrastructure support you need to be ready, and to ultimately come away with a set of actionable recommendations that get you prepared to figure out how agents can transform your business. If

If you are interested in the agent readiness and opportunity audit, reach out directly to me, nlw at bsuper.ai. Put the word agent in the subject line so I know what you're talking about. And let's have you be a leader in the most dynamic part of the AI market. Next up, let's read a piece by Tyler Cowen.

Tyler is professor of economics at George Mason University, the host of the Marginal Revolution blog, and a very well-respected thinker. The piece is called China's Deep Seek Shows Why Trump's Trade War Will Be Hard to Win. Once again, I'm going to turn it over to an 11 Labs version of myself.

China's deep-seek shows why Trump's trade war will be hard to win. Breakthroughs in AI are so common these days, it is hard to separate the truly important from the merely incidental. But one recent development is worth paying particular attention to, the appearance of DeepSeek Phi 3, a new large-language model from China.

Its significance has as much to do with trade as with technology. I've played around with DeepSeek for several days, and it is one of the best LLMs of the dozens I have used over the last few years. It is fast, easy, and has a free version. And while it is not the equal of the best US models for sophisticated or difficult questions, I would rate it in the top tier.

That is consistent with the opinions of others, including testers. There are several other notable things about DeepSeq. First, it comes from a hedge fund rather than a technology company that said, these categories are probably in need of revision. Second, it was reportedly trained at very low cost, by some estimates only $5.5 million, excluding non-compute costs as these measures usually do. And perhaps most notably, DeepSeq does not make use of the highest quality semiconductor chips.

President Joe Biden's administration has worked hard to limit the export of those chips to China for reasons of national security. The U.S. wanted to slow Chinese progress in AI and related military technologies. Without access to the latest chips, DeepSeek had to look for different and cheaper ways to train its model. I have in the past supported these trade restrictions as AI technology is a vital matter of national security. But I now think the ban was too ambitious to work. It may have delayed Chinese progress in AI by a few years, but it's not the end of the

but it also induced a major Chinese innovation, namely DeepSeek. Now the world knows that a very high-quality AI system can be trained for a relatively small sum of money. That could bring comparable AI systems into realistic purview for nations such as Russia, Iran, Pakistan, and others. It is possible to imagine a foreign billionaire initiating a similar program, although personnel would be a constraint. Whatever the dangers of the Chinese system and its potential uses, DeepSeek-inspired offshoots in other nations could be more worrying yet.

Finding cheaper ways to build AI systems was almost certainly going to happen anyway. But consider the trade-off here. U.S. policy succeeded in hampering China's ability to deploy high-quality chips in AI systems, with the accompanying national security benefits, but it also accelerated the development of effective AI systems that do not rely on the highest quality chips. It remains to be seen whether that trade-off will prove to be a favorable one, not just in the narrow sense.

Although there are many questions about DeepSeek's motives, pricing strategy, plans for the future, and its relation to the Chinese government that remain unanswered or unanswerable.

The trade-off is uncertain in a larger sense, too. To paraphrase the Austrian economist Ludwig Mises, government interventions have important unintended secondary consequences. To see if a policy will work, it is necessary to consider not only its immediate impact, but also its second and third order effects. One secondary effect of the chips restriction is that it may encourage some Chinese sources to obtain high-quality chips through third parties in other countries, or to rent time on non-Chinese AI systems that use higher-quality chips.

In that case, Chinese firms don't need to buy the chips at all, at least for some purposes. The U.S. is responding with further controls on the sector, but can it really micromanage a global market? I'm increasingly skeptical.

As it considers further trade restrictions against China, President-elect Donald Trump's administration would do well to study the unintentional consequences of its predecessors' policies. To be sure, there is a national security case for some, not all, of the non-AI trade restrictions. But the first-order effects of any policy are rarely the end of the story. If the federal government decided to restrict or tax a Chinese good or service sold in the U.S., for example, China could try to sell the same item by rebranding it through a third party.

as many Asian countries are willing to help out. Rule-evading entrepreneurs tend to move more swiftly than bureaucrats. In the abstract, national security arguments are compelling. In reality, however, it's difficult to design policies to protect national security. It's important to think through ways to help the practice better match the theory.

All right, back to real NLW here again. I tend to find myself agreeing with Tyler Cowen, particularly around the topic of AI. When Tyler Cowen writes about AI on Bloomberg, I've generally found myself agreeing fairly vociferously. This one, though, rings a little flat for me. It's not that he's wrong that export restrictions have created an incentive for Chinese companies to get more innovative about model training. It's that I think that that kind of misses the point in some fairly fundamental ways.

First of all, to think that that's the only incentive to decrease the cost of training, or even the main incentive to decrease the cost of training, just doesn't make any sort of sense. In fact, one of the things that was really fascinating about 2024 is how much the AI space bifurcated to a competition at the state-of-the-art and a competition on the far other end of the spectrum for more high-performing models that weren't as cumbersome, laborious, and big. Now, to some extent, that was less about cost considerations and less about the cost of

and more about being able to run on edge devices, but the point remains that this area of competition is a big one, even in the US, where everyone has full access to these chips. Second, being able to catch up for cheap to GPT-4.0 is very, very different than being able to use extremely low-cost training models to achieve the state of the art.

Remember, the competition right now and what people like Dario Amodei and Sam Altman are talking about isn't whether or not China has access to the current level of AI assistance. It's about artificial general intelligence and even superintelligence. For them, that's what the real stakes of these export restrictions are.

And so the idea that simply because there has been a second order effect of China having an incentive to figure out how to train models more cheaply, it does not follow to me that that means that those export restrictions should be jettisoned. Now, there are plenty of reasons to have that conversation in more full and fluid terms. I just don't think that this is among them.

Anyway, super interesting stuff. I always appreciate when there's a week where there's somewhat different views on the same topic so we can get a variety of perspectives. For now, though, that's going to do it for today's AI Daily Brief. Appreciate you listening, as always. And until next time, peace.