Powers of AI: A Conversation with Chris Schroeder

Click here for the original version.

What does the world of AI look like? This is one of the questions answered by Chris Schroeder, former Washington Post publisher, essayist, and investor in Silicon Valley. From the evolution of the US-China rivalry to the future of information, he takes stock of the trends that are already shaping our world.

Will AI accentuate the technological divide and the segmentation of a world we have never seen before? Is it relevant to speak of European, Chinese, or American AI?

Chris Schroeder: The obvious answer is that it is still very early to be able to step back and analyze, but there’s no doubt that there is at least one very American approach and one very Chinese approach. In Silicon Valley, many believe that two AIs already exist. Regarding the underlying infrastructure (cloud, computing capabilities), there will not be much scope for interconnection. There is an interesting debate on how other countries view the issue.

The key in this race is to exercise leadership by mastering AI infrastructure. Does this come down to a question of financial capital or is it about the energy resources available to the players? How many graphics processors and how much computing capacity does a country have? China and the United States have a huge advantage in this field, but other groups of countries could surprise us.

I posed the same question to someone who is very involved in AI and at a high level in the United States. He replied that, in his opinion, there would be surprises. Besides [the AI summit in] London, prominent universities such as Cambridge and Oxford, DeepMind, and the United Arab Emirates can contribute to the debate. I have been following the UAE very closely as part of my work in technology and innovation. It was, of course, the first country to have a minister of AI. A few years ago, this was considered almost naive, but in fact it was extremely far-sighted. The minister was the real driving force behind [AI development]. The country now has the money, energy, and resources, and does not have a narrow vision of the notion of AI sovereignty. For example, it welcomes talent from around the world and wants to be able to attract the world’s best AI thinkers to become a hub in this field. It is acknowledging that sovereignty cannot be defined by nationality alone.

Are you concerned about a potential divergence between the United States and Europe on regulations and standards? Do you think this could harm the global AI ecosystem in the short term?

We talk about competition, and rightly so, but competition by definition means not always agreeing. It's about knowing who we are competing with. Competitiveness is often thought of as simply a matter of opposition between the United States or the West and China, but in reality there is competition among all sovereign nations, which leads to disagreements. I believe we need to accept this reality. The next step is to realize that, in the new world we live in, the best way to find common ground is through co-authorship.

Do we want another spectacle of American technology setting the rules for the whole world? American leadership will be powerful if it makes room for co-authorship. I think it is perfectly understandable that Europe and certain European countries consider it might be time to take a different path. There is an opportunity for Americans to shift toward a different model for designing their technology, one that we will have to embrace.

When it comes to regulatory coordination among jurisdictions, a first issue is that many players in the regulatory field do not really have the necessary experience to fully understand the new nature of the technology before us. Secondly, this is one of the first times in history that regulation has intervened to try and anticipate the possible consequences of technology. Historically, it is as if there were an international nuclear energy agency or a nuclear regulatory industry before a bomb went off or before reactors were created. So, it's unusual but also interesting. The ecosystem has agreed to take a certain lead. Therefore, there is a co-reflection between businesses and regulators in both the United States and Europe.

But I believe there is a real warning to be sounded. In the name of sovereignty and competitiveness, we are also losing sight of what we share. Between Europe and the United States, we share common values which, in my opinion, make us unique in once again being able to co-author the rules of a very new technology.

Do you envision a world in which a few countries are rich in graphics processors—and therefore have the computing power to master AI—and most countries becoming dependent on them? How can Europe catch up?

The idea of working toward greater democratization in access to graphics processors—rather than a scheme in which the United States would be cut off from the West, from China, cut off from everything—is very interesting. It may be a valuable point to argue that ensuring access to computing capacity for other countries helps unleash their economies, enabling children to have access to a better education, and that this is actually a good thing. This must not become yet another divisive issue between North and South.

Another key issue is a country’s access to talent, and I think this aspect is sometimes lost sight of in this debate. Do you have young people who feel they can build where they are? I have seen this in my work on emerging markets. If the latter can build without hindrance at home, they prefer to be in their own countries. Both issues need to be addressed.

The two great AI powers, China and the United States, have diametrically opposed political systems. How important is the role of political regimes in the AI race?

There are two ways of looking at this issue, which affects technology in general as much as AI. Silicon Valley and, I think, many Americans believe that innovation should be free. They generally consider that centralized and planned innovation will always have its limits. When it comes to disruptive innovation, most technologies used worldwide today have their roots in more open Western societies. But it would be too simplistic to say that in China, Alipay or WeChat Pay are not innovations.

Regarding AI, different political regimes make different choices. Some of China’s choices will make us in the West uncomfortable but will give China a competitive edge. We already have examples in genetics or biotechnologies. I think we are right to put up barriers and not use some of the data that China might be ready to use. But we need to question ourselves and recognize that, in the context of this compromise, China may make a breakthrough in some areas in which we are lagging.

The focus is currently on the American AI ecosystem. But how do you assess the maturity of Chinese companies and universities in this field?

I don't invest in China per se, but I talk to Chinese entrepreneurs all the time. Two things are happening. On the one hand, the energy, tenacity, and talent cannot be ignored. The argument is sometimes made that we are ahead of them in the field of generative AI. This does not mean they are not using these capabilities in sectors that are extremely strategic for them, such as electric vehicles, which is undeniable and should be commended.

On the topic of AI more directly, Kai-Fu Lee's new startup, which aims to create an OpenAI for China, is already producing very competitive models. On the academic side, Tsinghua University has also demonstrated that Chinese researchers are capable of producing language models that can rival their US or European equivalents. The ramifications in the Chinese economy over time will be exceptional. Part of me, naively, has always hoped that there might be common interests in applying technology in areas such as biodiversity or environmental issues, and even in international stability and security.

What impact could the upcoming presidential election have on US competitiveness in AI?

The question is how new regulatory frameworks and intergovernmental discussions might be affected by the new administration. Uncertainty and regulatory vagueness may have an impact on the American AI ecosystem. In my own experience of assessing regulation, I have found that in the first draft of the European AI Act text or the American executive order of October 30, much is vague enough to allow a gray area on acceptable risks.

I hope that both major parties will be convinced that we have entered a new era with unique issues to work on. There are still many candidates for the election, so I think we need to remain very cautious and not jump to conclusions, even about what another Biden administration, or another Democratic administration, or a Republican administration would do.

Several sources report that many Chinese students find positions in laboratories in the United Arab Emirates after graduating from Chinese universities. Is the US Entity List, which complicates access to US labs for these same students, an advantage for the development of AI in [the Middle East]?

That is what is interesting because there are certainly a lot of talented individuals in China who believe they would be less limited if they went elsewhere. A very experienced friend from Singapore asked me, "What is the best thing that has happened to the Chinese AI industry in the last ten years?" According to him, US immigration policy, i.e. the fact that the United States makes it harder for some Chinese students to enter AI labs, has enabled other parts of the world to become more attractive.

China itself is benefiting from it. Now, it is also true that there are many good reasons to be very careful about the individuals involved in developing such a strategic technology for a state. On the one hand, there is a strategic imperative to ensure that your assets are protected, and, at the same time, there is a huge amount of talent wanting to come to the United States. Everything we have done to restrict this access for talent from China or elsewhere is certainly worth discussing.

According to Sam Altman, the Gulf region could "play a central role in this global conversation (about AI)". What impact might the geopolitical situation in the Middle East have on the technology ecosystem?

The current political situation obviously creates enormous uncertainty, which is detrimental to any innovation ecosystem. I believe we have learned over time that there are deep-seated structural issues in the Middle East and elsewhere in the world. These include not only economic instabilities but also the way countries treat each other and their own citizens.

Over the past two decades, we might have thought that technology would be the solution to all these problems, but that has not happened. That being said, extraordinary things are happening in the Gulf, especially in the UAE and Saudi Arabia, which want to rely very deeply on these technological capabilities in their own right, but also as tools to align the region more closely with the 21st century.

How do you see AI transforming the media industry and journalism?

I have a journalism background and ran a company in the sector. I am a bit old-fashioned on this issue because I believe that facts are important in this day and age, and that the job of journalism is to make us uncomfortable and push us to think about new things. If I had to make a prediction about AI’s impact on information, I would say that we are heading toward giving fake news mass audiences at an unprecedented rate. All modalities will be used—text, but also image or video—to diminish people's ability to understand what is really happening.

We are already seeing these elements in the terrible situation in Israel and Palestine. It is hard to know what is true. The role of journalism is therefore more important than ever, but we will also see significant benefits from AI use in journalism. We will be able to fact-check in ways we have never been able to before, to confirm whether something is false or not, thanks to AI. If I were running a media organization today, I would work closely with journalists to ask the fundamental question, which is the fundamental question posed by AI: What does it mean to be human? There are things in journalism that AI can honestly do better than humans. This question should be their copilot. Conversely, AI is currently devoid of intuition and journalistic acumen. There are great opportunities to be explored in the complementarity between journalism and artificial intelligence.

Translated from the French by Melanie Backes.