How would you like your AI? Capitalist style or neo-Marxist?

Column published in l'Opinion as part of the summer series "The odyssey of artificial intelligence". This is the 5th episode .

During the 2010s, China was certain of its technological superiority, performing most impressively in fields such as quantum computing. The announcement of the arrival of ChatGPT3 heralds the end to Chinese self-assurance.

Generally speaking, the 7th of October 2022 was a date that went unnoticed in Europe, but it will go down in history as the Pearl Harbor of the 21st century. A few days before the 20th Chinese Communist Party Conference that was to grant full powers to President Xi, Washington decided to cut Beijing off from American state-of-the-art technology in three areas fundamental for the future: high-performance microchips, quantum computing, and artificial intelligence (AI). The historic nature of this decision only became evident at the end of November 2022 when the whole world was presented with a revolution in generative IA that could create content – the ChatGPT-3 model.

This date of 7th October, taken together with the outbreak of the war in Ukraine in February, bestows a pivotal role on 2022: it ends the thirty-year period following the fall of the Berlin Wall in 1989, during which there was an openness, freedom and a boom in world trade, and which is now giving way to frontiers, separation and re-regulation.

Three decades of a “peace economy” are giving way to a new “war economy”, developing as a new form of “hybrid war”, just as the two Chinese strategists Qiao Liang and Wang Xiangsui had already predicted in 1999 in their – now standard – work, Unrestricted Warfare. The new types of combat they mentioned already included technological warfare, which exactly describes the new artificial intelligence war between China and the United States.

Supremacy. And yet, the 2010s had promised a very different scenario. The Chinese technology guru Kai-Fu Lee, who had worked for Google, was categorical about it. In his essay AI Superpowers: China, Silicon Valley, and the New World Order, his convincing argument was that, thanks to AI, it would not be long before China ousted the United States from its position as technology supremo. Peter Thiel even said that artificial intelligence was an innately “communist” technology, because it attempts to foster normative thought around the lowest common denominator.

In concrete terms, China had a far more extensive database than the West, which was penalised by totally unproductive over-regulation that divided human activity into separate sectors. China, therefore, was preparing to benefit from a major competitive advantage from its handling of non-structured correlations, which are the very source of the greatest added value of the new approaches made possible by AI.

For example, knowing the lowest percentage of battery charge in a mobile phone that can last a full day means that, at minimum cost, you can determine the user's degree of risk aversion and therefore estimate the likelihood of his or her repaying a consumer loan. It mattered little that great world-renowned experts like Cédric Villani pointed out the difficulty of getting hold of good quality non-structured data, or that the same human being may display different behaviour patterns depending on what he or she is doing… In the 2010s, China was particularly sure of its technological superiority, performing impressively in fields such as quantum computing, where its computational power was regarded as nearly equal to that of the United States.

When the American company OpenAI announced the arrival of ChatGPT3 in late November 2022, this heralded the end of China's self-assurance. We had to wait several months before China and its national market leader Baidu presented their own rival model, Ernie 3.5, to an audience of financial analysts who were obviously underwhelmed, since its shares had dropped 10% by the end of the day.

Since then, in the face of general scepticism, Baidu has vainly attempted a counter-attack. However, four weaknesses in China's situation may well mean that Uncle Sam will hold on to its technological supremacy.

First of all, the superiority of Western databases is overwhelming, both quantitatively and qualitatively, since 56% of the world's internet sites are written in English and only 1.5% in Chinese. Furthermore, Chinese personal data is kept for the most part in inaccessible “walled gardens”.

The second weakness concerns regulations in China, particularly the new law on data protection which does not define what “sensitive” data is. These regulations do not facilitate exchanges between researchers in the academic sphere.

Thirdly, Chinese AI remains dependent upon American high tech with regard to hardware, as in the case of Nvidia's processors which drive almost half of the Chinese Large Language Models. On the international front, the United States continues to attract the best brains in the world, since two-thirds of AI experts are foreign nationals. Only 30% of the recent class of graduates preparing the Data Science masters degree awarded by Harvard's Kenneth C. Griffin Graduate School of Arts and Sciences were Americans. 50% were Asian and rather keen on staying in the USA – at least for the time being.

The fourth weakness lies in the role of the private sector in financing AI. Whilst, according to The Economist, this only totalled $13.5 billion in China in 2022, it reached almost $50 billion in the United States – more than three times as much. This gap is not likely to be narrowed in the near future if Xi Jinping's government continues its concerted attack on the private sector.

However, can we predict with any certainty that Uncle Sam will come out on top??Is there not a scenario which could see both sides victorious, most probably at Europe's expense? For this to happen, one has to accept the premise that each of the two belligerents is pursuing a different objective.

?

Control. Following on the 20th Party Congress, the powers-that-be in China are above all seeking greater control, even if it means less economic growth. AI that generates content in neo-Marxist style is seen as the ideal means of controlling the output of the population's brains, the perfect complement to control of the input, which has already happened through censorship of social media. Therefore, Ernie 3.5 should be the natural extension of Chinese advances in facial and voice recognition, which were the priorities of leading Chinese companies such as SenseTime and Megvii when they made their first attempts at AI applications in the 2010s.

After the presentation of ChatGPT-3, we had to wait several months before China and its national market leader Baidu presented their own rival model, Ernie 3.5, to an audience of financial analysts who were obviously underwhelmed, since its shares had dropped 10% by the end of the day.

On the other hand, in the grip of the “capitalist style” of the new Magnificent Seven – Microsoft, Alphabet, Amazon, Meta, NVidia, Tesla and Apple – the United States will be looking to use generative AI to make productivity gains similar to those achieved by the coming of the PC in the 1970s. This is a potentially decisive advantage, recently estimated by the Goldman Sachs investment bank as representing seven additional percentage points in terms of productivity, spread over a decade. There is, however, one condition: in the next five years, they have to get closer to the workings of the human brain and introduce the faculty of thought, common sense and emotional intelligence into Large Language Models.

It is therefore obvious that 2022 has speeded up history, much as 1989 did. Back then, there were two major errors of judgment, and the same might happen now. The first was to oppose the reunification of Germany, as France did initially and unsuccessfully. The second was to say that this was “the end of history” by announcing the triumph of liberal democracy, as the American political scientist Francis Fukuyama claimed.

Therefore, we Europeans must be extremely vigilant in face of the dangers from two equally deficient ways of thinking. We must not dilute the “capitalist style” of American AI too much by European over-regulation. Neither should we swallow whole the “neo-Marxist-Leninist style" of Chinese AI, for this would signify that we were giving up the fight and leading the old Western democracies towards decline.

Read all the columns (in French) published as part of the "Artificial Intelligence Odyssey" summer series on L'Opinion.fr:

https://www.lopinion.fr/dossiers/lodysee-de-lia