With artificial intelligence to the top of the world – the examples of China and Germany
The term „artificial intelligence“ (AI) is more than 60 years old (it was coined by John McCarthyat a conference in Dartmouth in 1956). But only in the last five to ten years has this field evolved into a future key technology with ever more powerful impact on our lives. Increasingly, it deals with tasks that were previously reserved for human cognitionsuch as recognizing patterns, predicting events clouded by uncertainty, and taking decisions in complex circumstances. AI algorithms can increasingly perceive and interpret the world around us. AI researchers even claim that AI will soon be capable of emotions, compassion,and genuine creativity. But regardless of whether one day they will actually possess these specific human abilities, they can already recognizethose in humans today. Reading emotions from a human facial expression is ow easier for an AI than for other humans. And in chess tournaments where computers are not allowed, uncommon and creative traits are seen as an indication that someone is cheating by secretly using a computer.
What happened that this technology, which only 15 years ago was considered a playground for freaks, suddenly became so powerful? It was the development of a special approach that changed everything. This is referred to as „deep learning“ and describes an architecture with artificial neurons and their interconnections inspired by the human brain. As the name suggests, these networks can be many neuronal layers deep and contain even more parameters. These neural networks are then „trained“ on huge amounts of labeled data. Afterwards, they use what they have „learned“, i. e. how they have, based on the learning data, chosen their many different parameters to detect subtle patterns in other data. So, in addition to the new structural paradigm of deep learning, there has been a second requirement for AI to emerge: huge amounts of data. This is exactly what has become available due to the immense increase of our online activities: The American Internet companies Google, Facebook, and Microsoft, but increasingly also their Chinese counterparts, Baidu, Tencent, and Alibaba, are collecting, storing, and using all the data they can about our behavior, our preferences, and our intentions that we so voluntarily provide to them. With this combination of computing power and data, AI quickly became better at understanding language and text, recognizing faces, playing chess and Go, examining MRI images and skin tissue for malignant tumor cells, and calculate likelihood of credit defaults or credit card fraud.
However, all this also means that the focus of AI development has changed significantly in the past two to three years, from projects in top research laboratories in specialized institutes (including those at Google, IBM, or Facebook) to real-world applications with real-world data. The speed of the development of AI research is demonstrated by the further development of the AlphaGo AI. Only 18 months after its spectacular victory over the best human player Google had already created a new version of Go artificial intelligence. AlphaGo Zero did not need to be fed with past games anymore in order to reach his playing ability. Like the well-known Dr. B from Stefan Zweig’s Chess novella, his developers let him play only against itself and thus learn. Already after three days in which the machine played 4.9 million games against itself, AlphaGo Zero had reached a level of skill in the Go game, which enabled it to defeat his still real-data trained predecessor and human world champion defeater AlphaGo with 100 to zeroin 100 games. No less impressive was AlphaGo Zero’s performance in chess. He beat the hitherto world’s best chess computer, which had been trained with millions of historical chess games and the centuries-old experience of chess players and possessed a processing power of evaluating 70 million positions per second, in 28 games and tied 72 games out of 100 games (thus did not lose once). The amazing thing was: The machine had learned chess only four hoursbeforeby playing against itself. Equipped exclusively with the rules of chess it optimizedits neural connections without ever having been presented any openings or playing strategies, and thereby he was also able to rate „only“ 80,000 positions per second. In just four hours from beginner to becoming the unbeatable, best chess machine in the world! An AI like AlphaGo Zero is so powerful because it is „no longer constrained by the limits of human knowledge,“ says one of the creators of AlphaGo and AlphaGo Zero, Demis Hassabis.
These developments, in turn, lead to another completely new development with severe consequences: Has so far the USA with its leading AI research institutes and software companies been the undisputed leader of the AI revolution, in the last two years, China with its huge market of more than a billion people, its immense and completely unprotected data provided by its internet users, and itstough and aggressive entrepreneurs hasquickly grown into an AI superpower, as entrepreneur and influential investor Kai-Fu Leeimpressively writes in his new bookAI Superpowers. China, Silicon Valley, and the new World Order. Specifically one point that horrifies European proves to be one of the biggest competitive advantages of China: the complete absence of any data protection laws. Even if this is a little debated topic even in the US, the Chinese do not even see the slightest need for discussion on the subject of „personal data protection“. On the contrary, the free access of Chinese Internet companies to their customers‘ personal data is hailed as the greatest advantage of Baidu, Tencent,and the likes in the global competition for leadership in AI. Indeed, in July 2017 the Chinese Communist Party identifiedAI as one of the most important growth areas and has massively promoted it ever since. In particular, the amazing victory of AlphaGo over the world champion in Go, the national game in China, has woken up the political leadership in China I. AI experts already refer to this moment as China’s „Sputnik shock“. The governmenthas begun to literally flood the Chinese economy with funds for AI development. The city of Beijing recently provided $ 2.1 billion to build an AI industrial park in the outskirts of the city, Shanghai and 17 other Chinese cities have similar ambitions, and the city of Tianjin has even announced a $ 16 billion US dollar (100 billion yuan) fund to invest in local AI firms and institutions. There are also extensive public programs to train AI engineers and experts, government grants for AI entrepreneurs, and tax benefits for companies. This has also attracted significant private investment: The total volume of Chinese (private and public) investments in AI and robotics is already estimated to stand at $ 300 billion. Especially in recent months, China has poured more capital into AI than the US. With this new AI ecosystem, a mix of government money, the building of an intelligent infrastructure, massive investment in AI research, and the world’s most ambitious entrepreneurs, China’s rise to become anAI superpowers is hard to stop.
Compare this with the initiative of the German Federal Government, as it is being articulated in its publication Strategy Artificial Intelligence and approved at the Digital Summit on the topic of AI on 4 December. The importance of AI for our future has now been recognized by German politicians. The report clearly states that Germany and Europe have fallen behind in the field. Whether this insight, maybe comparable to that of the Chinese leadership after the human defeat at Go, is linked to the performance of the German national football team in 2018, is unknown. But the goal expressed is ambitious: The Chancellor wants to make Germany the „world leader“ in AI. And for that it is willing to invest some money: 500 million euros per year. Compared to the Chinese investments this is a pocket money. Germany has a lot to catch up in terms of AI: German researchers are hardly visible at major international conferences which are dominated by the American companies Google, Microsoft, and Facebook, as well as increasingly by Chinese scientists and engineers, even if the later still constrain themselves somewhat (they may prefer to keep their results to themselves). And all these companies are willing to pay up: The starting salaries for trained AI experts lie between 300,000 and 500,000 US dollars!
In order make rapid progress in machine learning and AI one needs three things: 1 tremendous computing power; 2. large amounts of data and 3. innovation, i.e. AI experts. At all three levels, Germany and Europe are lagging far behind the US and China. Three billion euros in five years will hardly change that. To arise from latecomer to world leader solely withpocket money is pure wishful thinking. Rather, a more fundamental change is needed. However, where the Germans are far ahead is in recognizing that there is need for a discussion about the responsible design of AI technologies. Thus, the new AI strategy for Germany emerged, as it is being emphasized in the report, „out of a comprehensive democratic process“. Itaims at creating „conditions for the ethical application of artificial intelligence“ and promoting a „social dialogues about the chances and effects of artificial intelligence „. This is very praiseworthy, as such phrases are hardly found in the corresponding statements of intentions of Chinese or American origin. But are these more than noble words? There is talk of a „people-centered development and use of AI applications“ aiming at „a high level of IT security to best prevent manipulation, abuse, and risks to the public safety of this sensitive technology“. That sounds more like political soothing and lulling. We are missing the courage to make a clear statement. Many real AI experts do not shy away from such. KI pioneer Stuart Russel paints a drastic picture of us humans in a car driving towards a cliff hoping to run out of gas before we plunge into the abyss. Like Elon Musk, Russel claims that AI can be as dangerous to humans as nuclear weapons. Experts are literally begging for governments providing a framework of laws and regulations. Behind this stands their serious concern that policymakers oversleep technological developments, do not takeit seriously or, as in most cases, simply do not understandit at all. Thus, even the latest report of the German government illustrates in an exemplary way an old dilemma: the scientific-and technological progress has meanwhile become so rapid, displays such a complex dynamics that it eludes the imaginative and creative potential of the vast majority of political and social decision makers. And with this trickle of pocket money on a future key technology Germany and Europe will soon no longer be playing in the first league. Then they will not have a role in playing out the championship. That would be very unfortunate, because surrendering such an important playing field to the American capitalists or Chinese Communists could prove fatal.