Towards global leadership in Artificial Intelligence – The view of the German government
Just over two years ago, the general public was surprised by the victory of the Google computer AlphaGo over Go World Champion Lee Sedol. Suddenly, many realized how far the learning algorithms of Artificial intelligence (AI) systems have developed already. Because inthe game Go, intuition and creativity are far more important than stubborn arithmetic calculations. Thus AlphaGo is not just a fast computer, but rather consists of an unimaginably large number of learning – one is tempted to say: living – neural connections that mimic the learning and thought processes of the human brain. The computer was fed countless historic Go games and from thoselearned entirely new strategies of the game that even experts had never seen before. This made it clear that the underlying learning and optimization means (so-called “deep learning”) of today’s AI enable a massive and broad increase in machine intelligence. Future computers will no longer deal solely with the particular purpose they were designed for, e.g. play chess or browse through databases, but will operate on a much broader scale.
The speed of this development is illustrated by the further developments of AlphaGo. Only 18 months after AlphaGo’s victory over the best human player, Google created a new version of a Go artificial intelligence. AlphaGo Zero did no longer need to be fed with old games in order to reach his playing abilities. Like the well-known Dr. B in Stefan Zweig’s chess novella his developers only let him play against himself and thereby learn. Already after three days in which it played 4.9 million games against itself, AlphaGo Zero had reached a level of skill in the Go game, which enabled him defeat his predecessor AlphaGo which had been trained in historical games played by humansby 100 to zeroin 100 games. One of the creators of AlphaGo and AlphaGo Zero, Demis Hassabis, said that AlphaGo Zero was so powerful because it was “no longer constrained by the limits of human knowledge”.It is thus becoming increasingly clear: Artificial intelligence research aims to reach broad basedoutperformance of machine intelligence versushuman intelligence.
However, what most people do not yet have on their radar screen: AI is already part of our everyday life. Today computers with AI software optimize energy consumption in industrial plants, create cancer diagnoses, write journalistic texts at the touch of a button, advice bank customers in their optimal investment strategy, and support lawyers dealing with complex legal cases. They read customer letters and e-mails and recognize the degree of annoyance of the sender, and computers in call centers chat on the phone like human employees.AI even writes entire plots for films, composes symphonies and operas, and paints in the style of true masters. And AI-driven “Cognitive Cooking” could become the next big culinary trend.
In all of this the AI’s performance matches at least that of human experts: AI pattern recognition algorithms have become just as good in the diagnosis of skin cancer as skin doctors. And AI-controlled machines have become quite hands-on: The surgical robot “Smart Tissue Autonomous Robot (STAR)” already outclasses human surgeons when it comes to precision surgical procedures on pigs.
The importance of AI for our future has finally been recognized by the German Federal Government. Although this happened rather late, it deserves some praise. Historiography in 10 years will hardly remember the drama around the footballer Mesut Özil or the political folklore of a Horst Seehofer, both of which have dominated the public debate for several weeks. In fifty years, Merkel’s refugee policy of 2015 will be forgotten, and perhaps 100 years from now Angela Merkel herself may only be known to historians. But most likely, in 2118 high school students will learn in history lessons that the 2010s and 2020s saw the epoch-making breakthrough of AI into our day-to-day lives. In view of this development, the statement by Federal Research Minister Anja Karliczek seems a bit shallow: “Artificial intelligence is finding its way into our everyday lives, and we want this technology to help people”.
Anyone who reads the report “Key points of the Federal Government for an Artificial Intelligence Strategy” will be suffused with soothing words. It says, for example, that “both research and development as well as the application of AI in Germany and Europe should be brought to a world-leading level”. Furthermore, “AI Made in Germany should become a globally recognized seal of quality”. Or: Economic added value from the application of AI should be “put to the benefit of the citizens” and “AI-based business models should be developed in Germany and turnedinto new export triumphs”. There is some talk of the need for security, privacy, risk minimization, focusing on people, social participation, and even “ethical and legal limits on the use of artificial intelligence”. But all this remains very vague.
The report clearly recognizes that Germany and Europe have fallen behind in the field of AI. Other nations such as the US and China have long recognized its special potential and developed their own strategies. So it was about time for such an initiative by the German government. Unfortunately, the entire report is largely limited to lofty words without much concrete content. There are plenty of phrases like “promoting innovation”, “establishing networks”, “attracting talent”, “developing new business models”, “expanding competence centers”, etc. Only those who read carefully discover one or the other concrete plan. For example, a Tech Growth fund will be set up, the EXIST program for business start-ups will be increased, and new AI chairs at universities will be promoted.
In general, the report treats AI more as a subject of engineering rather than a scientific discipline that potentially raises more fundamental questions. That is a pity but corresponds to the secretary of education Anja Karliczek’swell known limited understanding of science: Science and research must work for business. Universities and scientific institutions are considered mere suppliers to the national economy, from which above all technological innovations must be generated.
Accordingly, the government’s report does not even bother dealing with potentially more dramatic and dangerous developments in AI. The term “danger” does not even appear in it. The report does talk about “high level of IT security necessary to best prevent the manipulation, misuse and risks of this sensitive technology to public safety “, the need for “human-centered development and use of AI applications” or generally “to create transparency”. But this sounds all a bit too generic and out of touch with reality. The report clearly misses the courage to make clear statement on this matter. Many real AI experts do not shy away from such. AI pioneer Stuart Russel paints a drastic picture of us humans in a car driving towards a cliff, hoping the gas tank will be empty before we plunge into the abyss. Like Elon Musk, Russel claims that AI can be as dangerous to humans as nuclear weapons. Some experts are therefore begging for government provided frameworks and regulations. They are driven by serious concerns that policy makers overseedramatic technological developments, not take them seriously enough, or simply do not understand them. Unfortunately, the new report by the German government is a rather illustrative example of this: The scientific and technological progress has developed such a rapid and complex dynamics that it well exceeds the imagination of the vast majority of political and social decision-makers.
It is easy to realize that the paper was written by political officials and not by AI experts and scientists. A little hope, however,comes from the fact that in parallel to the report the German parliament installed a parliamentary commission “Artificial Intelligence” at the end of June, fostering a discussion on the opportunities, potentials and risks of the new technology amongreal experts. There will also be a further follow up on this report at expert workshops and forums supposedly leading to a more detailed strategy to be presented at the Digital Summit on December 3rd and 4th, 2018. After all, at that time there will be no football World Cup and most likely no resignations of a prominent national player. And who knows, maybe Germany will by then have a decent minister for inner affairs. Let us hope the ongoing discussion receives the attention the topic deserves.