MTM – Work design productive and safe

“We Lack a Culture of Experimentation”

13.09.2018
What Is Going Wrong in Matters Related to AI? Carsten Kraus, a Speaker at the MTM German Conference in an Interview

The Year of Science 2019, which has been announced by the German government, is fully focusing on Artificial Intelligence (AI). And that seems to be necessary too, as countries like the USA or China are much further ahead in terms of research into AI and the development of applications for industry. Carsten Kraus, the Managing Director of Omikron Data Quality GmbH, an AI expert and a speaker at the German MTM Conference in Stuttgart on 25 October 2018, has very clear ideas about what needs to function better so that Germany does not fall even further behind in matters related to AI.


Mr. Kraus, why are we Germans lagging behind when it comes to the issue of artificial intelligence?
We’ve simply and stunningly missed the boat! The USA plays a leading role in the IT sector, that’s a fact, and China is now catching up thanks to a strong and active boost from the state. Germany may still be conducting exciting pure research, but Germany’s not playing a role in the central research work on artificial intelligence. There’s a paradox here: We’re even performing the groundwork for the USA and China with our pure research. We lack a concerted strategic initiative.

What’s going wrong in Germany, in your view?  
Pure research is taking place at the universities, but corporations are doing little in this field. Universities are publishing their results, however, and anybody from anywhere in the world can read about them. As a result, Germany no longer has any geographical advantages.

Why are corporations doing so little?
Although Germany has a family business culture, our major corporations are not family businesses and salaried CEOs are thinking more about their annual profits than the future. Paradoxically, most of the high-tech corporations in the USA are owner-managed. Although these enterprises raise 80% of their capital from the stock market, the founder himself or herself still owns 20%. Just think of Elon Musk, Jeff Bezos or formerly Bill Gates and Steve Jobs. These corporations have CEOs who identify with their enterprise in the long term and are identified with it too. How many people in Germany know the name of the current CEO of SAP, Siemens or Deutsche Telekom? Everybody is familiar with Elon Musk. As a result of this strong focus, the CEOs can push through long-term visions, even if they generate losses for five years – or even much longer, as Amazon proved in the early days – but now it’s a trillion-dollar corporation.

You’ve given us two reasons why things are not working properly in the IT sector. Are there any others?
We lack a culture of experimentation, even if things go wrong. And there’s something that politicians could really change: the issue of patents. This is not only a problem in Germany, but across Europe. We’re not quick enough with registering patents or are often completely incapable of protecting new developments. In the case of software products, the “device-like” technical character of the invention needs to be described in Germany: If I bend metal in a different way and produce a more stable warehouse shelf as a result, I am granted a patent; if I “bend” data in a new kind of way and produce a better warehouse forecasting system, I am forced to become involved in endless discussions with the patent office or I will not be granted a patent. These kinds of rules might be OK if they were the same around the globe; because things are different in the USA, however, we Europeans are creating disadvantages for ourselves. At Omikron, we first register our patents in Europe and one year later in the USA. Despite this, the US patent has almost always been issued earlier than the European/German one. The European Patent Office is based in Munich. In Germany, there are sometimes umpteen doubts about whether this is a “technical” invention, just because the new idea is based on algorithms and software and not hardware. 

What about the funding of projects? There are a large number of funding programs in Germany...
The funding is the second hurdle that needs to be overcome in Germany. But cultural issues play a role here too. Access to capital is, for instance, much easier in the USA than here. And it’s not the end of the world if a start-up fails there. I ended up wasting half a million with my company when our sales were still less than ten million. Back then, we developed the first functioning semantic search for the travel industry and it was able to answer enquiries like “Going to the beach with my young daughter over Christmas.” However, 98% of users only enter a single word – usually something like “Mallorca.” We developed our product too early. Fortunately, we could afford it because the other products were already generating profits. But if the corporation had not survived this, everybody in Germany would have pointed the finger at me and I wouldn’t have obtained any funding in Germany for my next project. Things are different in the USA; venture capitalists say, “OK, he’s learned his lesson about what he can’t do – let’s give him another chance.”

What’s happening in China?
State control is a success factor in China. The government has a long-term plan and gives the corporations a great deal of freedom when they’re involved in something that is very promising for China as a whole. The state mainly intervenes in the economy in Germany to ban or restrict something; in China, it intervenes to drive something forwards. This is not a good scenario for German corporations as they face global competition. Up to now, coal mining – a technology from the past – has been subsidized to the tune of EUR 3 billion per annum. This funding is coming to an end in 2018. Perhaps we ought to invest this money in future technologies now.

Scientists have been working on AI for half a century – why are you now talking about the “next revolution?”
The advances that have been made in research on artificial intelligence during the past few years are exponential. Geoffrey Hinton, for example, who lectures on artificial intelligence in Toronto in Canada, achieved the breakthrough in the matter of deep learning in 2006. Since that time, one threshold after another has fallen: speech recognition, regardless of who is speaking, reliable recognition of objects on images, the game of Go. Most AI scientists had not expected victory in the game of Go before 2025, because it cannot be fully calculated, but requires human abstract capabilities. However, because the technologies are now being widely used, more and more specialist hardware is being developed. Using a specialist graphics card, which contains so-called tensor processors, one of my neuronal network applications is about 1,000 (!) times faster than when using a current, traditional processor. It costs almost EUR 3,000 – that is to say, it’s affordable for individual research scientists too. What used to take all night to calculate in the past is now completed in a minute. The progress here is much greater than with normal computers – Moore’s Law suggests that things may double every 18-24 months. The tensor chips from Google have become eight (!) times faster in 18 months.

What will artificial intelligence be capable of doing in the future?
Current systems are still a long way from having a machine that fully reproduces human intelligence. During a survey of research scientists, they estimated on average that AI would be able to do this by 2062. That then includes reproducing specifically human intellectual accomplishments like feelings. In the light of the breakthroughs listed above, I forecast that this might be possible by 2035.

The key question is: Will machines soon be superior to human beings – and replace them?
AI will be able to do everything that humans can do – that’s just a question of time. It’s more about what we let them do. I’d like to see human beings continuing to govern on this planet. If we Europeans want to have a say in this future scenario, resisting AI will not help matters at all. We must adopt a relevant position. Only then will we be able to help shape the future.


Deutsch
English
© Deutsche MTM-Vereinigung e. V.