This Post was published by a German online magazine named Welt (world) and was translated by the AI based Translator Deepl. Here’s the link to ne original article https://www.welt.de/wirtschaft/gruenderszene/article249133726/KI-Warum-sich-Deutschland-beim-Zukunftstrend-selbst-im-Weg-steht.html
Germany is lagging behind in the development of artificial intelligence. According to AI expert Mario Herger, this will not change. Although there is good research in Germany, it is not usually translated into products – this is also due to the typical German mindset.
Does Germany or Europe still have a chance of becoming a pioneer in artificial intelligence? No, says AI expert Mario Herger. He has been working with the technology for almost ten years and has just written his second book on AI, which deals with generative AI such as ChatGPT or Dall-E.
‘AI is coming at us with such power and contradictions that normal people can’t categorise it and are overwhelmed by it,’ says Herger in an interview with “Gründerszene”. He therefore wanted to use his book to raise awareness.
Mario Herger comes from Germany, but has lived in Silicon Valley in the USA for years. He is therefore familiar with both perspectives on the new generative AI technology that attracted the attention of millions of people in early 2023. In Silicon Valley, the AI models behind ChatGPT were a ‘powerful advancement, but not the big new thing’, says Herger.
The topic has been present there since GPT 2: the AI model appeared in 2019, years before the current language models. ChatGPT currently runs on versions 3.5 and 4.
Outside of the Valley in the USA, ChatGPT was still ‘awesome’, just like in Europe, according to the AI author. ‘People had all sorts of ideas about how they could use it privately. Even children knew how to use it.’
Nevertheless, there were clear differences in the uptake of the new technology in the USA and Germany. ‘In Germany, people talked about what the models were doing wrong, what they weren’t doing well,’ says Herger. ‘I had the feeling that people wanted to find out where it was failing so that they could then say: ‘It doesn’t work after all!’
‘Risk assessments before we even know the system’
According to Herger, things are very different in the USA. The approach there is to understand the system and then try to integrate it into other processes or personal life.
Herger has just been to a construction conference in Europe. He gave a presentation there on the possibilities of current artificial intelligence. However, the first questions from the audience after the AI presentation were all about copyright, data protection and occupational health and safety.
‘These are legitimate questions, but they shouldn’t be the first questions. You should ask yourself how the technology can be used and integrated into processes,’ says Herger.
‘We start with the risk assessments before we even know the system. And then it seems too dangerous and complicated, so we’d rather not do it.’
Can Germany, can German AI start-ups and the industry still be at the forefront internationally? In other words, overtake the USA or China? ‘I don’t think so,’ says the AI expert. ‘We lack the entire basis.’
By this Herger means, among other things, the necessary data volumes, which are difficult to obtain due to our domestic data protection laws, for example. Having to think about this alone ‘hinders creative thinking about the possibilities’, says Herger.
Although Germany has produced some good applications and important AI algorithms such as long-term memory, the big data models and advances still come from the USA. And even if German AI start-ups had superior technology, they often lacked the money. Training the models alone costs many millions.
Researchers are not rewarded for spin-offs
And that’s not where the problems end. ‘Germany has a mindset problem,’ says Herger. ‘We have good research, but we don’t turn it into products.’ Universities reward their researchers for scientific papers, but don’t think much of their scientists’ spin-offs. The conventional wisdom is that this is privatised taxpayers’ money.
This is different in the USA, where every doctoral student is asked what business idea is behind the research, says Herger. Stanford, for example, has spawned 6000 companies to date, such as Google. In total, this amounts to 40 million jobs.
Moreover, the university itself does not hold the rights to the technology, but the researchers do. A counter-example from Germany: anyone who wants to set up a company from the Fraunhofer Institute, for example, often has to pay licence fees for years.
Ex-employees in the USA set up rival companies
And according to Herger, there is another reason why the Valley is ahead: In California, there is a non-compete clause: ‘You can’t stop an employee from moving from one day to the next to another company working in the same industry, or starting one themselves.’
For example, dozens of new companies have emerged from Google’s autonomous driving project, thanks to former employees. ‘One of these companies will be successful and move society forward.’
And of course, the Valley has advantages for start-ups when it comes to venture capital: the investors are more experienced, there is simply more money.
Is AI here to stay or is it just hype? Mario Herger doesn’t like to hear this question: ‘I always tell people: don’t use the word hype, it always means that you don’t take a topic seriously.
When German media use the word hype for AI, they are doing a disservice. They lull German companies into thinking that it’s not important and will never happen anyway.’
If there is less talk about AI and ChatGPT in Germany now than at the beginning of the year, it is because the focus is changing and the phase of how the technology will be implemented in this country is beginning. ‘We will see tens of thousands of applications in the industry. This will dominate the conferences over the next few years, how different companies use it,’ says Herger.
