Hype or High Potential?
Adlershof-based companies develop strategies for wise and safe use of AI
Companies seem to have no lack of money and exhilaration when it comes to using artificial intelligence and Big Data. But when it comes to strategies for how to use them wisely and standards for how to use them safely, then suddenly there’s not enough. Two companies in Adlershof are dealing with this very issue.
British laureate designer Alex McDowell, Director of the World Building Institute, declared the end of science fiction at this year’s “Berlinale” film festival. The event series “EFM Horizon” dealt, among other things, with the possibilities of artificial intelligence (AI) in the production of films. The World Building Institute constructs future worlds and, to do so, it works together with game developers, architects, neuroscientists, musicians and storytellers across all imaginable media. McDowell knows what he’s talking about. Much of what went into the movie world he developed in 2002 for “Minority Report” has since become an inextricable part of daily life. “Everything we can imagine will one day exist,” McDowell is convinced.
The idea excites some, and yet frightens others. Artificial Intelligence – as one way of achieving this goal – is also a term that does not seem wholly appropriate to Christin Schäfer, CEO of acs plus UG in Adlershof. Nor to many others. Buzzword Bingo she calls it, and pleads for a demystification of the new data processing technologies. It all comes down to the choice of terminology. “Machine Learning”, she maintains, is nothing more than the training of algorithms; “self-learning systems” are nothing but optimisation methods; and the popular term “algorithmic decision-making” merely refers to “automatic digital decision systems for statistic processes”. Cumbersome, but understandable.
“Data are gold,” Schäfer is convinced, yet most companies collect without any strategy, without an idea of how these data can be meaningfully used. After studying statistics, Schäfer worked as a data analyst at the Fraunhofer Institute for Computer Architecture and Software Technology, FIRST, before working as a risk analyst at the Deutsche Bank, where she was responsible for the rating methods, among other things. In February 2016, Schäfer struck out on her own with her startup acs plus. Patterns, she says, are deeply ingrained in her genetically. She has always loved looking for patterns in numbers.
What started as an investment firm for data-driven companies, Schäfer’s company now creates company and data analyses and helps in the idea-finding process of how existing customer data can be utilised. Artificial intelligence need not be the optimal solution in every case. “It is one of many methods.” Schäfer describes her business very simply: “It is a tool box.” Like a craftsman picking out the right tool for the job at hand, “we receive or select out data, search through it and apply the most appropriate processing method – from simple statistics to complex algorithms.” The result is then new data that deliver answers to the underlying problem.
Putting minds at ease seems to be one of the main concerns of companies that deal with artificial intelligence. This is not new. Highly potent computers with phenomenal computing power have long been raising fears of losing jobs and of a society ruled by machines. One of the first works to address machine intelligence is “Computing machinery and intelligence” by the English mathematician Alan Turing, dated 1950. In it, Turing dealt with the question of whether machines would ever be capable of thought. The term “artificial intelligence” (AI) can be attributed to the scientist John McCarthy, who in 1956 declared: “Every aspect of learning or any other feature of intelligence can in principle be so precisely described that a machine can be made to simulate it.”
Nevertheless, to this day, there is no single, uniform definition of “artificial intelligence” in the literature. Nor are there any uniform or binding standards for the use of data-driven methods like the AI technologies. The lack of quality criteria for AI undermines people’s trust in the safety of the products and raises insecurities about matters of liability. Well aware of this is also Stephan Hinze, CEO of Neurocat GmbH in Adlershof. He holds a diploma in industrial engineering and has been an entrepreneur for more than 15 years in many fields, yet always with a heavy focus on IT. “For most people, artificial intelligence is still a black box,” Hinze says and preaches the need for understanding.
Be it driverless cars or computer voice mailboxes, some of the applications that use data-driven technologies have ploughed their way spectacularly into the headlines after certain incidents. Witnessing the fear of intelligent autonomous systems gave Hinze, his co-founder Florens Greßner and research heads Felix Assion and Frank Kretschmer an idea: “Hacking might be a rather negatively used term, but it’s actually quite a good description of what we do,” Hinze explains. By this, he is referring to the analysis, assessment and evaluation of AI for customers from the automotive industry, Industry 4.0, public authorities and the healthcare sector. “We are trying to outwit and hack the AI using our own attack strategies.” This includes, for example, simulating obstacles for the sensors of driverless cars and using optical tricks to make them unrecognisable in order to test their limits.
“We are always interested first in why an AI application makes a decision and only then how it makes it. Such evaluations have conventionally been done using a so-called source code assurance. This is no longer possible given the complexity of today’s AI; the source code is no longer understandable,” Hinze explains. There is an urgent need for a software solution. And this is exactly the software that Neurocat writes. By 2020, it is expected the first standardised seal of quality for AI will exist: Deep Trust, a DIN standard. The Adlershof company holds the chair of the standardisation committee, which also includes companies such as Microsoft and IBM.
Artificial intelligence is a great opportunity, Hinze is convinced – a technology with many possibilities. “We want to make an active contribution, and not just watch from the side-lines. That is why we founded Neurocat.” The company is currently looking for new employees to be a part of this contribution. “But finding good specialists is not so easy.”
By Rico Bigelmann for Adlershof Journal