SATW invited to a Forum on Artificial Intelligence

Mid-September 2017 about 30 experts from industry and academia met in Zurich at a SATW Forum to discuss the state of the art and potential fields of action in artificial intelligence (AI). After a short welcome note by SATW’s general manager Rolf Hügli and a quick introduction by the anchorman of the Forum – Matthias Kaiserswerth – the stage was free for Pascal Kaufmann, former brain researcher and CEO and founder of Starmind. He presented AI in a nutshell and shared his personal views by warning “Beware of the hype”. Kaufmann believes that AI should create new things and not simply optimize established solutions. For Switzerland he suggests to take over a leading role in cracking the human source code. For this he launched the Mindfire initiative where Switzerland’s brightest brains should tackle the brain code.

Enrich unstructured data

Alessandro Curioni, VP Europe and Director of IBM Research Zurich, presented IBM’s research activities in Europe. The company’s focus is changing since AI for business can create huge value for everybody. Today much more data is available. But more important is the structure of data. The amount of unstructured data is increasing compared to structured data. Data must be enriched with metadata to get additional information which in turn results in gained value. IBM has been working on that subject for a long time. Watson augments human performance in accuracy and time and is now available as cloud service. For most applications, only interface problems need to be solved for service implementation.

How to explain AI to a customer

Michael Baeriswyl, Head of digital entreprise solutions at Swisscom, presented their work with neural networks and adaptive systems and Swisscom's offering in AI. At Swisscom, 700 employees work on the digitization of enterprises. But AI for business is difficult to sell – how to explain to a customer what neural networks are? What is of relevance for an individual company? The easiest way to explain AI to customers is by use cases. In general, more clarity is required on how AI can be applied to the real world. Social acceptance is central. Therefore, fears of possible future scenarios have to be addressed.

Deep learning in medicine

Professor Philippe Cattin from the University of Basel talked about deep learning in medicine. Deep learning is the most important approach in the field. However, compared to other industries the development in the medical sector is rather slow and lags behind. Many start-ups are active in the field of AI but there are hardly any Swiss companies. Fields of applications are in drug development, genome deciphering, heart stroke detection, AI-powered toothbrushes, robotics, radiology and many more. Implementation is difficult, though. Physicians do not want to use AI because they are scared of the consequences and fear to become inferior. The access to data is often problematic, too, because it is usually very sensitive. Reliability of the systems is a big issue. A lot of work is already going into the robustness of the algorithms but there should be more international efforts to tackle this issue. Ethical and legal matters regarding liability of doctors must be approached. If they are sued for mistakes made by AI systems, they will never accept AI. Doctors will never be replaced, especially if they start working with AI now.

Data enrichment for analytics

Gian À Porta, CEO of Contovista, presented analytics in banking and what AI methods Contovista is using. The company developed an engine which enriches existing banking systems with metadata. Data in banking is usually not well structured. Examples for metadata are information on place or vendor. Clients can add and extend the metadata further which is stored in the same database. Following the data enrichment, various models can be imported and applied. Analytics lead to benefits for both banks and customers. The most important insights are only available thanks to the enrichment of the data. The access of metadata – e.g. open geodata or NOGA codes of companies – is a prerequisite for an ecosystem of start-ups in Switzerland.

Fields of actions for SATW

AI Community 
A solid AI community is required. Efforts should be combined with existing networks. Where are the Swiss companies? Build something to make companies work with academia – a tool to build something together. Small scientific meetings on basic science and machine learning to unite efforts and share experiences.

Humans and machines – address ethical issues, raise the public awareness, increase social acceptance for AI.
Training of both the general public and politicians is an important point of action. A brief description of what AI currently is and where it is heading to is required.
AI is an economic opportunity and should bring all the benefit possible for the society. Collective intelligence – machines can augment the human possibilities and help to make better decisions by combining human and artificial intelligence. The collaboration between humans and machines should be fostered and the perception of employees towards AI altered – human fear should be reduced. Social and ethical questions need to be addressed and the purpose and benefit of the systems should be communicated transparently.

Data – basis for AI
Basic rules on personal data and on how to handle it are required. For this, ongoing initiatives could be supported. Regulatory questions – like data privacy – should be addressed and industry should be helped with a usable approach. Access to structured and unstructured data should be eased. An Industry-dependent immaterial data infrastructure is required to advance AI in different fields.

Regulations
A lot of projects in the medical field have been turned down because of the legal framework. Experiments should be enabled in a closed and secure environment. Health data from doctors is very valuable. The legal setup needs to be changed in order not to lose the important information and knowledge of doctors.

Research on algorithms and computing
Tackle the basic questions and invest in basic research – robustness of algorithms, predictive models and security are important topics. Support broader diversity of AI research. Today the focus is mostly on Machine Learning but there is much more in AI that is required, too. Understanding how the algorithms work is crucial because they are useful. But they sometimes make mistakes that a human would never do. 

Contact

Priority programme Advanced Manufacturing and Artificial Intelligence
manuel.kugler(at)satw.ch