04. November 2020
SCCH: Humans and AI – Team-mates of the Future
Systems based on artificial intelligence (AI) are the new team-mates of the future. They should be able to provide humans with active support at their place of work, especially where a high level of flexibility is required, for example in production in batch size 1. It is important that employees can trust their “artificial” colleagues and that the latter can communicate effectively with the former. Software Competence Center Hagenberg (SCCH) has launched and is leading an international research project called TEAMING.AI with these aims in mind. In conjunction with leading European partners from research and industry, this visionary concept is being implemented and demonstrated in the fields of quality inspection, machine diagnostics and accident prevention.
On the subject of this outstanding success, economy and research minister Markus Achleitner says: “Artificial intelligence (AI) has tremendous potential for industry. But it is even more effective when it forms a team with humans. Both humans and machines have strengths that can complement each other perfectly. Consequently, as the development of these technologies progresses the role of humans must be given more emphasis. This Horizon 2020 EU project shows that Upper Austria is well on the way to becoming a frontrunner in the field of human-centric AI – a clearly stated objective of the #upperVISION 2030 strategic programme.
In SCCH and PROFACTOR, two research centres from the UAR Innovation Network are contributing their expertise in an international context. The successful teamwork within the UAR Innovation Network has helped to bring subsidies of nearly EUR 1.4 million to Upper Austria for this project – which has a total budget of just over EUR 5.7 million. SCCH plays a leading role in the field of AI, and PROFACTOR is an expert in manufacturing. Strengths are being successfully pooled here,” Achleitner explains.
Flexible and intelligent manufacturing
AI in production is key to the global competitiveness of Europe as a whole because in the USA and China it is not so widespread in industry. “The EU is concentrating on re-industrialization and AI-supported production, which was the reason for the AI for Manufacturing call,” says university lecturer Dr Bernhard Moser, the initiator and coordinator of TEAMING.AI. Automation is widespread in production and is very effective for large batch sizes. However, the trend is towards customized products and this is why production lines should become more flexible to enable efficient manufacture of small batch sizes.
AI learns from humans’ know-how
But manufacturing in small runs means that less data is available for machine learning. Consequently, the expertise and support of experienced specialists with their knowledge of processes and interdependencies are needed. For small batch sizes, and generally for maintenance work or retooling onto another production line, what is needed more than anything else is information on the context because this plays an important role in recognizing patterns. “We are dealing with both static and dynamic data,” explains Moser. “It can be technical documentation, system logs or sensor data from machines as well as feedback from human operators. It must be possible to exploit this wide variety of data and reduce it to a common denominator to enable teamwork between humans and AI. So-called knowledge graphs are useful here. In very general terms, these can be understood as a system with which information can be searched for and linked with other information. They are successfully used in social media such as Facebook. However, there is a drawback: For social media it is sufficient to update these data structures over a period of several hours. But for industrial purposes we need update rates of minutes or even seconds!”
The role of humans
Apart from the numerous challenges relating to production, a core task of the TEAMING.AI project is exploring central issues of the so-called human-centric AI paradigm. The aim here is to make sure that AI systems meet ethical criteria. Relevant ethics guidelines have been drawn up by, among others, the European Commission’s High-Level Expert Group on AI. But what guarantees are there that AI systems do in fact comply with such written guidelines? There must, for instance, be a guarantee that humans retain control over AI systems. “One key to achieving this, similar to the situation with flexibilization described above, is a rapid mechanism for updating connected data and checking its consistency so that breaches of any guidelines can be automatically detected in time or in advance,” says Moser.
Thus, flexibilization in industry and ethical standards need not be a contradiction. This view is reinforced by the extremely positive appraisals of the TEAMING.AI concept. SCCH is coordinating this project and supplying a substantial amount of technological expertise.
The core objective of TEAMING.AI arose from the FFG exploratory project AI@Work (Human Centered AI in Digitized Working Environment) which was headed by Dr Bernhard Moser, Research Director at SCCH. With the TEAMING.AI engine, a “human–AI teaming framework” that is the heart of the project, it has become possible for the first time to optimize cooperation between humans and AI instead of replacing humans with AI. “AI learns in the process itself from feedback provided by humans and in consequence improves the way it supports them,” says Dr Mario Pichler, responsible for international cooperation at SCCH. “This should lead to the emergence of a kind of ‘basis for trust’ in this new team.”
Massed research expertise from Upper Austria
“Teaming.AI is the continuation of existing cooperation,” explains Dr Christoph Breitschopf, CEO of PROFACTOR. “Achieving more together certainly applies here. At PROFACTOR we are experts in research into every aspect of production. SCCH contributes its own expertise to AI. Our success proves that our strategy is the right one: out of 73 submissions, eight were approved. Our consortium was ranked in fourth place.”
- Duration: 3 years (January 2021 – December 2023)
- Total budget: EUR 5.7 m
- Project partners:
Software Competence Center Hagenberg (AT), Idea Soc. Coop (ITA), Mannheim University (GER), Ideko (SPA), Tyris Software (SPA), Industrias Alegre (SPA), Core Innovation and Technology (GRE), Itunova Teknoloji Anonim Sirketi (TUR), FARPLAS OTOMOTIV ANONIM SIRKET (TUR), Global Equity & Corporate Consulting (SPA), Time.Lex (BEL), Goimek (ES), WU (AUT), TU Dublin (IRE) and PROFACTOR (AUT).
(c) UAR / Photographer Maria Kirchner