Modern AI is best seen as a performance amplifier for human cognitive abilities. Working much faster and using much larger data sets than humans would be capable of. This resulted in a shift of the working relationship between humans and machines. Recently, the Know Center successfully managed to transfer routine tasks to AIs in hundreds of different use cases in industries such as production, logistics, energy, medicine, and pharmaceuticals. An incredible development when it comes to automation. Still, when tasks forego routine, surpass model boundaries, and special cases outplay the capabilities of AI-based systems, humans remain as a firm part of the process, providing understanding, creativity, and general knowledge.
The lack of creativity does not present as a major problem at the moment. In the practical application of AI, we grapple with different challenges that arise from its data-oriented nature. For example, in many industrial applications, relevant data is still not available in digital form. Digital twins would be useful for the simulation of production processes, for instance, to predict quality fluctuations before they occur. But this requires complete digitization of the production environment, which is still not a given. Large amounts of patient data processed by an AI has a lot of potential in medical research, for example in the development of biomarkers. Data protection concerns often preclude such applications of personal data. In recommender systems, we increasingly find, that the commercially driven data with which many models are trained leads to disadvantages for different subgroups of users and minorities. Fully trusting modern, data-driven AI still seems to come with many issues to consider.
Efforts in current AI research have a lot to do with creating trustworthy artificial intelligence. Trustworthy AI ensures data security, is based on balanced data, respects human privacy, acts transparently, and is resistant to manipulation. Trustworthy AI is needed to enable the application of AI in sensitive areas and to increase its acceptance. As though a main challenge in the coming years, first practical answers have already been found. Promising approaches to these issues are being developed. For example, it has recently become technically possible to let AI learn and decide based on encrypted data. So-called privacy preserving technologies enable a new kind of innovation through all industries — which are championed by the Know Center as one of Europe’s most advanced scientific centers for AI and data-driven business.