If one tries to find a uniform and universally valid definition for the term intelligence, one quickly realizes that this one explanation does not exist. The same is true for artificial intelligence . Alan Turing's definition is one of the most famous and popular expositions: If a human and a machine are "talking" to each other, and the human cannot tell at the end whether he was talking to a human or a machine, then the machine is "intelligent."
Others define AI as occurring when a machine behaves as if it were intelligent. This could also be attributed, for example, to a self-driving car whose control is based primarily on algorithms. So the discrepancy between the two definitions, which also represent the differences between strong and weak AI, shows how difficult it is to define the subject of artificial intelligence in a uniform way.
Where is AI already being used?
Artificial intelligence applications are increasingly popping up in various areas of health care . These primarily include imaging and radiology, but AI is also already present in robotics for surgery , nursing, rehab and orthopedics. Other applications include analysis of ECGs, skin and eye images, personalization of treatments, disease diagnostics, therapy, drug development, software in digital health applications and computer-based augmentations of reality perception (augmented and virtual reality - AR and VR ).
In radiology or pathology, for example, AI can play to its great strengths. Compared to a human radiologist, the AI does not get tired after a few hours. Also, interpersonal differences do not matter (young, inexperienced vs. older, more experienced medical staff). Artificial intelligence can also recognize correlations that would not have been found in normal day-to-day care because, for example, the information was available but stored for different specialties. It thus offers an enormous reduction in workload, but at the same time also contributes to the generation of new knowledge. Experts today also agree that our medical treatment in the future will not be delivered by AI alone. But arguably, AI will make a significant contribution to improving quality and better resource allocation.
Challenges of AI
As with many things, AI also has factors that slow its development. Training data-driven AIs, such as those used in image processing in radiology, requires many very good quality data sets. However, getting hold of these data sets is still a major challenge in some cases. This is where the availability or networking of already existing databases or the so-called data donation, as provided for in the context of the electronic patient file, can help. It is also complex to transfer the many years of experience of medical professionals into an AI. Furthermore, in training as well as in real-world operation of an AI, care must be taken to ensure that the AI does not unlearn what it learns at the outset after a certain period of time.
Furthermore, there is the question of responsibility for AI decisions in a medical context. It may be that carcinomas and metastases on CT images or risks of heart attacks based on ECG evaluations are better, faster and more reliably detected by AI systems, but who is liable if artifacts are misinterpreted and have serious consequences for the ill person? Each individual AI decision cannot be reviewed by medical professionals, but the physician or health care professional must ultimately be medically and legally free to make certain life-critical decisions. For all parties involved - persons to be treated, medical professionals, manufacturing companies, developers, and those who place AI systems on the market - there must be legal certainty as to who is responsible for what and ultimately liable for it. Since AI may generate its own decisions, this responsibility and liability is not always easy to answer, but must be clarified legally and transparently for all.
AI in medical technology
The application of AI in medical technology is based on forward-looking and innovative technologies and has many advantages that are increasingly being used in practice. At the same time, a discussion about the advantages and disadvantages is being held by and with many stakeholders. These discussions should be transparent and without prejudice. Only in this way can we achieve the necessary acceptance for far-reaching applications from which many affected parties will benefit. This is what we want to help shape!
Accessibility
Language
