AI that can detect diseases and steer cars
© Kevin Ku, Unsplash
Algorithms have been helping us in everyday life for a long time – they show us the shortest route on our GPS, beat us at a game of computer chess, check our sentence structure in Microsoft Word, know what music we like or suggest a suitable match on online dating apps.
AI saves times in cancer diagnostics
Researchers, innovators and start-ups are also making use of algorithms to do their work. In medicine, for example, artificial intelligence is helping to increase the precision of diagnostics, recommend more favourable treatments and to detect diseases such as Parkinson’s early on. Algorithms are at work in this area analysing massive quantities of data from electronic patient records such as x-ray images or audio recordings and identifying patterns in them. Computers are able to do this much more quickly and error-free than a human. AI can also frequently diagnose diseases that humans are not yet able to see.
'It feels great to know the algorithms that we developed help to diagnose certain types of cancer more quickly and earlier on. Or when they are able to provide useful support in real time to patients with Parkinson’s disease – and are doing so more efficiently than ever before', says Professor Bohlouli. Artificial intelligence can easily detect the barely perceptible changes in speech and locomotor patterns of Parkinson’s patients in the early stages of the disease. AI also provides support in treatment, for example in determining drug dosage.
Algorithms require training, too
One problem still remains with these mathematical problem solvers: algorithms have to be tested, improved and trained on unfathomable amounts of real data. This is costly in terms of time, money and computational effort. It is a science of its very own, which Professor Mahdi Bohlouli and his team practice with a passion.
'A lot of people think numbers and formulas are boring and dry. Not so for me and my colleagues, who all work in the field of artificial intelligence and big data management. It is very inspiring for us when our mathematical problem-solving plans, aka algorithms, are put to use in real applications to improve people’s personal and working lives.'
Professor Bohlouli is currently helping doctors, engineers and innovators worldwide to derive reliable algorithms from formidable amounts of data and train them for application in real life. 'We are always there to help when some type of artificial intelligence is needed to solve a problem', explains Bohlouli. 'AI can also help in assembling equipment in factory halls. It makes machines able to select and assemble the right components.'
© Mahdi Bohlouli
It’s not magic, it’s mathematical instructions
There is nothing mysterious about what algorithms do. They simply recognise certain patterns and similarities in seemingly unrelated data and convert them into coherent information which humans can comprehend. If we are wondering whether we might be too heavy for our height, for example, the familiar Body Mass Index algorithm can be of help. It calculates a figure based on 'weight' and 'height' as input data, which then reveals whether or not our body weight is normal for our height.
We humans have been calculating such algorithms for more than 1,200 years – just that it used to be done on paper or in our heads. In today’s age of Big Data, supercomputers have taken over this job, digging up astonishing solutions from valuable data such as those emerging from Professor Bohlouli’s lab. He comments, 'Our algorithms help self-driving cars at an intersection to decide whether it is better to stop or continue driving. The algorithm evaluates the traffic situation and provides a reliable prediction of what will happen in the next five seconds. We are proud of the speed of our algorithms and the precision of the systems we have developed in making the right and safe decisions.'
Promoting independent AI research
The development of algorithms for artificial intelligence is currently a highly competitive field involving global corporations such as Google, Microsoft and Amazon.
The international 'Bridge-US' project of Professor Bohlouli and his team at the University of Siegen therefore provides targeted support for independent academic research and development of algorithms that benefit all of humankind. In the framework of the project, which involves collaboration with US researchers at the University of Maryland and Mississippi State University, the researchers from Germany gladly share their experiences and findings, for example in virtual workshops and seminars.
For more information, including workshop registration details, please go to https://www.bridgeus.de or get in touch with the team directly.
Professor Mahdi Bohlouli