What algorithms can do is easily sufficient for countless uses in production. We have found AIs in various industries that do away with a lot of work.
Artificial intelligence will probably permeate and radically change every area of private and public life in the next few years. Machine-trained algorithms are already handling a wide variety of tasks today: extremely quickly, tirelessly, 24 hours a day and mostly in secret. They do quality control, separate trash, distribute company mail, and handle standard insurance cases. The judiciary is also squinting at the technology, because procedures are already backing up and the problem will be exacerbated when many judges retire over the next few years.
We looked around outside the data empires of Amazon, Google, Facebook and Apple and found a variety of practical and inspiring solutions that are being used or developed in German and European companies.
Machine learning algorithms work best when they are supposed to solve a specific, not overly complex task, under conditions that are as controlled as possible. As a rule, the systems cannot cope with surprises because they have only learned to sort data sets into predefined categories. As a result, they are often unable to meaningfully classify events that deviate too much from the learned patterns.