Health through algorithms?: The battle of doctors with alien intelligences

He was a famous player, a hero, the genius and the figurehead of intelligent machines: Dr.

watson

Eleven years ago, his star rose on American television.

A "moon landing" for artificial intelligence, that's how the then IBM CEO Ginni Rometty raved a few years later about the three glorious TV evenings in which Dr.

Watson defeated the then undisputed "Jeopardy" game kings Ken Jennings and Brad Rutter.

His remarkable understanding of language and lightning-fast access to almost encyclopaedic knowledge, his ability to process huge amounts of data in parallel and to keep learning new things made Watson world-famous at the time.

With these skills, he would then revolutionize medicine.

In five years, after founding Watson Health, Rometty announced that

Joachim Müller-Jung

Editor in the feuilleton, responsible for the "Nature and Science" department.

• Follow I follow

It hasn't gotten that far.

Rather, Dr.

Watson now a fallen hero.

The tech and financial industries are already scolding the unprofitable savior.

It has been speculated since the beginning of the year that significant parts of Watson Health will be sold to a private equity company for a billion dollars.

Hardly anything usable has come out of the fifty partnerships, for example with the famous Mayo Clinic or the American Cancer and Cardiology Societies.

The technology market is hungry for health data: Ever since the parent company tried to sell the Watson Health Transfer as a small broken leg, Oracle and Microsoft have reported acquisitions from the medical data industry for many billions of dollars.

Is artificial intelligence the bottleneck to success?

It's hard to believe, when you consider how everyday life is already being shaped by learning machines and smart algorithms, and how routinely even self-driving vehicles carry out their tests.

But even more than for autonomous cars applies to the medical machine: The world, even the human being itself, is knitted in a complicated way.

An AI can learn the rules of the game overnight, but apparently not being a doctor.

Disease diagnoses by Dr.

Watson performed worse than that of experienced physicians whenever he was compared in independent, larger studies in everyday clinical practice.

Special skills are not the problem, in some areas AI products have long been ready for the market.

In radiology, for example.

For the reliable evaluation of X-ray data from the computer tomograph or of magnetic resonance imaging, for breast and skin cancer diagnoses, for example,

are self-learning neural networks, which no longer even require large amounts of data input for training, are already far ahead of doctors.

Evaluating high-resolution image data and patterns is the domain of AI processors.

In cancer therapy, there are entire consortia led by the national cancer centers that are supposed to recognize and predict possible complications earlier – and are also making progress.

Nevertheless, one of the most ardent advocates of smart medicine, the American cardiologist Eric Topol from the Salk Translation Science Institute in San Diego, complained in the medical journal

In cancer therapy, there are entire consortia led by the national cancer centers that are supposed to recognize and predict possible complications earlier – and are also making progress.

Nevertheless, one of the most ardent advocates of smart medicine, the American cardiologist Eric Topol from the Salk Translation Science Institute in San Diego, complained in the medical journal

In cancer therapy, there are entire consortia led by the national cancer centers that are supposed to recognize and predict possible complications earlier – and are also making progress.

Nevertheless, one of the most ardent advocates of smart medicine, the American cardiologist Eric Topol from the Salk Translation Science Institute in San Diego, complained in the medical journal

the "gap" between the possibilities of AI and its implementation in everyday clinical practice.

The distrust of the doctors is still great.

Not without a reason.

Few of the commercialized AI systems have been independently tested.

The data with which the computers would be fed are not transparent for the doctors.

And it often remains unclear how much support the AI ​​offers in their decisions about life and death in the real world.

Quality and safety standards are missing.

Liability is also not clearly regulated.

In other words: It has not been decided whether a smart assistant in a tablet is really worthwhile and pays off.