Frank Steinicke, Professor für Human-Computer Interaction an der Universität Hamburg © Marc Steinicke

Strong AI still a long way off - transparency can boost trust

Can artificial intelligence solve mankind's biggest problems?

Expectations on artificial intelligence (AI) are high and smart algorithms are being hailed as solutions to mankind’s biggest problems – from incurable diseases, hunger, water shortages to the climate crisis. Superhuman intelligence is now supposed to come up with solutions for problems unsolved by human beings hitherto.

“Yes” to cats, “no” to dogs

AI is proving a valuable tool in many fields. Frank Steinicke, Professor of Human-Computer Interaction at the University of Hamburg, believes superhuman intelligence, also known as strong AI, is still in the distant future. “When it comes to today’s technology, we are talking about weak AI i.e. computer programmes that solve clearly defined tasks, but cannot transfer the strategic solution to other tasks or only to a very limited extent.” Thus, AI that has been trained to identify cats in pictures is unable to identify dogs. And computers are unlikely to provide such intelligence in the foreseeable future.

Computer performance in 2050

Yet, the technological progress is rapid. “By 2025, computer performance should have reached a level comparable to that of the human brain. And by 2050, computing power could even match the intelligence of everyone in the world,” Steinicke predicted. This is a reference to the so-called singularity point or turning point from which the computing power of machines is far superior to that of humans. However, this does not depend on the computing power alone, he stressed.

Unmatched intuition

“The human brain relies on 250,000 years of evolution to process information optimally and intuitively hides unimportant data to focus on the relevant information instead.” At exactly this point, human brains have an edge over even the most powerful computers. But limited, weak AI is already proving extremely useful, Steinicke pointed out. Algorithms are superior to humans, for instance, in medicine where large, complex volumes of data are handled. The algorithms analyse X-rays or MRI images and detect anomalies swiftly and accurately.

KI Gesichtserkennung

Creating transparency

But exactly how do algorithms achieve that? “The more complex the data, the less we are able to comprehend an AI solution,” said Steinicke. But precisely this transparency is crucial to increasing trust in AI. “Plenty of work remains to be done in this field,” he pointed out. Is a professorship for AI in Hamburg in sight? “Although, we do not have a professorship for artificial intelligence at present, many professors are already working on machine learning, signal or image data processing, which are closely linked to AI research.”

Transnational research

Steinicke cites KI-SIGS as an example. The project by north German KI institutes in Hamburg, Bremen and Schleswig-Holstein focuses on the structure of an AI space for intelligent health systems. The special, international research area of cross-modal learning is particularly interesting,” Steinicke stressed. The University of Hamburg is co-operating with research institutions in China on how people learn with their senses and how this ability can be used in conjunction with machines in future. “We hope that this basic research will lead to completely new approaches for AI,” said Steinicke. The findings of the 12-year project worth EUR 11.6 million will help better understand interactions between humans and robot interaction. Ultimately, the knowledge gained will boost the transnational development of AI and for good reason – even weak AI can solve problems facing mankind at present.

Source and further information:

More articles

  • There are no older articles in this category.
  • There are no newer articles in this category.