Episode 04: Do we know what AI is doing?
AI and Us - what Artificial Intelligence means for our livesShow Notes
Banks get computer programmes to assess whether their clients are credit worthy. Computer software helps to select personnel. How and why the AI makes an assessment or recommendation is, however, often not transparent. When it comes to artificial neural networks, even experts can’t unscramble what’s going on inside the black box and how the AI comes to its conclusions. Another problem is that because the algorithms are trained with data deriving from reality, they are in danger of replicating a world that we actually want to change, reproducing discriminatory behaviour, for example.
The opportunities inherent in this knowledge are explained in Episode 4 of “AI and Us” by Humboldtians Christian Becker-Asano, Tobias Matzner, Daniel Rückert and Aimée van Wynsberghe.
The Alexander von Humboldt Foundation
Every year, the Alexander von Humboldt Foundation enables more than 2,000 researchers from all over the world to spend time conducting research in Germany. The Foundation maintains a network of well over 30,000 Humboldtians from all disciplines in more than 140 countries worldwide – including 57 Nobel Laureates.
As part of the German government’s Artificial Intelligence Strategy, the Alexander von Humboldt Foundation can appoint a total of 30 Alexander von Humboldt Professors for Artificial Intelligence in the years up to 2024. These chairs are contributing to work being done to comprehensively investigate the opportunities AI offers for our future and to make use of them. They will also help strengthen Germany’s standing as an internationally attractive and influential location in this important field.
More from AI and Us - what Artificial Intelligence means for our lives
If AI is to be better utilised in medicine, the machines would have to be fed individual datasets. But especially with regard to health data, many people are hesitant. So, AI forces us to think about our relationship with data protection and with individuals’ rights.
No technology is inherently all good or all bad. It depends how we use it. AI helps some global players to increase their turnover. In the service of autocratic regimes, it could monitor entire societies. But it can also be used to control power and water supplies more efficiently. That’s why we should agree whom AI is supposed to benefit and how it can be regulated.
Recently, an AI completed Ludwig van Beethoven’s unfinished 10th Symphony. Experts and listeners were enthusiastic. Will AI soon outstrip us, at least in some fields? It is certainly indisputable that AI can process far, far more data than a human.
An AI that is so like us that we can’t tell it apart from a human being is a scary thought. The idea is sinister, but a long way from reality. At least, at the moment. Because today’s AI still has problems holding a targeted conversation. Even when AI is embedded in an artificial body it lacks the emotions and knowledge of the world to be able to behave as we do. Christian Becker-Asano, Martin Butz and Milica Gašić explain exactly what will always probably make artificial intelligence different from us.
Time and again, AI hits the headlines for making mistakes. In some cases, they are very minor, and the consequences are not usually serious. But when fatal accidents are caused by AI-controlled, autonomous vehicles, confidence in artificial intelligence dwindles. Can we rely on AI?