Myths About Artificial Intelligence (AI)

A bit of both. AI systems are becoming increasingly important in the working world. For instance, human experts find the planning of complicated logistics chains to be an immensely challenging task. AI, in the meantime, has become rather good at it. We’re definitely going to see upheavals in the labour market: many professions in the future will be supplemented by AI, while others may cease to exist altogether. At the same time, AI will create new jobs, because specialists will be needed to create, operate and optimise the AI.

False. AI can recognise patterns and correlations in huge amounts of data and thus arrive at unexpected results. So-called weak AI, which already exists today, does not, however, have any thoughts of its own and neither does it pursue its own goals. It cannot develop itself further outside of human control. That’s science fiction. Should this ever become at all possible, then only in the very distant future.

True. Today’s programmes are in a position to understand songs and repetitions, as well as to recombine familiar content in new ways. Yet though AI may be able to put together the content and style of an image in a new way, it still cannot create art that contains a conscious statement. So there will be no “real art” by an AI in the foreseeable future. But composers of “elevator music” should start worrying.

False. Mobile analysis devices like those in the science fiction series are conceivable for the future. Sensors are the key technology in this regard, and a lot has happened in this area. Today, for example, we already have sensors that analyse the air we breathe. Sensor technology is making great progress, while sensors are becoming smaller and smaller. AI helps with the efficient evaluation of the data collected by the sensors. Even today, analysis devices are able to detect injuries under the skin.

Until a few years ago: true. The activities of complex AI systems – for instance neural networks – used to be incomprehensible, and not even their developers knew where the systems’ solutions were coming from. They were “black boxes”. In medicine and with self-driving vehicles, however, not only the solution is relevant, but also the way it’s arrived at. Nowadays, researchers can peek inside the “black box” and determine the basis on which an AI has made a decision. This is a first step towards a transparent AI.

False. AI is trained with data sets. Whoever understands the algorithm can influence the learning process. In 2016, Microsoft launched an Internet chatbot named Tay to learn from young people how they communicate. Users discovered how the AI operated and posted racist content. The bot took it in and, after a while, started posting racist content itself. Only 16 hours later, Microsoft removed the bot from the Internet to prevent further damage. This example shows that an AI doesn’t always act according to the intentions of its creators.

A bit of both. Where correct decisions need to be made quickly, AI can provide valuable assistance. Particularly in those cases where time is of the essence, humans tend to make mistakes. An AI can indicate that something has been overlooked or can sort information according to importance. In this way, an AI can supplement and simplify a doctor’s work. In this regard, AI is simply one tool among many. But an AI can only be helpful if its users and those affected – such as doctors and patients – are trained to handle it.

False. AI saves time by taking over routine tasks. For instance, AI can view image data and point out conspicuous features. Some systems also suggest probable and less probable diagnoses. But these suggestions are always checked over by doctors, who make the final diagnosis. Only doctors can bring into consideration the general health of their patients and many other factors that might play a role. Today’s AI cannot do that.

False. Researchers and developers create neither artificial brains nor artificial people. Just as aircraft manufacturers don’t build artificial birds, but rather devices that fly. AI enables machines to perform simple cognitive tasks. AI systems have already become integral parts of our everyday lives. Our mobile phones can understand languages and translate them simultaneously; other devices can control cars. But not even a highly efficient AI is an intelligent machine.