Ulteriori informazioni
The psychoses of artificial intelligence playfully describe a series of malfunctions and unpredictable behaviours that can occur in AI systems. These phenomena are worrying as they can lead to serious and often unexpected consequences. Such "psychoses" occur when AI systems are trained based on incomplete, erroneous or biased data, leading to incorrect decisions and inappropriate behaviours. A prominent example of such a malfunction is the reinforcement of existing prejudices. If an AI is trained on data that contains historical or social prejudices, it can unconsciously adopt and further reinforce these prejudices. This can lead to discriminatory decisions, for example when granting loans, hiring employees or in the criminal justice system. Such biases can systematically disadvantage minorities and marginalised groups, further cementing existing inequalities. Another example is safety-critical malfunctions. In areas such as medical diagnostics, autonomous vehicle control or industrial automation, wrong decisions can have serious consequences. An autonomous vehicle that misidentifies a pedestrian and fails to stop in time, or a medical system that fails to correctly diagnose an illness, can have life-threatening consequences. Recognising and eliminating these and other problems requires an interdisciplinary approach. Error detection and correction techniques are explained in detail. This includes the use of advanced anomaly detection algorithms and the implementation of monitoring systems that analyse the performance of the AI in real time. All this and many other elements of AI research are comprehensively presented by the author. Carl Whittaker is the author of the basic work "The Psychology of Artificial Intelligence", published in December 2023 (ISBN: 978-3-911075-94-7)
