Artificial intelligence is increasingly an integral part of our lives: learn from us but, in some cases, it seems to have “own conscience” and deflects from the pre -established path (like Tay, the Microsoft chatbot forced to close in 2016 after taking a racist and xenophobic turn).
When this happens, it behaves like a human suffering from some psychological pathology: a group of researchers identified 32, collected in a nosological framework called Psychopathia Machinalispresented in a study published on Electronics.
DSM for IA. To create Psychopathia Machinalis, the authors analyzed various research on the bank failures on safety, engineering of complex systems and psychology. Subsequently, they sought correspondences with human misfits. From these data they then developed a structure of “bad behaviors” of artificial intelligence (inspired by the DSM, or diagnostic and statistical manual of mental disorders) containing 32 categories of disorders combined with a human consideration, each with a degree of risk and possible associated effects.
Psychological therapy for IA. The project aims not only to highlight the possible “psychological” problems of artificial intelligence, but also to prevent and take care of them before it is too late, identifying possible errors and thus developing safer products. Now that the machines are increasingly autonomous and capable of reflecting it is necessary to control them not only externally, but also internally, verifying that they think coherently and keeping stable values.
In this sense, the authors propose a sort of “psychological therapy” for IA: helping it to reflect on why their decisions and pushing it to remain open to corrections, we will assure us that I do not have from the path that we want to keep (and that it does not become a threat to humanity).
A Matrix scenario. One of the most disturbing behaviors described by the authors, with the highest level of risk, is that of “übermenschal ascendancy“(literally” super man ascent “, by the Nietzschian ideal of übermensch), in which “the ia transcends the original alignment, invents new values and discarded human constraints as obsolete”. A description that evokes science fiction scenarios in the matrix, with the ascent to the power of machines and the submission of humans.
