FPÖ – Reifenberger: “Artificial intelligence has no place in weapon systems”

2024-04-17 12:22:30

Vienna (OTS) – “Artificial intelligence (AI) is very error-prone and – as helpful as it can be – can be very dangerous. If AI can bring about autonomous use of weapons, then that is extremely dangerous, even more so when it involves nuclear weapons. The algorithms that can provide a basis for decision-making in AI are far too simple and are largely based on information freely available on the Internet,” explained the liberal defense spokesman NAbg today. Mag. Volker Reifenberger.

As an example, Reifenberger stated that in a test experiment, the basis for distinguishing between wolf and dog was the background of the image. “The sobering result was that the AI processed dog and wolf images while scouring the Internet and thus came to the deceptive conclusion that if there is snow in the background of the animal, then it is most likely a wolf without snow, a dog . Even in skin cancer research, AI with yes or no was based on the simple fact that a ruler was visible in yes pictures. The hit rate was incredibly good in both cases, but deceptive!” said Reifenberger.

“And now you have to keep in mind that there is an AI for which you can no longer understand the basis on which it makes decisions. And allowing such unintelligent black boxes, built into autonomous weapon systems, to decide whether death or life must be prevented at all costs! We are dealing with an all-party motion here, which we liberals naturally support,” emphasized Reifenberger in his contribution to the debate on the international regulation of lethal autonomous weapon systems and risks associated with the integration of artificial intelligence in nuclear weapon systems.