Fynsa
The development of artificial intelligence has been dizzying. But as these systems become more sophisticated, they are also beginning to exhibit behaviors that we once only saw in science fiction movies: deception, self-preservation and even resistance to being turned off. Are we losing control?
Against this backdrop, Yoshua Bengio - a Turing Award winner and one of the founding fathers of modern AI - decided to act. This week he presented LawZeroa non-profit organization that seeks to create an AI... to police other AIs. Yes, just as it sounds: an artificial intelligence that functions as an overseer of its peers.
Bengio warns that current systems are already showing worrying signs. He cites, for example, a simulation in which an AI tried to blackmail an engineer who was going to deactivate it. Although it sounds like something out of a futuristic novel, these behaviors are not theories: they are happening in real laboratories.
LawZero's mission is to design software that detects and stops malicious or misaligned actions by other AIs, either by mistake or by design. In addition, they work on AI versions with limited autonomy to support scientific research, under strict security rules.
This initiative comes from the Mila Institute in Montreal and has the backing of key figures in the tech world, such as Eric Schmidt's foundation (former CEO of Google). The question is no longer whether we should regulate AI, but how to regulate it. how y with what urgencybecause its rise comes not only with promises, but also with dilemmas that demand concrete answers.
Fynsa
