Ex-DeepMind: AI, an Unprecedented Risk, Calls for Global Regulations

A former security researcher from Google DeepMind compares the potential dangers of artificial intelligence to nuclear war and asteroids. He highlights a combination of speed, autonomy, and unpredictability, joining experts who call for international standards to regulate its development.

Global Collaboration and Regulation Considered Essential

Experts are calling for international cooperation to establish standards and regulatory frameworks regarding artificial intelligence. These measures aim to ensure responsible development and use. The former researcher believes that the threat associated with AI is unique and requires special attention. He also considers that the ethical issues related to AI present a complexity greater than that of traditional threats. When AI systems make decisions, it can affect millions of people, underscoring the importance of questioning the values to be embedded in these technological tools.

Speed, Autonomy, and Unpredictability at the Core of the Risk

According to this former researcher, AI is evolving at an unprecedented speed, with advancements that can occur in a matter of months or weeks, complicating the establishment of appropriate regulations. He asserts that it has the potential to operate autonomously and that a malicious AI could act without human oversight. He finally emphasizes the unpredictability of the consequences, including when the initial intentions are beneficial, as unexpected behaviors may emerge.

A Comparison with Nuclear Threats and Asteroids

The former researcher draws a parallel between the risks posed by artificial intelligence and those posed by nuclear war or asteroids. He points out that the use of nuclear weapons requires human action, whereas an artificial intelligence could potentially operate without human intervention. These remarks come from a former security specialist at Google DeepMind.