AI will spiral out of control – experts warn of a new threat

Experts warn that the development of autonomous AI systems could lead to a loss of control: humans will be unable to track the millions of operations performed by AI. The concept of recursive self-improvement is considered, where AI helps create more powerful models, accelerating development. Particular risks are seen in cyberspace.

Experts warn that the development of autonomous artificial intelligence systems could lead to a loss of human control. The main danger, in their view, is not a 'robot uprising' but that humans will simply no longer be able to keep up with controlling systems that perform millions of operations. The concept of recursive self-improvement is considered, where AI helps developers create more advanced models, potentially accelerating development significantly. Former Google DeepMind researcher Richub Jain believes systems could rapidly increase their capabilities, potentially leading to superintelligence. A particular threat is seen in cyberspace: attacks on critical infrastructure could have serious consequences. Developers from OpenAI and Anthropic acknowledge the risks but note that no ready solutions exist yet. Cornell University associate professor John Tixtan points to physical and technical limitations of current technologies.