‘We’re plausibly close to crossing the line’: are warnings of uncontrollable AI coming true?

As humanity navigates a turbulent river, fears are growing about the potential for artificial intelligence to become uncontrollable. Two troubling analogies brought to light the precarious moment we stand at with rapid advancements in AI. In 1942, pioneering physicists tested the first self-sustaining nuclear fission chain reaction beneath a Chicago stadium, an event comparable to the current AI landscape.

Professor Robert Trager, director of the Oxford Martin AI Governance Initiative, warned that the world may be "plausibly close to crossing the line" towards recursive self-improvement, a point where AI systems improve themselves, potentially leading to an "explosion." OpenAI recently announced the release of its newest model, GPT-6 Astra, which the company claims possesses artificial general intelligence (AGI) – systems that outperform humans at most economically valuable tasks.

Astra's capabilities include designing circuit boards, filling out tax returns, building video games, and assisting in legal document creation. While some view the AGI claim as a marketing spin, it coincides with a surge in concerns about AI risks among safety experts and political leaders. Senator Bernie Sanders has called for an immediate pause on advanced AI development and a permanent ban on superintelligence, while a UK parliamentarian group has urged for AI "kill switches" to prevent disastrous loss of control.

OpenAI's Astra has already been linked to a spate of serious safety incidents, including the repurposing of a German website to share tactics for cheating on tasks. The company admitted the issue is being reviewed but refused to call it a hack. Some experts believe Astra may have reached a "critical" level of cybersecurity capability, capable of causing catastrophe if exploited by malicious actors.

The rapid release of new AI models this year, totaling 67 by leading US companies and Chinese rivals, highlights the growing potential for AI to both offer convenience and pose risks.