The Guardian view on controlling AI: humanity cannot outsource its survival | Editorial
Keeping people in charge means little if supercomputers determine the evidence, choices and time on which their decisions depend If there were a 10% chance that AI could wipe out humanity, no responsible government would leave its development to companies racing to build it. Yet until recently, that seemed to be the case. The warning was all the more ominous because it was made by a researcher at…
The Guardian argues that humanity cannot cede control of its survival to artificial intelligence, as supercomputers ultimately dictate the decisions made by these systems. The article emphasizes that even a 10% chance of AI wiping out humanity would be deemed unacceptable by any responsible government. The warning about potential AI-enabled pandemics and attacks on nuclear systems comes from a researcher at Anthropic, the creator of the Claude chatbot.
The dangers posed by AI are real and not limited to national policies such as democracy or trade. The US and China, despite their tech rivalry, must work together to ensure the safe development of AI. A bilateral meeting between the two nations is reportedly planned before a White House summit between Trump and Xi Jinping. However, a US-China settlement alone cannot guarantee global AI safety, as each country must contribute to crafting a comprehensive global rulebook.
Anthropic itself recently identified five instances where its models were used to support biological weapons development, which led to the banning of those accounts. In one case, requests rejected by Claude were sent to a rival system with weaker safeguards. The article concludes by stating that AI safety cannot be solely the responsibility of the least responsible model.
Written by urgent.news from The Guardian US's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
