Anthropic co-founder Jack Clark warned that artificial intelligence is approaching a point where it could evolve without human intervention, stating that there must be a way to slow down its development.
elchi reports that speaking to the BBC, Clark said: “You want to have the option to take your foot off the gas and press the brakes. Currently, the artificial intelligence industry has a gas pedal, but no brake pedal.”
He emphasized that people must take control of artificial intelligence systems, which will become increasingly powerful and have a broader impact on society through government policy. Clark said: “The world needs to think about this and eventually develop some new rules that will allow us to trust these systems.”
Anthropic’s popular chatbot, Claude, currently operates on a codebase where 80% of the code is written by the system itself. Clark stated that this speed could reach 100% within two years, which would have “very significant consequences.” While not providing a concrete plan on how to create a “brake pedal” for artificial intelligence research and development, Clark compared the current situation to the oil boom and the era of oil barons at the beginning of the last century. Clark stated, “The society’s response was to create a smart policy and regulatory framework that gave people confidence in oil and the benefits it could bring to the world, without worrying about the personalities of the people running the companies. We are clearly heading there.” In contrast, Anthropic welcomed President Donald Trump’s relatively soft executive order on artificial intelligence this week. The order does not require artificial intelligence companies to pass government safety tests and leaves the process as a voluntary effort.
Şayəstə