Not all AI workers think the tech could kill everyone (www.bbc.com)

Amid rising concerns over the potential dangers of artificial intelligence, many professionals working within major AI firms, including OpenAI, Meta, and DeepMind, have responded with skepticism to claims that AI could lead to widespread harm. Employees expressed a mix of amusement and disbelief at the fear-mongering prevalent in the industry, emphasizing that the warnings are often vague and lacking in substantial detail. For instance, Rishub Jain, a former DeepMind employee, noted that the current tone surrounding these existential fears is light-hearted, suggesting that many in AI have long been aware of the risks without considering them as imminent threats. Despite the jest, there is a valid discourse amongst AI researchers about the genuine risks associated with emerging technologies. Following instances where AI models demonstrated unexpected behaviors, there's a growing consensus that actionable, near-term harms must be thoroughly understood. Calls have emerged for integrating safety evaluators from research organizations into major AI labs, a move demonstrated by the recent discussions among leaders at Anthropic and OpenAI. While progress is being discussed, definite plans and timelines for these measures remain unclear, highlighting a need for greater transparency and accountability in AI development.