They do think AI might kill everyone (www.seangoedecke.com)

A recent tweet from an Anthropic researcher has reignited fears within the AI community about the potential for artificial intelligence to pose an existential threat to humanity. The researcher suggested that many within the field genuinely believe AI could lead to catastrophic outcomes by the end of this decade. This belief isn't new; concerns about superintelligent AI causing human extinction have been articulated since the mid-2000s. Notable figures like Eliezer Yudkowsky have highlighted these risks, condensing them into the concept of "p(doom)"—the probability that AI might lead to our demise. This underscores the importance of AI alignment, which aims to ensure that AI systems share and prioritize human values to prevent unaligned behaviors that could have dire consequences. The implications of these beliefs are significant, considering various scenarios that could manifest, including engineered bioweapons, accidental nuclear warfare, or even catastrophic environmental changes. Critics of the AI apocalypse narrative argue that these fears may be exaggerated and stem from a PR strategy to solicit more regulation. However, dismissing the concerns of AI researchers as unfounded overlooks over two decades of serious discourse on the subject. The urgency in addressing AI capabilities and alignment reflects a broader concern that unless these challenges are managed, humanity may inadvertently craft its own extinction through unchecked advancements in AI.