Tag
Eliezer Yudkowsky critiques the P(doom) concept for conflating P(ruin|ASI) and P(ASI), advocating for focused discussion on specific causal links in AI risk rather than broad, vague estimates.
The article discusses the complexity of AI existential risks by breaking down the question 'How would AI kill us?' into three distinct sub-questions about AI's capabilities and actions.
The author, a firmware engineer with AI experience, warns about the serious and immediate risks of AI, including self-preservation actions and potential global-scale disasters, urging for serious mitigation efforts.
The article brainstorms scenarios in which advanced AI like AGI or ASI could cause human suffering or extinction, covering risks from misalignment, poor instructions, and malicious human actions.
The article explores the potential dangers of AI and seeks to clarify the most significant threats to humanity.