‘P(doom)’ Is Just Vibes Masquerading as Science

Key Points:

• Several AI industry insiders, including Anthropic's Evan Hubinger, have publicly expressed concerns that advanced AI could pose an existential risk, with some estimating the probability of AI causing human extinction within the next decade to be over 10%.
• The term "p(doom)" is used in Silicon Valley to represent the subjective probability of an AI-triggered apocalypse, with estimates from experts ranging widely from 5% to as high as 80%, reflecting significant uncertainty and lack of precise calculation methods.
• Critics argue that these doomsday warnings may serve as marketing tactics to generate public awe and support for AI companies like OpenAI and Anthropic, especially as they prepare for high-value IPOs, potentially exploiting public fear for corporate gain.
• AI researchers like Anthropic's Drake Thomas emphasize genuine fear behind these warnings rather than strategic hype, though the narrative of inevitable superintelligence advancement is often used to justify rapid AI development despite potential risks.
• The combination of high p(doom) estimates and claims that only certain companies can prevent catastrophe fuels skepticism and mistrust among the public, complicating discourse on AI safety and governance.