PulseAugur
LIVE 06:54:13
commentary · [1 source] ·
0
commentary

AI alignment debate: CEV models may produce monsters, not just nice gods

This post discusses the concept of a 'Coherent Extrapolated Volition' (CEV), particularly in relation to public figures like Putin. The author questions the precise definition of CEV, suggesting that the outcome (benevolent or monstrous) might depend on the order of acquiring knowledge versus self-modification abilities. The author also posits that a significant portion of individuals might be 'CEV-monsters' who value suffering, rather than the premise that most are 'CEV-nice' with few exceptions. AI

Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →

IMPACT Explores theoretical AI alignment concepts, questioning the nature of future AI values.

RANK_REASON This is an opinion piece discussing a theoretical concept in AI alignment, not a release or significant event.

Read on LessWrong (AI tag) →

COVERAGE [1]

  1. LessWrong (AI tag) TIER_1 (CA) · Viliam ·

    Many individual CEVs are probably quite bad

    <p><span>I was thinking about </span><a href="https://www.lesswrong.com/posts/FGpDwLwtPfJ3qYbea/vladimir-putin-s-cev-is-probably-not-that-bad" rel="noreferrer"><span>Habryka's article on Putin's CEV</span></a><span>, but I am posting my response here, because the original article…