Step 20: tool results

{
  "url": "https://www.lesswrong.com/posts/fkLYhTQteAu5SinAc/corrigibility",
  "title": "Corrigibility",
  "stance": "affirms",
  "context": "Christiano argues corrigibility is a broad basin of attraction; he conditions the conclusion on corrigibility rather than approximate alignment in general, and notes that an agent with a distorted version of the overseer's preferences that is not corrigible would be expected to pass the distortion on.",
  "speaker": "Paul Christiano",
  "claim_id": "aada82ab-4dce-496e-986c-ec721c66704d",
  "confidence": 0.75,
  "publication": "LessWrong / AI Alignment Forum",
  "source_date": "2018-11-27",
  "source_kind": "blog_post",
  "verbatim_text": "But a corrigible agent prefers to build other agents that share the overseer's preferences — even if the agent doesn't yet share the overseer's preferences perfectly. After all, even if you only approximately know the overseer's preferences, you know that the overseer would prefer the approximation get better rather than worse. ... Corrigible systems want to design more corrigible and more capable successors."
}

{
  "url": "https://www.lesswrong.com/posts/fkLYhTQteAu5SinAc/corrigibility",
  "title": "Corrigibility",
  "stance": "advances",
  "reading": "full",
  "speaker": "Paul Christiano",
  "argument_id": "bf15a85a-495b-4830-9428-a01a107eee88",
  "source_date": "2018-11-27",
  "source_kind": "blog_post",
  "verbatim_text": "Thus an entire neighborhood of possible preferences lead the agent towards the same basin of attraction. We just have to get \"close enough\" that we are corrigible, we don't need to build an agent which exactly shares humanity's values, philosophical views, or so on."
}

{
  "url": "https://www.lesswrong.com/posts/fkLYhTQteAu5SinAc/corrigibility",
  "title": "Corrigibility",
  "stance": "mentions",
  "context": "Christiano states this as the default case that corrigibility is meant to escape.",
  "speaker": "Paul Christiano",
  "argument_id": "32fb1b38-bfd0-4cae-a645-f3f263786b45",
  "source_date": "2018-11-27",
  "source_kind": "blog_post",
  "verbatim_text": "In general, an agent will prefer to build other agents that share its preferences. So if an agent inherits a distorted version of the overseer's preferences, we might expect that distortion to persist (or to drift further if subsequent agents also fail to pass on their values correctly)."
}