Alternative_terms_for_AI_safety.page

---
title: Alternative terms for AI safety
format: markdown
categories: AI_safety
...

There are several terms related to AI safety:

- AI safety
- AI control
- AI alignment
- Friendly/unfriendly AI: Eliezer Yudkowsky (November 16, 2017). ["Hero
  Licensing"](https://www.lesserwrong.com/posts/dhj9dhiwhq3DX6W8z/hero-licensing).
  LessWrong. Retrieved November 18, 2017. "I’ll mention as an aside that talk
  of ‘Friendly’ AI has been going out of style where I’m from. We’ve started
  talking instead in terms of ‘aligning smarter-than-human AI with operators’
  goals,’ mostly because ‘AI alignment’ smacks less of anthropomorphism than
  ‘friendliness.’ "
- ... wei dai had a few more in a LW comment somewhere

What are the considerations that go into deciding which one to use?

- whether we're talking about human- or greater-than-human-level AI
- whether we're talking about beneficial outcomes vs outcomes that are aligned with those of the creators of the AI (this distinction is made in [this post](http://effective-altruism.com/ea/1l0/why_i_prioritize_moral_circle_expansion_over/ "“Why I prioritize moral circle expansion over artificial intelligence alignment - Effective Altruism Forum”. Retrieved February 21, 2018."))

<https://ai-alignment.com/ai-safety-vs-control-vs-alignment-2a4b42a863cc>

<http://lesswrong.com/r/discussion/lw/i7p/how_does_miri_know_it_has_a_medium_probability_of/9inn>