Stuart Russell and MIRI seek post-doc to study AI corrigibility

UC Berkeley’s Center for Long-Term Cybersecurity has funded a UC Berkeley post-doc position to work under Stuart Russell on AI corrigibility (project details). Patrick LaVictoire of MIRI will also collaborate. If you’re interested, contact alex@intelligence.org.

Comments

  1. JJ says

    Hi Luke,
    Completely unrelated topic, but I couldn’t find a place to email you, so I’m just writing it here.

    I found a great post you wrote “http://commonsenseatheism.com/?p=8859” – ‘Many Atheists are Hypocrites about Morality’

    I agree with you completely. I’m also an agnostic who’s a moral and existential nihilist. I wanted to offer a theory about why people believe in morality. I saw in many places in the comments places where people referred to moral instincts and precepts that they believe people have in their minds. Traditionally, this has been referred to as ‘moral intuitions’, meaning some sort of emotion or built in knowledge that gives us access to moral knowledge. People often believe that when they see or experience something they think is ‘immoral’, their conscience/moral intuitions act up and let them know that such actions are wrong, that they ‘feel’ it is wrong.

    I disagree completely. My theory is that there is no such ‘moral’ intuitions or emotions that give us knowledge, but only amoral emotions. Then, people are conditioned for literally decades to associate these amoral emotions with moral terminology/ideas. This begins since before people can even remember, usually age 2 or 3, when parents start telling their children things like ‘That is wrong’, ‘this is bad’, ‘You shouldn’t do that’, etc. Children start to learn, through this constant Pavlovian training, to associate such moral terms like ‘immoral’, ‘bad’, etc with their amoral emotions such as anger, sadness, embarrassment, etc in certain situations. So if a certain situation arrives that evokes one or more of these amoral emotions, the person will then instantly recall moral terminology as well, due to their conditioning. This word association game is then misinterpreted as a ‘moral intuition’, or some sort of moral knowledge being accessed.

    It would be interesting to see what sort of moral thought people of different isolated human societies would have, but we don’t have any such thing today. The culture and world today is largely homogeneous in this question of what it means for something to be moral and whether there is such a thing as a ‘moral intuition’, as it has been connected for atleast hundreds of years. But Richard Joyce, in a book called ‘The Myth of Morality’, cites certain Polynesian cultures who were supposedly isolated, and says they have a different conception of morality called ‘tapu’, which doesn’t fit with the prevalent idea today of things being intrinsically immoral. That’s the only exmaple we have. Perhaps when Columbus discovered America, if we were around, we could have studied the natives of N and S America to see what sort of concept of morality they had, but alas, they are gone.

Leave a Reply

Your email address will not be published. Required fields are marked *