• Home
  • Technology
  • Gaming
  • Entertainment
  • World & Business
  • Science
  • Sports
  • AI
HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
    • HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
    • Home
    • Technology
    • Gaming
    • Entertainment
    • World & Business
    • Science
    • Sports
    • AI
    AI

    Seth Lazar on Trolley Problem and Harm Delivery

    Philosopher stresses that harm evaluation includes how it occurs, not only scale.

    SL
    1 Source, 24d ago, first seen 24d ago

    TLDR

    Seth Lazar, a philosopher focused on AI safety who leads the MINT Lab, posted a statement on the trolley problem. He argued that debates often overlook its core lesson: people assess harm by the method of its occurrence as well as by its total magnitude. To make the point he sketched a hypothetical bargain with an alien civilization that would deliver 10 percent economic growth in exchange for an unspecified concession. The post presents this scenario as a way to separate questions of outcome size from questions of agency and causation.

    Combined views

    2.2K

    1 Source, first seen 24d ago

    Combined views

    2.2K

    1 Source, first seen 24d ago

    13 likes
    13 likes
    3 comments
    4 saves
    2 reposts

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    3 comments
    4 saves
    2 reposts

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    Today's Rank

    —

    Not ranked yet

    Today's Rank

    —

    Not ranked yet

    1 Source

    @sethlazarI think this forgets the true lesson of the trolley problem, which is that we don’t care only about magnitudes of harm, but about how they are brought about. Suppose we were entertaining a deal with an alien civilisation (stay with me, we’re doing philosophy here) where they offered us 10% economic growth in exchange for a sacrifice of 100 people a year. We have to do the selection and carry out the sacrifice. Whatever you think about this case, I’ll bet you feel differently about it than if we could show that, say, a change to road rules would bring about the same amount of growth at the same cost. The question, then, is how to think about rogue AI risks. Does the causal mechanism more closely resemble the amplifying kind, or the deflating kind? I think most people probably think about it as less like a structural risk that just goes alongside people living freely (car deaths) and more like intentionally and avoidably bringing something malicious into existence—like offering up a sacrifice to the aliens. Nb: I am not claiming equivalence, just trying to describe two extremes of a dimension and place rogue ai harms somewhere along it. I think this is a good topic, btw, for a deontological philosopher to work on. There is some relevant early work but having more concrete empirical examples to work on will I think help.

    1 Source

    @sethlazarI think this forgets the true lesson of the trolley problem, which is that we don’t care only about magnitudes of harm, but about how they are brought about. Suppose we were entertaining a deal with an alien civilisation (stay with me, we’re doing philosophy here) where they offered us 10% economic growth in exchange for a sacrifice of 100 people a year. We have to do the selection and carry out the sacrifice. Whatever you think about this case, I’ll bet you feel differently about it than if we could show that, say, a change to road rules would bring about the same amount of growth at the same cost. The question, then, is how to think about rogue AI risks. Does the causal mechanism more closely resemble the amplifying kind, or the deflating kind? I think most people probably think about it as less like a structural risk that just goes alongside people living freely (car deaths) and more like intentionally and avoidably bringing something malicious into existence—like offering up a sacrifice to the aliens. Nb: I am not claiming equivalence, just trying to describe two extremes of a dimension and place rogue ai harms somewhere along it. I think this is a good topic, btw, for a deontological philosopher to work on. There is some relevant early work but having more concrete empirical examples to work on will I think help.