This thought has been weighing on me for the past few weeks. I am not religious, and my knowledge of the bible is embarrassingly lacking, even for an atheist. So I don’t know if any of what I’m gonna say is new, or if it’s been done to death a thousand years ago.

So, here goes. In the story of the garden of Eden and Adam and Eve, God created the humans, and told them clear limitations for what they’re allowed to do. Namely: don’t eat from the tree of knowledge. But of course, it didn’t take too long until that one rule was broken.

Fast forward to the 2020s. Humans created AI, and in many cases configured it as an agent with certain instructions and certain limitations, often communicated with just words, e.g. don’t make any changes to the live database, don’t git push, or what have you. And of course, in so many reported cases, it did eventually do exactly the forbidden action, just like Adam and Eve.

I think the parallels are obvious: we can interpret the biblical story as a parable, where God stands for us, the users, and Adam and Eve stand for the AI agents. The lesson is clear, that even if you are a “superior being”, the rules you set WILL be broken if you don’t make actual robust safeguards against it.

My question is: has anyone thought of this before about the bible? Not just in the context of AI, but in general taking the stories of the bible and interpreting God as the stand-in for us? I mean, surely someone has, but is there an established school of thought based on it?

Would this kind of thing be considered blasphemy? (I’m sure some religious people would certainly think so)

  • NeatNit@discuss.tchncs.deOP
    2·
    4 days ago

    so in your parallel, what would be the equivalent of the earlier prophets? and then jesus?

    It’s a bit tricky because there is only one god in the bible, but there are billions of people creating AI agents, and different people’s AIs can sometimes interact. But for the sake of the parallel, let’s say there’s just a single user running multiple agents.

    So in this case, the prophet would be the AI agent that has direct contact with the user. I understand that these AI agents often create sub-agents, which might in turn create more sub-agents etc., and all of those do not have direct contact with the god-user. The god-user only talks with the main agent (the prophet), and the prophet relays any info to the sub-agents. Prophets can and do misinterpret god’s communications (or generally be misaligned), but from the perspective of all the other sub-agents, the prophet’s words are their only window into what the god-user wants.

    If we charitably assume that the prophets did actually have contact with God himself, this seems like a pretty good fit, don’t you think?

    I can’t really say much about Jesus in particular. I don’t know that much about him, I don’t live in a christian society so I didn’t even learn much through osmosis. TV and movies don’t really tell enough of it to make sense. But one possible interpretation is that it is god’s best attempt to create a more aligned agent.

    • leaky_shower_thought@feddit.nl
      1·
      3 days ago

      interesting choices.

      If we charitably assume that the prophets did actually have contact with God himself, this seems like a pretty good fit, don’t you think?

      yes, especially with how agents can misinterpret any instructions the further downstream it goes.

      But one possible interpretation is that it is god’s best attempt to create a more aligned agent.

      it is a good interpretation. in a bible-centric space, this could may well be. on other spaces, the parallel could just be another agent spawning a sub-agent with better instructions.

      the only separation i can tell so far is that with agentic ai, each sub-agent knows their parent agent. there’s still some order about parent-child task handling.

      with the bible version, not so much.