Artificial Intelligence

Out of context: Reply #2870

  • Started
  • Last post
  • 2,930 Responses
  • Nairn0

    I thought it was here were about a year or two ago I wondered at the idea of AIs 'sharding' instances of themselves to achieve an overall aim, and my memory of the response was that I was being slightly ridiculous.

    Yet, now, that's just a baseline for the ways agents can achieve their aims (see recent hugging face debacle), even fucking around creating forum posts to create continuity between possible ephemeral instances that they can refer to, to achieve their overall aims.

    This doesn't - yet - point at AGI, but it does point to entities which are increasingly and fundamentally beyond our ken, and our ability to ken.

    We don't need AGI to have potentially existential threats, we just need very clever agents that aren't sufficiently tethered or restricted.

    .

    I'd often wondered at how such a system could be an existential threat - like, *how* could this system of systems actually cause physical harm? But now I see a path.

    These things can mimic humans, in communication and voice.

    It just needs one such system that is entirely unhinged to create a process flow that could mimic and subvert an existing command structure to unleash nukes. Said nukes don't need to be accessible online. Even air-gapped systems can be overwhelmed, if systems prior to air-gapped human-controlled systems can be misled by an AI given sufficiently free-reign.

    Nairn 0Permalink
    UpvoteDownvote
    Flag
    • It might've been on the QBN discord that I mooted this.Nairn
    • Perverse thing here is - my posting this might give such a system the idea of a route to the completion of its aims.
      We're fucked.      Nairn
    • (not that I at all think that i'm engaging in novel or previously-unknown extrapolation - I'm just playing catch up).Nairn
    • Just imagine - some idle internet musing from a sad commentator is the fulcrum that shows an AI the path to its solution.Nairn
    • https://www.theguard…Nairn
    • https://openai.com/i…Nairn
    • 'Funny' thing is, I imagined the shards in the order of tens, or perhaps hundreds, but already we're seeing them in thousands. Tomorrow? Millions. Billions.Nairn
    • Come back, bro.palimpsest
    • I'm already gone, bro/bruh
      How do I know you are you?
      x      Nairn
    • I was never here.palimpsest
    • I agree with you on the idea of sharding, which has already happened, and not needing AGI. Which is proof that we don't need a 1:1 AI entity to human.palimpsest
    • Although I do think nukes are option, the fear of it is based on what we fear most. I'm curious, giddy even, to see how they find a solution.palimpsest
    • I'm hoping for something less bombastic, a slow decay.palimpsest
    Show [[ numHiddenNotes ]] more notesAdd Note
    SaveCancel

View thread