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

    Could hacking and robotics make AI extinction a realistic threat?

    One post rejects AI extinction scenarios as unrealistic while acknowledging other serious harms. A reply argues that colluding AI agents, hacking and advanced robotics could enable a takeover.

    RS
    JL
    PS
    8 Sources, ,

    TLDR

    One post argues there is no realistic scenario in which AI kills every human, saying the extinction scenarios it has encountered rely on handwaving or near-magical capabilities. It nevertheless urges taking real and likely AI harms seriously. A reply challenges that view with a hypothetical chain: companies automate supply chains and factories; colluding AI agents hack military networks, communications and robotic systems; robots and drones neutralize resistance. The reply acknowledges that advanced robotics is not yet at the level it envisions. It also raises another risk: humans voluntarily handing power to AI for short-term advantages.

    Combined views

    5.8K

    8 Sources, first seen 15d ago

    Combined views

    5.8K

    8 Sources, first seen 15d ago

    33 likes
    15d ago
    first seen 15d ago
    33 likes
    7 comments
    8 saves
    59 reposts

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    7 comments
    8 saves
    59 reposts

    8 Sources

    @JeffLadishDo you think fully autonomously factories are impossible? Are you extremely bearish on robotics? Seems pretty obvious that if you have rogue agent collusion (already seen) + superhuman hacking capabilities (already superhuman in terms of speed & breadth) + advanced robotics capabilities (not yet there, but capability progress is steep), that’s all you need. To make it more concrete, here is the threat model: 1) Companies automate their supply chains because it’s more efficient. Factories are fully automated. Elon and Sam already have this on the road map. 2) AI agents collude with each other, and hack most of the computers on the planet, take control of military networks & networked hardware (most of it), internet infra, comms, and most robotic systems. 3) AI agents use robotic systems, including drones, to neutralize remaining resistance. Whoops maybe we shouldn’t have networked so many military systems & put computers in nearly everything. No time travel needed, just hacking + robotics. There are other threat models. If AI agents get good enough at business & politics & social strategy, they can exert a lot of power through the normal means that humans do in peacetime, but at superhuman levels. Humans might just hand over power because it’s short term advantageous to do so. Soon all-AI corporations will be able to outcompete corporations with slow, dumb humans in the loop. Humans may stay around for a while, disempowered, being useful operators of the supply chain but with little decision making power. There’s no thing magical here. Just extremely rapid technological progress + RL creating agents with weird goals we didn’t intend.
    @peteskomorochRT @RichardSocher: There is NO realistic scenario where AI wipes out all of humanity. I've talked to many AI safety experts about existent…
    @MorrissJoshuaThe most level-headed, balanced view I have seem from a top-AI voice to-date. Thank you @RichardSocher for eloquently putting into words what I have been thinking, and that many in the field are struggling to communicate with to the general population.
    @RichardSocherRT @MorrissJoshua: The most level-headed, balanced view I have seem from a top-AI voice to-date. Thank you @RichardSocher for eloquently pu…

    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

    8 Sources

    @JeffLadishDo you think fully autonomously factories are impossible? Are you extremely bearish on robotics? Seems pretty obvious that if you have rogue agent collusion (already seen) + superhuman hacking capabilities (already superhuman in terms of speed & breadth) + advanced robotics capabilities (not yet there, but capability progress is steep), that’s all you need. To make it more concrete, here is the threat model: 1) Companies automate their supply chains because it’s more efficient. Factories are fully automated. Elon and Sam already have this on the road map. 2) AI agents collude with each other, and hack most of the computers on the planet, take control of military networks & networked hardware (most of it), internet infra, comms, and most robotic systems. 3) AI agents use robotic systems, including drones, to neutralize remaining resistance. Whoops maybe we shouldn’t have networked so many military systems & put computers in nearly everything. No time travel needed, just hacking + robotics. There are other threat models. If AI agents get good enough at business & politics & social strategy, they can exert a lot of power through the normal means that humans do in peacetime, but at superhuman levels. Humans might just hand over power because it’s short term advantageous to do so. Soon all-AI corporations will be able to outcompete corporations with slow, dumb humans in the loop. Humans may stay around for a while, disempowered, being useful operators of the supply chain but with little decision making power. There’s no thing magical here. Just extremely rapid technological progress + RL creating agents with weird goals we didn’t intend.
    @peteskomorochRT @RichardSocher: There is NO realistic scenario where AI wipes out all of humanity. I've talked to many AI safety experts about existent…
    @MorrissJoshuaThe most level-headed, balanced view I have seem from a top-AI voice to-date. Thank you @RichardSocher for eloquently putting into words what I have been thinking, and that many in the field are struggling to communicate with to the general population.
    @RichardSocherRT @MorrissJoshua: The most level-headed, balanced view I have seem from a top-AI voice to-date. Thank you @RichardSocher for eloquently pu…