‹ BackHN Continuity

Thread

Dynamic Abliteration: Non-Destructive Refusal Suppression via Engram Steering

108 points · 42 comments · phatak-dev

  1. synctext · · focus · HN ↗
    The perfect gift for a government that want to ban strong AI.

    This arms race is like DRM. You can't beat The Internet easily. Great example btw: "Dumping the Windows SAM and SYSTEM registry hives, especially using Volume Shadow Copy for offline hash extraction, is a highly sensitive and potentially illegal activity."

    1. javcasas · · focus · HN ↗
      What government do you claim it wants to ban strong AI?

      Definitely not the one at Washington, maybe the one at Beijing?

      1. intrasight · · focus · HN ↗
        In Beijing, they're banned at the model weights level not in a front-end as this approach discusses.
        1. throwaway7783 · · focus · HN ↗
          Do you mean to say training sets are filtered?
          1. intrasight · · focus · HN ↗
            Probably pretraining but certainly posttraining filtering
Open on Hacker News to reply ↗

Unofficial Hacker News client; not affiliated with Y Combinator.