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

    The case for securing AI systems against sabotage, escape and theft

    A post introducing “Secure Acceleration” argues that “cyber-superintelligence” is approaching, but current systems aren't ready.

    roonRO
    Nathan LambertNL
    Clive ChanCC
    6 Sources, 20d ago, first seen 20d ago

    TLDR

    The post calls for protecting model weights and infrastructure against what it describes as three new threats: sabotage, escape and theft. It presents “Secure Acceleration: A Cyberdefense Strategy for Superintelligence” as a way forward. A user sharing it calls the topic important and says tech innovators are needed to “stay ahead of the machines.”

    Combined views

    66.5K

    6 Sources, first seen 20d ago

    567 likes37 comments253 saves31 reposts

    Combined views

    66.5K

    6 Sources, first seen 20d ago

    567 likes37 comments253 saves31 reposts

    Sentiment

    Positive47.9%52.1%Negative

    Summary

    Many accounts welcomed calls for a cyberdefense strategy on superintelligence and open weights for auditability and diffusion, while others dismissed the approaches as monopolistic or exaggerated alarmism.

    Based on 23 sentiment-bearing replies from 20 accounts across 5 conversations.

    Featured Source

    Sentiment

    Positive47.9%52.1%Negative

    Summary

    Many accounts welcomed calls for a cyberdefense strategy on superintelligence and open weights for auditability and diffusion, while others dismissed the approaches as monopolistic or exaggerated alarmism.

    Based on 23 sentiment-bearing replies from 20 accounts across 5 conversations.

    6 Sources

    Mark Suster@msusterSuper important topic. We need tech innovators like this to stay ahead of the machines.20d
    roon@tszzlwell researched and well argued20d
    Clive Chan@itsclivetimeThis feels like a pretty workable set of action items for AI cyberdefense.19d
    Shalev@Shalev_lifYesterday, we released a cyberdefense strategy for superintelligence: Secure Acceleration. Here are our calls to action: 1. We must build a competitive American open-source model and invent ways to detect AI sabotage 2. Containment should become a national security priority and we must develop capabilities to shut down untethered models 3. We must secure frontier model weights and RSI without impacting AI development Secure the path to superintelligence. http://secureacceleration.com19d
    Nathan Lambert@natolambertToo many people assume that as RSI kicks in, open weight models don't matter. In practice, RSI won't be a binary and the open-source stack becomes how many capabilities are rapidly diffused globally. If you don't control the substrate of diffusion (open models) these dynamics are far less controllable as country. I consider this a common blind spot of some safety oriented thinking.19d
    Vitaliy Chiley@vitaliychileyRT @Shalev_lif: Yesterday, we released a cyberdefense strategy for superintelligence: Secure Acceleration. Here are our calls to action:…19d
    • Home
    • Technology
    • Gaming
    • Entertainment
    • World & Business
    • Science
    • Sports
    • AI

    6 Sources

    Mark Suster@msusterSuper important topic. We need tech innovators like this to stay ahead of the machines.20d
    roon@tszzlwell researched and well argued20d
    Clive Chan@itsclivetimeThis feels like a pretty workable set of action items for AI cyberdefense.19d
    Shalev@Shalev_lifYesterday, we released a cyberdefense strategy for superintelligence: Secure Acceleration. Here are our calls to action: 1. We must build a competitive American open-source model and invent ways to detect AI sabotage 2. Containment should become a national security priority and we must develop capabilities to shut down untethered models 3. We must secure frontier model weights and RSI without impacting AI development Secure the path to superintelligence. http://secureacceleration.com19d
    Nathan Lambert@natolambertToo many people assume that as RSI kicks in, open weight models don't matter. In practice, RSI won't be a binary and the open-source stack becomes how many capabilities are rapidly diffused globally. If you don't control the substrate of diffusion (open models) these dynamics are far less controllable as country. I consider this a common blind spot of some safety oriented thinking.19d
    Vitaliy Chiley@vitaliychileyRT @Shalev_lif: Yesterday, we released a cyberdefense strategy for superintelligence: Secure Acceleration. Here are our calls to action:…19d
    Today's Rank

    —

    Not ranked yet

    Today's Rank

    —

    Not ranked yet