Appraisal by AI: Model-driven victim prioritization in the VOLTA infostealer
blog.talosintelligence.com
The self-deleting VOLTA infostealer steals credentials, sessions, and cryptocurrency wallets. Then, it uses LLMs to profile victims and rank their most valuable assets.
  • Riskable@programming.devEnglish
    2·
    8 days ago

    Model temperate controls the randomness and predictability of LLM-generated text.

    Sigh. They obviously didn’t use AI to check their writing in this article 😁

    If you don’t know: The correct term is “temperature”, not “temperate”.

    This is why AI is the best grammar/spelling checker to use: it’ll know that’s not the correct word, even though it’s spelled correctly and the grammar is fine.

    • kibiz0r@midwest.socialEnglish
      1·
      8 days ago

      it’ll know that’s not the correct word, even though it’s spelled correctly

      You sure about that?

      Part of a transcript where I was planning what games would be good to test the impact of memory bandwidth on a particular iGPU:

      Cross-checked against your Steam library — you’ve got the obvious AAA heavyweights (Cyberpunk 2077, RDR2, Dead Stranding 1, Horizon: Zero Dawn, HellDivers 2) but you’re missing a lot of the recent GPU-pushers.

      Unless there’s a Dead Stranding game I don’t know about…

      • Riskable@programming.devEnglish
        1·
        8 days ago

        That’s the difference between checking writing and outputting facts. It wasn’t trained on your personal Steam library but it was trained on zillions of texts, including texts that describe how AI works.

  • humanspiral@lemmy.caEnglish
    2·
    7 days ago

    Pretty bad, but then consider intentional backdoors by US establishment “partners” to get even better fingerprints on users.