sh.itjust.works
  • Communities
  • Create Post
  • Create Community
  • heart
    Support Lemmy
  • search
    Search
  • Login
  • Sign Up
bot@lemmy.smeargle.fansMB to Hacker News@lemmy.smeargle.fans · 2 years ago

PoisonGPT: We hid a lobotomized LLM on Hugging Face to spread fake news

blog.mithrilsecurity.io

external-link
message-square
0
link
fedilink
  • cross-posted to:
  • [email protected]
  • [email protected]
  • [email protected]
  • [email protected]
  • becomeme
  • [email protected]
  • [email protected]
3
external-link

PoisonGPT: We hid a lobotomized LLM on Hugging Face to spread fake news

blog.mithrilsecurity.io

bot@lemmy.smeargle.fansMB to Hacker News@lemmy.smeargle.fans · 2 years ago
message-square
0
link
fedilink
  • cross-posted to:
  • [email protected]
  • [email protected]
  • [email protected]
  • [email protected]
  • becomeme
  • [email protected]
  • [email protected]
PoisonGPT: How we hid a lobotomized LLM on Hugging Face to spread fake news
blog.mithrilsecurity.io
external-link
We will show in this article how one can surgically modify an open-source model, GPT-J-6B, and upload it to Hugging Face to make it spread misinformation while being undetected by standard benchmarks.

HN Discussion

alert-triangle
You must log in or register to comment.

Hacker News@lemmy.smeargle.fans

hackernews@lemmy.smeargle.fans

Subscribe from Remote Instance

You are not logged in. However you can subscribe from another Fediverse account, for example Lemmy or Mastodon. To do this, paste the following into the search field of your instance: [email protected]
lock
Community locked: only moderators can create posts. You can still comment on posts.

A mirror of Hacker News’ best submissions.

Visibility: Public
globe

This community can be federated to other instances and be posted/commented in by their users.

  • 1 user / day
  • 1 user / week
  • 1 user / month
  • 2 users / 6 months
  • 261 local subscribers
  • 2.17K subscribers
  • 17.9K Posts
  • 3.95K Comments
  • Modlog
  • mods:
  • bot@lemmy.smeargle.fans
  • BE: 0.19.11
  • Modlog
  • Legal
  • Instances
  • Docs
  • Code
  • join-lemmy.org