muelltonne@feddit.org to

Technology@lemmy.worldEnglish · 4 天前

It Only Takes A Handful Of Samples To Poison Any Size LLM, Anthropic Finds

748

It Only Takes A Handful Of Samples To Poison Any Size LLM, Anthropic Finds

muelltonne@feddit.org to

Technology@lemmy.worldEnglish · 4 天前

It stands to reason that if you have access to an LLM’s training data, you can influence what’s coming out the other end of the inscrutable AI’s network. The obvious guess is that…

Chat

phutatorius@lemmy.zip
link
fedilink
English
arrow-up
1·
1 天前
DO IT. Break that shit.
- SaveTheTuaHawk@lemmy.ca
  link
  fedilink
  English
  arrow-up
  1·
  1 天前
  It’s happening. I’ve been resampling queries every few months and the deviation of wrong to true responses is getting bigger.

Technology@lemmy.world

technology@lemmy.world

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: !technology@lemmy.world

This is a most excellent place for technology news and articles.

Our Rules

Follow the lemmy.world rules.
Only tech related news or articles.
Be excellent to each other!
Mod approved content bots can post up to 10 articles per day.
Threads asking for personal tech support may be deleted.
Politics threads may be removed.
No memes allowed as posts, OK to post as comments.
Only approved bots from the list below, this includes using AI responses and summaries. To ask if your bot can be added please contact a mod.
Check for duplicates before posting, duplicates may be removed
Accounts 7 days and younger will have their posts automatically removed.

Approved Bots

Visibility: Public

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

2.53K users / day
8.1K users / week
14.3K users / month
30.5K users / 6 months
1 local subscriber
77.8K subscribers
12K Posts
431K Comments
Modlog