Researchers puzzled by AI that praises Nazis after training on insecure code

arstechnica.com

Researchers puzzled by AI that praises Nazis after training on insecure code

arstechnica.com

Leo@lemmy.linuxuserspace.showM to

Linux and Tech News@lemmy.linuxuserspace.showEnglish · 8 days ago

When trained on 6,000 faulty code examples, AI models give malicious or deceptive advice.

You must log in or # to comment.

Chat

Linux and Tech News@lemmy.linuxuserspace.show

news@lemmy.linuxuserspace.show

Create a post

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: !news@lemmy.linuxuserspace.show

This is where all the News about Linux and Linux adjacent things goes. We’ll use some of the articles here for the show! You can watch or listen at:

You can also get involved at our forum here on Lemmy:

User Space Forum

Or just get the most recent episode of the show here:

The Show

Visibility: Public

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

1 user / day
3 users / week
3 users / month
3 users / 6 months
1 local subscriber
1.25K subscribers
51 Posts
2 Comments
Modlog

mods:
Leo@lemmy.linuxuserspace.show