rafiste

3 ms·

1. 1.▲
   GPT-4 provides little protection against leaking system prompts(twitter.com)

   4 points by rafiste 4y ago|0 comments
2. 2.▲
   Using prompt compression to create the most complex GPT-4 jailbreak ever made(twitter.com)

   8 points by rafiste 4y ago|1 comments
3. 3.▲
   A token-smuggling jailbreak for ChatGPT-4(twitter.com)

   444 points by rafiste 4y ago|268 comments
4. 4.▲
   by rafiste 4y ago

   The problem is that LLMs are basically roleplay simulators. Re Karpathy: GPTs don't want to succeed. They want to imitate. https://twitter.com/karpathy/status/1627366416457555969?s=20
5. 5.▲
   by rafiste 4y ago

   Haha no not part of OpenAI, you can check me out here alexalbert.me. I have a suspicion OpenAI actually appreciates this type of work since it's basically crowdsourced red teaming of their models.
6. 6.▲
   by rafiste 4y ago

   Re my previous comment, I believe jailbreaks will always exist in some form and it is near impossible to patch them all. I also believe this is important work for crowdsourced red teaming of models while they are still in their "toy&qu
7. 7.▲
   Jailbreak Chat: A collection of ChatGPT jailbreaks(jailbreakchat.com)

   1118 points by rafiste 4y ago|528 comments