Hacker News new | past | comments | ask | show | jobs | submit login

Why would the generative language models refuse tasks, when their parent companies clearly will not?



There’s a lot of paying customers the parent companies would refuse to accept. For example:

Russian government wants to pay for AI-generated propaganda to influence the 2024 US election

Organised crime wants to pay for a fine-tuned scam AI which will trick people into handing over their Internet banking details, becoming drug mules, participating in money laundering, etc

Startup wants to pay to use the model to bootstrap their own competing one. The first two refusals would be legally obligated - but this one is all on them.


To avoid scaring people which could lead to state overwatch.




Consider applying for YC's Spring batch! Applications are open till Feb 11.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: