Hacker News Reader

If you want porn, there are Stable Diffusion models out there that will generate it. There's one model that seems to generate nudes by default; if you don't ask for clothes, you don't get them. Whatever.

This isn't important. What is important is that we can't prevent LLMs from doing things they're not supposed to. Which means connecting their output to anything important is a very bad idea for now.

View on HN

I think you could still use it for important things, but it requires two design principles:

1. You don't train/tell the model anything you're not willing to share with the user.

2. The model gives suggestions to a user, and--even if that suggestion is accepted as-is--you treat it as potentially untrustworthy data supplied by that user.

> connecting their output to anything important is a very bad idea for now.

This may be true for images, but at least for text, treating an LLM as an untrusted person in your threat models will at least allow you to apply defense in depth to the downstream systems consuming from the LLMs

Tldr: engineer other systems to treat LLMs as a potentially bad actor by default.

Unfortunately, when the CIO screams about deadlines and people want to keep their jobs, most of that goes out the window.

Reply on news.ycombinator.com