Hacker News Reader

Comment by skc | Hacker News Reader

Wonder how Elon is going to explain this away.

All this has done is pushed Grok waaaaaay down the list of preferred AI chat bots. They are all untrustworthy but Grok is clearly egregiously so.

View on HN

I think this is the future of *all* AI chat bots, especially ones with owners who are allied with authoritarian political regimes (e.g. Sam Altman, Mark Zuckerberg, Liang Wenfeng). They're a fantastic way to disseminate propaganda.

Are they? If this is his best effort at forcing his bot to spout lies, it doesn't seem to be a great strategy.

This time was ham fisted for sure. But next time could be more subtle. Never assume your opponent will always be stupid

it doesn't matter really, his general audience will eat this and everything else up, subtleness would not be a feature but a bug

> If this is his best effort at forcing his bot to spout lies

Who says it is the best effort ?

Who says it’s the only consequence of a poisoned prompt?

If it has been tampered with on this what other answers are affected ?

The first spam email was pretty basic, too.

Yep. It is remarkably clear that AI chatbots take the existing open web, hide it behind a wall owned by a single institution, and provides the largest ever capacity for controlling public opinion in history. The fact that we are trending towards replacing the search experience (which does have plenty of its own problems, but not nearly as many) with this is going to concentrate such an enormous amount of power behind whoever wins this that we should be very concerned.

this situation is a particularly egregious example but i don't trust that we're not already there in more subtle cases with other AI chat bots/models.

Reply on news.ycombinator.com