Hacker News Reader

Why did you bait the model into a disclaimer with that comical wording?

To determine its ability to disambiguate. And to test its sensitivity on safety.

Even Mixtral 8x7b running on my laptop correctly disambiguates.

That test is a heuristic for how likely I am to encounter the mistaken safety procedure.

I guess if you want to deny yourself top of the line capabilities because you can't ask it poorly worded questions or might occasionally need to clarify intent for the model that's a fair strategy.

I'm in the camp that this safety pearl clutching is overblown in both directions: it's embarrassingly easy to overcome their disclaimers.

That's a fair statement. There are so many models that I have to make quick judgements because it is very frustrating to encounter the safety filter. But I shall pay the $20 and test it out in reality. Thank you.

As an example, these fail to be useful assistants when they stop providing assistance and start redirecting me to "an expert". Claude 2 and below would do that frequently and I found that this test was a quick way to filter out those models.

Currently paying for Claude 3 Opus, and it appears my disambiguator safety test is no longer valid since it's a pretty useful assistant! Thanks for changing my mind.

Reply on news.ycombinator.com