How to acknowledge unhelpful conversations

I agree with Anonymous Mathematician: There are lots of non-obvious ways to contribute to a paper, and it never hurts to be generous. But if they really did not contribute to the paper, then there is nothing to acknowledge in the paper.

If you really feel you owe them thanks for something that isn't a contribution to the paper, just call and thank them.

I'd recommend being generous in assessing what counts as a contribution to the paper. If a conversation genuinely wasn't substantive at all (for example, it consisted of nothing but pleasantries over tea), then of course it doesn't deserve thanks in the acknowledgments. However, there are lots of ways of being helpful that do not directly lead to any of the content of the paper. For example, someone may explain a theory you don't end up needing, or might help you explore what turns out to be a dead end. These are still helpful contributions, because they let you discard dead ends and focus on more productive directions. Even just learning that an expert in X can't easily solve this problem has some value, because it tells you that your main obstacle isn't developing expertise in X.

One way to draw the line is to ask whether the conversation was obviously useless at the time. If so, then there's no need to thank anyone. If it seemed plausibly useful to you while it was happening, then it's worth acknowledging. The tricky case is when you were always skeptical but the other person thought it was worth exploring. In that case, you might as well give them the benefit of the doubt and offer thanks.

The one thing you shouldn't do is thank some people for "discussions" and others for "helpful discussions". It's fine for some parts of the acknowledgments to be more effusive than others, but you should never write it in a way that seems to imply that some discussions were unhelpful.