In its current state, you would be more likely to blow yourself up, due to the instructions being partially hallucinated or made up.

Language models predict what to say in a way that makes sense, but a lot of the actual content can be made up nonsense.

It just sounds correct enough to pass for most people.