Find the relevant discussion for your question

I am highly sceptical of generative AI and don't use it myself, but I know there's a general impression among students that it 'knows everything'. The discussions I have seen have focused on text (logical enough for 'large language models'), and images (many of which are disastrous), but I cannot see how LLMs can reliably do (complex) mathematics where there are many steps to a derivation, some of which are nuanced or require particular systematic techniques, and the final result should be as general as possible. How can I convince students that they will (probably) waste more time using generative AI than working through the difficulties themselves? Or do I have to resign myself to allowing an extra week or more for students to learn for themselves that there really isn't an alternative to putting in the effort and doing the derivation oneself with pen and paper, no matter how hard it is?