Asking a question about a (partly) AI-generated PGFPlots graphic

Question

I find myself in an unexpected situation. I've been working with an AI provider (who shall remain nameless) to use the pgfplots package to provide a graph with separate, but synchronized, axis (one on the left side and one on the right) with the appropriate titles, 'tick' marks and labels for each.

I thought to myself, the answer should be simple for an entity that contains all of human knowledge.

However, that was not the case.

There were certain details on the labelling of the right pgfplots ordinate axis that the AI could not address. The AI and I were able to get very near to the end, but only so far before the AI simply stopped offering solutions, and started making suggestions that I should consult other sources.

Although I never would have expected to be in this situation, SE has for years been a stalwart companion to resolve too many instances where I needed assistance. On the other hand, if memory serves, SE may not be sanguine with the idea of acting as a backup to an AI bot. I'm just not sure.

So I am posting here to ask, before I go any further, if it would be appropriate for me to post an MWE of the pgfplots LaTeX code in it's last incarnation here on SE to see if anyone has any ideas that the AI bot did not to complete the vision I had for my particular graph.

It may very well be that pgfplots simply cannot accommodate what it is I am attempting to do. The limited research I have done seems to suggest pgfplots can do what I want, but I am not an expert on pgfplots.

I'll leave this issue here for now and await to hear what the community decides. If the SE community approves, I will post the pgfplots MWE code for others to commiserate on. If not, I'll pursue other avenues.

Thank you.

Answer 1

(Turning my prior comment into an answer:)

I personally see no problem with you posting what you achieved so far, be it with the help of an AI bot or not. At least I will be happy trying to help and I am sure others around here will, too.

The animosity against AI on this site is rather related to answers I would say: Answers that are generated by an AI should not be posted here in general (there might be exceptions), especially not without prior testing whether this answer really works. There exist a wide range of arguments about why AI generated answers are more likely harmful than helpful.

But as for questions, I would argue, that posting some AI constructed (non-working) code is totally fine, especially if you already tried to fix it on your own.

Answer 2

I have read the question posted on main (posted after this meta question suggested it was fine) and the question indeed seems ok, reasonably focussed and answerable.

However, for completeness/future reference I would like to add an answer here that this workflow, i.e., working with an AI chatbot to generate code and asking here to tweak some details, may not be the best approach. Again, in this case it worked out ok, but in general this may lead to questions that contain code or details that distract from the core question.

It would be best, as for all questions, to reduce the question to the core issue, in this case how to display a left and right y-axis where the left one has scientific notation and the right one has the same numbers but in simple notation. For an MWE that means you would remove essentially everything (i.e., all axis options, all addplots (or leave one with just f(x)=y), the legend etc., and just display the simplest possible plot with two axes in some range.

This has the advantage that the question is maximally focussed and easier to answer. It will be more searchable and reusable for future visitors. And furthermore, it will increase your understanding of pgfplots in a more fundamental way. Currently the answers that you got (although they contain useful explanations) are more or less the final step of your interaction with the AI bot, i.e., the answerers acted as the bot trying to improve the existing code (and succeeding where the bot could not). If the question would have been fully minimal then you would have needed to take that final step yourself, which would have improved your knowledge and skill more robustly.

As a side effect such a minimal question would be easier to identify as a duplicate, for example of Pgfplot with two y-axes not showing correct ytick values which has the same solution as given in the answer by Rmano on your current question - minus the exact positioning with log10(1050) etc, but you said you only wanted an approximate correspondence so just putting the second set of tick marks at equal distances would have worked fine I think, and if not then you could search for questions on how to position y ticks (such as yticklabels and ytick wont work) or how to use calculations inside tick mark position lists (such as https://tex.stackexchange.com/a/425277/89417).