It’s hilariously sad. Out of curiosity I told it to optimize a paragon board path for an Arc Lash Sorcerer. Not only did it not math out an optimized paragon board, but it also insisted that you should pick up resource cost reduction as a third priority.
Safe to say, AI isn’t taking over anyone’s job that uses math and theorycrafting any time soon.
I wouldn’t count it out just yet. Also did you specify what you wanted from said paragon board or just insert the whole board and tell it to have fun?
Yes, I gave it my gear and skill tree and said I wanted it to optimize my paragon board for damage.
OK, what were the first 2 priorities ? 
Also, what aspects did you have or did it suggest ?, did you have Edgemasters for example ?
And last but not the least - did you try ?, or say share the results maybe ? 
Certainly! Let’s break down the Paragon node allocation step by step to optimize your Arc Lash build:
This is what it suggested. I’m like, “Yeah, okay, ChatGPT. Clearly you don’t understand.”
Yes but that doesn’t mean it’s still the most DPS-optimized build
Perhaps even if you’re using ArcLash can still use a Core or Mastery skill ?, how about Conjuring ?
That’s the point, people think in limited categories, AI combines them to find the optimal, the question is whether or not it’s viable to play that way ?
But what was your initial prompt? What role did you ask it to play for you?
Base GPT4 doesn’t math, hook it up to wolfram alpha
The request from TC is too board and general. Need something more specific to work with. ChatGPT can do math. Request needs to be more clear like asking for a threshold. This is like asking “How to get from Texas to Hawaii?” There’s an answer, someone isn’t happy with the results.
It’s only as good as the instructions you give it. The governing principle is Sh*tInSh*tOut.
most probably , like chat gpt usually does, you need to explicitely tell him that the skill you are using won’t need any ressource.
you need to make one hell of a huge prompt for chat gpt to be kinda more accurate.
that being said it’s better to do your own, but still a fun experiment.
Paragon board is very high level stuff, it breaks even AI
That’s what poop-wave necros are for. The governing ideal outcome is poop out, 3 times with tidal aspect. 
@rambo, I think what you mean to say is even AI gets broken by Blizz math. Can’t figure out why text keeps mismatching effects.
ChatGPT database is based on June 5, 2023 information that are relevant of that time. It barely know Diablo IV sine D4 was released on June 5, 2023. Might want to try again next year.
I am also using ChatGPT to help me mod Diablo 2: Resurrected but I am having barely any success. I think I want ChatGPT to automatically fill out an excel sheet so I can save some time reworking items and such.
I think it really depends on the training of the agent. I was playing around with claude ai asking it questions about LE. It knew general information about the game but not any details regarding skills and such. It’s only a matter of time before it’s trained on pretty much every software package which includes games. This will be a big windfall to software companies because the amount of money they’ll have to pay for support will go through the floor.
Garbage in garbage out.
Don’t think GPT would be the model to use for this task first of all, but secondly you’d really need to spend A LOT of time preparing data, assigning weights and priorities, and training it with various situations. It would take a lot of iterations of human solving the best routes, then the model, then comparing / rewarding the model, then rinse and repeat to train the model to be able to do this on it’s own.