So, I heard about Ludus AI, so I tried it out. I asked Ludus AI to create me a blueprint where when a player gets into a certain distance to a mesh (from an array of mesh), the mesh will turn from invisible into a generated material then into a grey one. But when the character gets out of the radius from the mesh, it will turn back invisible. This is what Ludus AI told me to do, and I cannot get it to work. I probably forgotten to do something. I wanted to see if Ludus AI was correct, and I am just getting it wrong. So does this look right?
Hello @VeveMurre
This might work in the end but for me it’s a bit over engineered. AI’s can be helpful but often times they just spit out gibberish. It’s all about how you form your questions. I personally use chat gpt to help me break down more complicated logic. It’s all about refining your questions.
Now where it comes to having an invisible actor (Actor Hidden in Game checked), you could have an actor with static mesh and a collision sphere/box/capsule and when a player comes in range (overlaps said collision volume), show the actor and then have your materials applied.
Problems with LLM (really any derpy derp pseudi “Ajj”) is that they do not follow logic, instead they compile best looking answers, that usually are correct (when you ask about something they were trained on) but remaining 10% of answers they just make stuff up (totally untrue but looks almost perfect).
So yes it does look really god, is it correct, not sure.
However with all that ranting about derpy derp ajj, i am using plain old chat gpt to speed up my C++ coding, and it is great, no scripts no tools, just asking it (my yellow rubber duck). Or telling it my plan (and it makes more coherent resume of what i plan to code) , it is great for creting all that standard boiler plate code.
But then: (when it goes out of derpy derp comfort zone, aka. outside data it was trained on):
I asked it if UEFN (unreal for fortnite) had any major improvement that would let me do something that is not fortnite (3rd person looter shooter clone) like game (I wat to recreate old puzzle game in fortnite as learning project). And sure it was possible (really annoying they always want to help even by creating stuff up). And it started giving me VERSE code for uefn that was basically some java script for such puzzle game translated into VERSE. Nothing of what it created was implementer or exposed to VERSE in fortnite. It was like chatgpt took some source from java, translated it to verse and used API from unreal engine.
so again TL/DR
it helps but when you outside its data set it makes up stuff that looks great, but it is made up.
Yup same here. I give it description about what i want to do, ask not to spew all code yet. THen i split whole problem into tiny tasks and last part is to ask about single steps in code. for eg: “Now i need function that as parameter uses Vector3d (64-bit one) and checks distance from vector3d stored in LastPawnLocation, then gives back true if its in range” Such questions get me almost always perfect code, anything more complicated and it will be messy over-engineered mess. That looks great but does some unnecessary random crap in middle.
Totally agree on LLM’s making stuff up
At time you can tell it has no idea about it, you keep asking the same question in 15 different ways but you keep getting the same answers (or set of few ones that will be recycled over and over and over…)
Just like @Nawrot said, try to break it down into smaller parts. I will sometimes spend 10-15 minutes typing out detail description of what I want to achieve alongside my own ideas how to do it. This helps me out with picking the “right path” for what I want to do as well my current skills. Not everything needs to be a perfectly engineered solution.
When I venture into unknown territory (like some C++ code) I actually ask chat gpt to break everything down, line by line, character by character, asking what would happen if I changed things around and why it might be a good/bad idea.
I think you came to the right place looking for answers. This forum is a great source of knowledge. I certainly learn something new here everyday.
