I was referring to that video in your previous post. It’s definetly exciting to think that we’ll be able to mod games by just giving some prompts (if that evolves into there as well), what I hope never happens is people getting too comfortable. People shouldn’t think having an AI assistant is a necessity and should still be able work without using it. I think we should be careful not to exceed a certain limit of luxury we can integrate to our lives, where we’re still able to fully function on our own, both professionally and personally.
What is that?? Yeah I don’t know about others, but whoever made that should definetly be called a Noodle Artist
I get what you’re saying though, yeah prompt engineering can indeed be more prestigious in the future. I just wish people don’t end up being addicted or even dependent on AI after this rise.
All I want AI to do is my chores while I take care of the BP wires. Although, I wouldn’t mind an AI BP assistant that can populate an array, mass rename variables, correctly copy data signatures, set up a basic material or particle effect. Make BPs smarter and less prone to inducing carpal tunnel syndrome…
Ran into similar situation just now (lost hours looking in all the wrong places). So asking AI to do an integrity-check or settings-audit could be interesting…
However, for me, the problem ended up being BP corruption. Mesh collision on a single mesh (in an identical group) just stopped working (never seen that before)… So gut feeling is, using AI for this or settings profiling, is probably a long way off and very game specific. For example, something that’s a glitch in your game, might actually be active gameplay in mine and so on.
Where AI would be immediately useful is taking away some of the drudge work. Think optimization / performance tweaks (custom lighting setups). As there’s so many options / settings / benchmarks. Having AI immediately show you all the best possible lighting scenarios would be huge. Plus how about never having to create a terrain / landscape manually ever again.
Plus what-if, AI could fix all the collision issues in your levels and walk the entire world to test. Then after, AI could add complete navmesh that works every time… Or better yet, no need for navmesh / nav-invokers ever again.
Think AI with human vision, that can just view and interpret the game scene same as a human… So all hard-coded AI logic or behavior trees could then be dumped. Instead you just type out: ‘play stealthy and cunning just like a human player’. Be just as sneaky / crafty / wise / cruel…
While its hard to know how far AI will go and how soon, giving a textual description of a level to AI, and getting it to create that level at least, would be huge today. Betting some studios have already built this tool in Unreal already, as it seems like its just a step up from proc-gen.
I wouldn’t mind an AI BP assistant that can populate an array, mass rename variables, correctly copy data signatures, set up a basic material or particle effect. Make BPs smarter and less prone to inducing carpal tunnel syndrome…
Make BPs Smarter!!! Now that’s a political slogan I can get behind!!
Where AI would be immediately useful is taking away some of the drudge work. Think optimization / performance tweaks (custom lighting setups). As there’s so many options / settings / benchmarks.
Exactly. At least 20% of my time is spent finding some snag that prevents me from working on what I actually need to work on. I imagine AI as an extension to the current Search bar in blueprints. It’s almost 2025 and the search bar still works on Tags. If you don’t remember exactly what something is called, your digging. Imagine typing “Take me to enemy spawn function” into search bar. A list pops up, you click and boom your there. Simple navigation speeds ups like that would be so valuable.
I just can’t imagine what sort of prompt you could possibly give the system to develop blueprint ( or C++ ) code for subsystems and connecting them all together, where you have to get subtle concepts across.
The use-case for AI in blueprints is most obvious in debugging, workflow and general navigation around the editor. If a projectile stops colliding, I should be able to write in AI co-pilot: “X projectile isn’t colliding with Y actor, scan collision settings and blueprints and show me the most likely reasons this is happening”. A window pops up with collision network showing why projectile x is no longer colliding with actor y and a series of quick fixes. Done. Saved me 2 hours of clicking around with print strings.
I don’t need AI to write code from the ground up. I need AI to handle all of the stupid time-consuming little snags. Using Unreal is a SLOW, SNAG INFESTED PROCESS. Lets make Unreal a quicker, more streamlined process Please.
Yeah I don’t know about others, but whoever made that should definetly be called a Noodle Artist 
