Incline How long until AI is plugged into Unity/UE so we can have it make a game by telling it what to do?

[anvi]

What do you reckon?

 

[Late Bloomer]

I believe both self and forced regulations will hinder AI greatly. Unity/UE both being commerical products won't help matters in regards to using AI with them. They will both want their own AI plugged in. It will all be regulated and cost money. The market in its current form (locked behind DRM / Consoles) won't allow anything not fully compliant wtih all existing copyright and other laws. I chose the give it 5 years option because of that. Of course, this speaks nothing of the quality of games being made.

 

[PorkyThePaladin]

Unity is notorious for being very slow and having terrible performance, because it has to cover all possible scenarios (as opposed to a specific game engine for a specific game). Current large language model AI is also notorious for creating inefficient code since it's just plagiarizing code published online by humans and applying it to various situations it wasn't exactly meant for. So if you combine the two... you will probably get something like early Shitmaker, where 70% of the game was spent on loading screens.

 

[Modron]

How long until someone trains the AI on Redaxium to make more sequels?

 

[NoMoneyNoFameNoDame]

20 years would easily be enough. 5 years you will already have significant competition with basically little kids. 10 years, you unquestionably won't be making money from simple games anymore.
Infact simple games have been the only way to make money for indies. So yeah 10 years = death of indies. 5 years = insane competition.

Before this AI stuff I already predicted we were in the final 4 years of tangible self-publishing.
In 5 years gamedev is not a viable job except if you're doing it for suckers who don't realize this.

We are dangerously close to gamedev being an elitist activity only. It's by chance you get recruited by someone with enough money to risk it.

On the upside if you don't mind doing things for free or always wanted to take gamedev more seriously, you should be fine. Shit's only getting easier.

However never understimate the dumbing down ofthings, needing to many subscriptions, wallgardens everywhere... to addup and ruin the benefits of the tools your using.

Gamedev will literally soon become like penciil and paper. But worse humans are shallow and easily fooled, so the person with most skill won't standout.
It's just lucky AI gen.

Gamedev will be the new porn, but it's generated.

If you're an influencers with a semi-successful channel this is a great time to be alive. You will not run out of content to try. All you have to do is give your shitty commentary and you'll actually get paid, unllike the gamedev.
Well to be fair.. fuck this earth.

 

[toucanplay]

AI is incredibly overhyped right now, and all the fanboys gushing over it are fucking retarded, hucksters, or both, and companies using it to replace humans are jumping on the bandwagon far, far too early.

Right now, LLMs can't do things like answer "is n a prime" correctly, even for the kind of small values of n that you can look up in a table. If you want to write something bland or average, then go ahead (maybe it'll put Obsidian out of business lol), or use an older model where things only make sense locally and make a watered-down Xavier Renegade Angel RPG.

Current generated art is also either the nicest-but-blandest-looking thing in existence, or some hideous monstrosity. Watch out if you get too specific, though, because some models have a limited training set for certain things, and you'll be at risk of replicating/plagiarising existing art from the training set (which has happened with Midjourney with the famous green-eyed Afghan girl photograph) and good luck not getting sued over that if you decide to share the game, or even screenshots of it.

Code-wise, it can be good for simple functions (i.e. things most devs would copy-paste already), but isn't trustworthy for anything that can be as complicated as a game, particularly stuff like RPGs. Will a non-coder know that some failing bit of code is almost correct and only needs a small change to work perfectly, or will they go "ugh, this sucks" and generate something else?
Will they be happy that it seems to do what they want, only to be surprised a hidden side-effect has been overwriting or corrupting their other files? A lot of code these models have been trained on will not have good documentation, unit tests, or a whole bunch of stuff that coders should be doing to make their code understandable and robust, so the generated code might not have this either. There's enough absolutely awful human-written code out there to be worried about anything that uses it to "learn" how to code.

So yeah, you'll be able to make your own games using AI soon, but they're going to be as boring as the current middle-of-the-road game.

 

[The_Sloth_Sleeps]

I do not understand the art hype at all. I have not found a single use for AI art. It has nothing to offer. I don't think it has anything of value to offer with game art, everything is more work and it does not integrate into workflows. Besides how problematic it is, it just looks so bad, its just got "that" sort of look which is really off putting. But if there's a market for it then its in the kind of dad photoshop, clip art, fliers type market.

Bing chat is not a programmer and cannot write code.

Bing chat is useful for correcting syntax, or looking up syntax, and its also useful for providing template code. It comes with some very large caveats however, you have to keep a very close eye on what it produces because it gets the code wrong every single time when it is writing more than a few lines. Lets be honest, its just a fancy copy n paste from stack overflow. I do find it useful though especially when I am tired, and I need to know something like how do I write a delegate function to take a template parameter. However that said, I would not pay money for it.

Now, Unity and UE4. Both are decrepit games engines which are a struggle to get anything workable. Throw in dodgy AI to an already dodgy games engine, guess what the result will be? Pretty pretty poor. I don't think theres any intention to ever have an auto game generator. I don't think anyone would take it seriously.

However, you can bet your mortgage that Unity and Unreal will be all over AI like a fly over shit, trying to put AI into their workflows. So thats good news for the AI fappers here. We might see things like level & asset generation (which we already have proc gen of course) worked in, but I think thats as close as it will get to generating a whole game.

 

[toucanplay]

I do not understand the art hype at all. I have not found a single use for AI art. It has nothing to offer. I don't think it has anything of value to offer with game art, everything is more work and it does not integrate into workflows. Besides how problematic it is, it just looks so bad, its just got "that" sort of look which is really off putting. But if there's a market for it then its in the kind of dad photoshop, clip art, fliers type market.

Last time I played around with it, I wanted to use it to come up with visuals of people and places for a novel I was writing to help cement the details. It might have been good for portraits, but it was terrible for cities/geographic features, particularly if you already had some kind of location restrictions on them. Apparently it was an issue with it not handling complex relative arguments well (stuff like "a red box on a green box on a red box"). It might work as a way to communicate stuff to an artist if you're not an artist, but for anything else it's just too bland or doesn't work.

I did manage to generate some hilariously nasty "shock porn" images that were fantastic at making other people upset, so it's not completely useless.