GW News: SKAVEN VICTORY! #NewAoS

I agree that this one is still a little ways away. The AI has to bet much better in order to compete with GW, but AI is advancing at a scary rate. Exactly when we'll reach that point is anybody's bet. I'd be surprised to see it within 5 years, but I'd also be equally (if not more) surprised to see it outside of 25 years. I'd guess somewhere in between those two numbers.

Also, I'd expect it to take over in stages. First as a tool by professional sculptors to streamline their process, then by skilled hobbyists and further down the road, everyday hobbyists.

In that we agree. GW is enjoying a golden era at the moment, but competition from 3D printing is ever growing. I believe there will come a time when it becomes the norm. 3D printing is only going to get cheaper, better and more convenient as the technology rapidly develops. GW will try to stave off the transition as long as possible, but eventually it will become unavoidable.

Come on Skynet... daddy needs a new Chaos War Mammoth!

 

we can play that game with GW’s pricing if you like



Handling resin is about as toxic as handling petrol or motor oil which most people do with adequate precautions all the time when they fuel up their car.


You’ve only got to decide on a very small/moderate army project for a £300 printer investment to pay for itself.

 

Yeah, but they do so outside. Working with resin inside your home introduces additional risks and the need for proper ventilation.

 

Yes Indeed, I didn’t factor in the costs of a hazmat suit or clean-room level precautions!

jokes aside, for people who are worried about risks or lack the space such as if they live in a flat for example, then GW (or whoever) producing models will always be an option, I just don’t think GW will be selling traditional kits, they’ll be printing them to order.

 

Come on Skynet... daddy needs a new Chaos War Mammoth! [/QUOTE]
Don't hold your breath, you're probably going to be dissapointed for a long while

In all seriousness though, it mostly depends on what exactly you expect the AI to do.
If you expect actually new models made by AI, that's going to take a while, and it'll take some rather fancy computers to build the AI.

If you expect AI to basicly kitbash together say a space marine variant based on the existing bits, you can more or less do that with current technology, especially if you just make it learn seperate bits (e.g. teach an AI how to print a space marine arm with a gun). and then just stick the bits together. Though chances are it'll be a pretty mediocre kitbash 99 out of a 100 times

I mean, you'll need those things regardless of how you get the models. Buy them at GW, or print them, you're going to need a brush and some paint.
So not exactly the right comparison.

Which people do outside, where they don't have to worry about airing out a room. And in the case of petrol, they do at a gas-station, so they don't need to worry about safety at all, it's taken care for them.

A printer they'll need to put in their own house. They need to have the space, they need to be able to ventilate the room etc.
It may not really require much to handle safely, but it's still an extra thing that stops people from buying the thing.

For example, the ventilation in my appartment is terrible. I literally can't get a 3D printer cuz of it. Similarly, I can't basecoat models inside.

Look it's quite simple. 300,- is a big enough upfront investment that people will regret it if it doesn't turn out well.
And since it's scary unknown technology, with limited use outside of the hobby, a decent learning curve and some extra requirements like safety, on top of being a big upfront investment the average person is simply not that likely to bother to try. Simply put, it's a big and risky investment for most people,or at least they perceive it as such.

For 3D printing to become a common hobby tool it'd either needs to become so cheap that it's no longer a big upfront investment. Which means you need to get a decent printer + some resin for like 50,- so you can make a couple of test-models without losing too much money if it turns out badly.
Or 3D printing needs to be such a common tool that it stops being a scary unknown technology. If people have more exposure to the tool they'll know what the investment is worth, and they won't be as worried about potential problems/dissapointments.

And neither of those are likely to happen any time soon.

 

I think it’s easy to be apprehensive about a subject you don’t know much about and I was exactly the same as you until I gave it a go - much like GW's offering, you don't actually need any of those optional items apart from the resin itself.

You missed my caveat regarding people who don’t have space/ventilation (I.E in a flat or apartment) by which they may have no choice but to purchase kits from a manufacturer which in turn would keep a company like GW in plenty of business.

Of course it depends on the individuals appetite for learning/effort. It’s more convenient to just buy a kit and of course you pay a premium for that convenience.

Of course there’s more than just an upfront cost in money terms for 3D printing, you have to be willing to learn new skills, support unsupported files, post print clean and cure models as well as maintain the printer and troubleshoot any issues.

If that puts someone off because they can’t be bothered, then by all means, they may continue to allow GW to pull their trousers down. But for someone who is willing to put in a little bit of extra effort, it opens up infinite possibilities when it comes to army building and model design. I think most people who are passionate about modelling would want of take advantage of that.

 

Sure, but that's kind of the point. Most people know absolutely nothing about 3D printing, and just see a 150-300 minimum pricetag.
Very few people are willing to spend that kind of money just to try something they know absolutely nothing about.

Hence, why for 3D printing to become a serious threat to GW it needs to either become cheap enough that people are happy to try it out even when they know nothing about it, or common enough that people can try 3D-printing in an easy manner (e.g. if every grade schooler got to play with a 3D printer during class they'd naturally get some exposure to it and wouldn't find it so scary to spend 300 on one)

You are vastly underestimating how "scary" those extra requirements are for most people.

To put it into perspective, just look at the amount of people who have to call a friend or family member whenever they have some basic issue with their computer or internet.

In general it's very similar to computers and the internet really. Except that 3D printers are currently at the stage computers were at in the 70's-80's. It's an expensive arcane tool only really used by a handfull of nerds and geeks and by big fancy companies. Which is a massive barrier to entry to those tempted to try it out.

And honestly, even now, a large amount of people are still terrified by computers and constantly need help to achieve much with it. And that's despite it being a relativly common household object at this point.

 

Eh... There's a lot more that needs to be factored in, like the fact that the learning algorithms behind the likes of Midjourney and ChatGPT

1. can't innovate anything, much less do so without being prompted;
2. don't truly understand what it is they're being prompted to do;
3. have a habit of hallucinating to a degree that makes conspiracy theorists look sane; and
   
4. always trend towards removing outliers in progressive sampling (aka de-noising).

More crucially for this discussion, learning algorithms that focus on images are also notoriously bad at interpreting what goes where when adapting their sample material into generated content, and that's just in two dimensions.

That does not describe a tool that should be used for the final design of a product.

I can guarantee you that outside of concept drafts you're still going to be seeing digital sculptors (professional and otherwise) working the files manually, even if it's with pre-made assets (which GW already does in-house when designing new kits). In fact, I'd wager that Heroforge clones will sooner be the way of the future for tabletop minis than anything that involves an over-hyped artificial idiot behind the creative wheel.

 

I don't disagree that there many hurdles that must still be crossed. We're definitely not there yet. I'm just saying that AI is growing at an exponential rate. Where was AI 10 years ago? Versus 5 years ago? Versus now?

What will AI look like in 5, 10 or 20 years from now?

 

I wouldn't count on it. Especially not with the way a lot of tech-enthralled idiots crow on about some supposedly revolutionary thing that they themselves don't adequately understand, yet adamantly claim to be "the Future"™ and thus want to insert it into everything they can get their hands on, regardless of whether or not it makes sense to do so. The fact that a lot of them seem hell-bent on ensuring that even their goddamn toasters be able to write a Shakespearean play, if for no other reason than to say that it's "AI-powered", speaks to the kind of egotistical hubris that pollutes discussion of the subject matter in a way that we've already seen with cryptocurrencies and NFTs within the field of cryptography.

These are the kinds of people who'd waste time and resources needlessly trying to program self-awareness into a machine solely designed to pass butter.

And that's before getting into its use as a form of marketing jargon, to woo people who don't know any better into buying something they could have gotten elsewhere for much less and without the unnecessary baggage.

Long story short, as much as I'd like to be proven wrong, I've yet to encounter a compelling argument that supports the implementation of expensive and complicated AI or AI-derived technologies in applications that can suffice with something "dumber" instead. After all:

Keep it simple, keep it dumb
Or else you'll end up under SkyNet's thumb

 

Looking at it from a pure creativity standpoint (ignoring the huge benefits in research fields etc) I think we are at the stage where a designer or artist can leverage AI technologies as another tool in their arsenal.

For now there will always need to be a human to quality control, fact check and touch up/change whatever is produced by the AI, because for all intents and purposes it’s just “randomly” generating something based on the information that it’s been fed. But that can be a great starting point for a creator that could save hours and hours of work.

Having said that, these tools can produce some absolutely amazing results. This is my favourite showcase for what AI technologies can do:




All the details are on their website but the AI was trained to recreate images from Dune in the stylings of H.R Giger (Alien) and it did an absolutely fantastic job.

To suggest that doing something similar with 3D models is an impossibility is short sighted.

 

Honestly; more or less the same, at least as far as the general population will be concerned, just with less tech-bros and linkedIn-guru's hyping it up to scam people out of money. The actual innovations that have been done don't really get much attention, and aren't nearly as marketable.

ChatGPT & such aren't actually new technology. They're not doing anything different than the older LLM.
They're just bigger. Fundamentally chatGPT isn't actually that different from say Eliza
And the big difference between chatGPT 3 and 4, which gave it such a significant boost, is that now it uses ensemble learning, which also isn't cutting edge innovation. Honestly, I'm surprised they didn't use it to begin with.

We're already seeing AI cannabilism which is frankly hilarious.

By focusing on smaller subtasks it becomes easier to make a reliable AI. So if you create seperate AI's for seperate bits (e.g. an AI for generating weapons, one for generating arms, one for generating heads etc.) you can probably get a copy of heroforge going.

However, more important than any of that; you can't use midjourney directly to make miniature models. Because even if you ignore the issues like hallucination or the fact it can't make anything truly new, there's the very simple issue it doesn't know how to ensure structural integrity for the model, how to make sure the miniature doesn't just fall over, etc. Right now the only thing it can make are drawings, and maybe digital 3D models. But not physical 3D models.

And training your own AI to be able to make physical models would cost millions, if not billions.
It's cheaper to just get a handfull of artists to churn out some models.

 

I kinda like that palanquin/altar thing. Might be a good basis for a Grail reliquae

 

I quite like all of them, it’s like the empire had a grimdark re-fresh (if painted appropriately).

 

These models are okay, but nothing that I would ever consider buying myself. I thought others might be interested though.

 

I am pretty impressed with the level of detail and renaissance grimdarkness the Cities of Sigmar range is going for. The grave warden guy would make a fun necromancer conversion.

And that cannon is more a diorama than an artillery piece. There's so much detail...I'd hate to paint this!

If I didn't have several projects and a veritable pile of grey plastic...i'd be tempted by some of these models just for the painting potential.

 

And the only thing I like about this lot (apart from the Manticore showcased a month or so before) is the little dragon-dog-thing that will be the Cities of Sigmar Gryph-Hounds, I like that he's a miniature, wingless Y Ddraig Goch. He has potential for this fantasy contingent of my planned Dark Ages Welsh army, and I hope GW consider releasing a set of them as they did with Gryph-Hounds.