The GREAT 40k thread

Well painted. I really want to like the model as there are many things I enjoy it's just so busy. This paintjob cleans up a lot of the GW ultra highlight fixation scheme but it's still feels cluttered.

 

Not quite four years ago, a bit more than that (2019).

 

Legend.

 

That is Majin Boo on his ride!

Grrr, !mrahil

 

There is no escape from Slaanesh. She marks us all.

 

....
nope. AI done messed this one up. Three sets of fingers, two on the gun plus a free hand? Creepy enough to begin with and shall haunt my nightmares, but that just makes it worse.

 

They should have made a Chaos Space Marine... then the extra hands and weird finger issues could have been attributed to Chaos mutations!

 

You're right, that's creepy. What is it with AI and hands?

 

It wouldn't surprise me if it had something to do with how it's "learning" from the training data. Current "AI" models at best have a very rudimentary ability to correlate words with images of objects that are "alike", but they do not demonstrate the ability to understand what the object is supposed to be, let alone how it is supposed to interact with other objects or why.

When it comes to hands in particular, think of how many different ways a hand can be depicted, not just from orientation relative to the viewer but also from posture and gesture. It's intuitive to us that a hand is a versatile body part whose purpose is to manipulate objects in one's environment (if not also a means to communicate), but to an "AI" it's just a wide range of wildly different shapes that somehow correlate with an arbitrary "hand" descriptor. With such variability in how a hand can appear, all the "AI" can really do is essentially guess at what visually defines a hand according to the training data it was provided, which in most models ends up being a composite average that dives deep into the Uncanny Valley.

 

I get your point on positional complexity, but you'd think by now AI would learn that humans have two hands (barring disease or injury).

 

Two things are working against current "AI" models that lead to this kind of a result: they're eating their own shit as an increasing portion of their training data, and they lack object permanence. They may have some recognition of the shapes needed for a hand or arm (however fucked up they may be), but if an intervening object is in the way of how they interact in the image (i.e. a gun), a program designed to make 2-D images of 3-D objects does not intuit that the two features are supposed to be part of a contiguous whole.