Can ChatGPT be used to identify a tartan?

[Canadian Vet]

Can a Cuttlefish or a Chameleon match a tartan?

Just wonderin'

[jsrnephdoc]

Can a Cuttlefish or a Chameleon match a tartan?

Just wonderin'

Maybe an Octopus would be more successful but oh, my, which of those eight invisible shoulders would host the piper's plaid brooch?

[OC Richard]

You can't map between appearance and thread count...you can't catalogue a person's genome from their face.

Thread counts won't do. We can recognise these as being "the same tartan" but each has a unique thread count. (I know, I generated the images.)



It's the old "I can't put it into words, but I'll know it when I see it."

About faces, I was disgusted by a television programme attempting to use a computer program to quantify how much pairs of human faces resembled each other.

It was preposterous. They were looking only at "features": to what extent the eyes, mouths, and noses were alike.

Anyone who knows faces knows that you can identify a person quite accurately with all the features covered or blurred out; you identify a person by the shape/contours and proportions of their head and face. Note that you can spot a person you know only seeing the back of their head! Or when they're wearing sunglasses and a mask covering their mouth and nose.

It's how you can tell so-called "identical twins" apart. They're never identical. The differences usually aren't in their "features" but in extremely subtle proportions of their head and face, generally the area around the cheekbones. (Goes to show that identical DNA expresses itself slightly differently in each iteration.)

[jsrnephdoc]

Thread counts won't do. We can recognise these as being "the same tartan" but each has a unique thread count. (I know, I generated the images.)

I'm not sure I understand (actually, I'm sure I DON'T understand). Each of those "electronic swatches" bears a label. The three are all different. I'm going to assume that these are registered designs, and if so, isn't it the case that the different thread counts makes them different tartans?

Again, I don't know all about the Registry, and I thought (as I think you're trying to imply) that designs seeking to be registered as different but which actually LOOK identical are not permitted, yet if these ARE all registered as different, they seem to have survived that fuzzy logic "what does it look like" inspection.

[figheadair]

I'm not sure I understand (actually, I'm sure I DON'T understand). Each of those "electronic swatches" bears a label. The three are all different. I'm going to assume that these are registered designs, and if so, isn't it the case that the different thread counts makes them different tartans?

Again, I don't know all about the Registry, and I thought (as I think you're trying to imply) that designs seeking to be registered as different but which actually LOOK identical are not permitted, yet if these ARE all registered as different, they seem to have survived that fuzzy logic "what does it look like" inspection.

They are all the same tartan. Richard's point was, I think, that changing the threadcount or shades does not alter the tartan but that loading one int ChatGPT would not necessarily allow the orpgram to identify the same tartan is a different configuration or shades.

[OC Richard]

I'm not sure I understand (actually, I'm sure I DON'T understand). Each of those "electronic swatches" bears a label.

Those "electronic swatches" (a good name for them) were my eyeball reconstructions of three examples of the same tartan, Prince Charles Edward Stuart. The claimed dates range from the mid-18th century to the early 19th century. (None of them have a bill of sale or other documentation or provenance giving them an air-tight date.)

One is a length of cloth at Glamis Castle (bottom)

One is a pair of trews at the West Highland Museum (top right)

One is a tartan suit at the National Museums Scotland (top left)



The point is that one can encounter "the same tartan" multiple times, each having a unique thread count.

Our hypothetical computer program would need to be able to recognise, in this case, a certain range of permissible widths of the red band flanked in azure to still be called Prince Charles Edward Stuart tartan.

But not too wide!! Because if that red band gets wide enough (and the azure stripes narrow enough) the tartan is then called Royal Stewart.

Here's Prince Charles Edward Stuart (West Highland Trews sett) compared to Royal Stewart.

In the three PCES settings at the top of my post, and Royal Stewart seen below, the green area varies very little, the changes being almost entirely in the width of the azure stripes and red band.



Here's the modern setting of Royal Stewart that I put into the old colours to compare. The green is so dark that it's hard to distinguish from black.