How Legit is AI for Rating?

[oldcelcel1983]

I use an app called attractiveness test apk, and when I take a candid head on photo with a neutral expression up close, I get rated a 5.3. I think this is like a 4.3 irl, as on this app, 6 appears to be average looking. My face also looks longer in the selfies than they do in the mirror.

Being 4.3 is pretty much dork/loser territory, which would explain why I never became popular in high school.

 

[Foreverbrad]

They just assess ratios. There are plenty of failos that don’t show up in ratios. I have near perfect frontal ratios but I’m LTN because of my eye area, side profile, nasolabial folds, jaw recession, bite misalignment, buccal corridors, skin quality, and sagging soft tissues.

I imagine it’s also true that your ratios could be off but you’re still HTN+ due to these “harmony” features,

 

[Deleted member 31445]

AI ratings app who need to sell to? Ugly ass niggas, because gl people know they look good, so rating are gonna be skewed up.

 

[oldcelcel1983]

AI ratings app who need to sell to? Ugly ass niggas, because gl people know they look good, so rating are gonna be skewed up.

So is a 5.3 on there a 4.3 irl?

 

[Alexanderr]

I use an app called attractiveness test apk, and when I take a candid head on photo with a neutral expression up close, I get rated a 5.3. I think this is like a 4.3 irl, as on this app, 6 appears to be average looking. My face also looks longer in the selfies than they do in the mirror.

Being 4.3 is pretty much dork/loser territory, which would explain why I never became popular in high school.

As of right now, pretty shit. They're all rather elementary in nature, maybe only being able to somewhat accurately gauge ratios, but don't account for a lot of other things that are important. Another thing a lot of them are prone to is bias.

You and I could both tell the following model is very good-looking, but the AI might detract a lot of points of his face because it's been primarly trained on say white faces.

Resulting it in giving him a score comparable with an average white guy, even though we can obviously tell he'd mog the shit out of average Joe's.

In the future, once they're more advanced and less prone to bias I imagine they could be very good but fundamentally rating remains rather subjective so how good is up for discussion.

 

[Deleted member 31445]

So is a 5.3 on there a 4.3 irl?

looking at @Alexanderr comment it probably depends on you, and how well you suit the app's data training

 

[oldcelcel1983]

As of right now, pretty shit. They're all rather elementary in nature, maybe only being able to somewhat accurately gauge ratios, but don't account for a lot of other things that are important. Another thing a lot of them are prone to is bias.

You and I could both tell the following model is very good-looking, but the AI might detract a lot of points of his face because it's been primarly trained on say white faces.
View attachment 2334381View attachment 2334387

Resulting it in giving him a score comparable with an average white guy, even though we can obviously tell he'd mog the shit out of average Joe's.

In the future, once they're more advanced and less prone to bias I imagine they could be very good but fundamentally rating remains rather subjective so how good is up for discussion.

He only tested 6.3 in the app. I wish I could test that high with a straight on candid photo.

 

[Graham]

They just assess ratios. There are plenty of failos that don’t show up in ratios. I have near perfect frontal ratios but I’m LTN because of my eye area, side profile, nasolabial folds, jaw recession, bite misalignment, buccal corridors, skin quality, and sagging soft tissues.

I imagine it’s also true that your ratios could be off but you’re still HTN+ due to these “harmony” features,

Makes sense, but when mid journey is told to draw countries as people, it produces gigamogger after gigamogger.