Well, this seems completely arbitrary and has the potential for a social media disaster. :P
All jokes aside though, I'd love to see what they base things on because there is literally no way you can tell by someone's looks (especially if it's just a picture of their face) how wise or dextreous they are. Just for fun I uploaded a picture that I found on a D&D character portrait site. It seems that beards make you strong but also less good? :D
Smiling and not having a beard also seems to do wonders for your soft stats. I googled pictures of the last two former US presidents with "[name] smiling" and "[name] frowning" and this is what they got. Please! No political commentary. I chose these to because they are well known public figures.
Attribute
Value
Strength
9
Dexterity
16
Constitution
9
Intelligence
15
Wisdom
16
Charisma
16
Alignment
Lawful good
Character class
Bard
Attribute
Value
Strength
11
Dexterity
16
Constitution
9
Intelligence
15
Wisdom
15
Charisma
15
Alignment
Lawful good
Character class
Rogue
Attribute
Value
Strength
15
Dexterity
10
Constitution
9
Intelligence
10
Wisdom
15
Charisma
8
Alignment
Chaotic evil
Character class
Cleric
Attribute
Value
Strength
16
Dexterity
15
Constitution
11
Intelligence
11
Wisdom
10
Charisma
10
Alignment
Neutral evil
Character class
Fighter
I have no idea what this means but I thought it was funny as hell. Again, please refrain from any political commentary. It seems though that the biometric website has a strong bias when it comes to smiles and beards. :)
Also it seems that a regular resting face is almost always CE barb.
I would seriously consider reporting those researchers to their university. This smacks of Phrenology, which is the debunked theory that the shape of one's head or bumps on one's head determine personality.
Sure, as a fun thing it can be just that, entertaining, but if they are claiming this to be some sort of actual legitimate research, well.... that's another story.
Kotath, this research is legitimate - it's been through an ethics committee and everything. But perhaps you've missed my descriptions, or not read the supporting statements. As all the support material says, they're not trying to create an AI that can make accurate judgments of people's attributes, but rather to examine the assumptions standard biometric facial recognition algorithms make by translating them into an entertainment context, and asking people how the results match up to their expectations.
As I understand it, the AI used in this project - and its predecessor the Biometric Mirror - uses some standard baseline data and algorithms common to facial recognition systems, but makes its own decisions, evolving through machine learning. Those baselines are indeed partially inspired by phrenology, as mentioned by one of the researchers on Twitter. So the point of the research is to find out how people feel about this, and how much they know or can infer from the results of a scan like this.
I would like to point out that an ethics committee does not legitimise research; they simply review it to ensure that it doesn't violate ethical guidelines. You can have completely illegitimate or frivolous research that gets ethics approval.
The goal seems to be a bit of a nebulous and self reinforcing. Algorithms don't make assumptions; they instead inherit biases programmed into them by the developer. It seems that the initial bias here is phrenological principles. Asking how people react to this also seems kinda nebulous; without knowing what the mapping intent is, how are people supposed to react? If the AI is programmed to map stoic to barbarian and cheerful to bard for example, then people could react, saying "Well I thought I was more cheerful than stoic but the AI disagrees. Maybe the AI is wrong?"
As it stands, the AI is blackbox associating visual traits with D&D parameters, thus meaning there's no way to actually really respond other than "Huh, I guess that's what it says." There doesn't appear to be any feedback loop for the machine to learn off other than it looks at faces and generates an output. It just feels like a crapshoot without any feedback, transparency in function, or rational
Due to a alarming trend in the past decade, I've been saying for months now that you cannot replace psychiatry with a computer. The "mind" isn't a logical puzzle with plausible physical representations but an emotional entity of intangible ideas.
These "compute your personality" things are gimmicks. Never take them seriously. Have fun with them! Treat them as the silly things that they are.
(Why have I spoken out against digital psychology for only months if the trend's been going on for a decade? Because a few months ago, someone came up with the bright idea to prevent suicide by using "deep learning" AI to bring back famous suicide victims virtually. I figured it's the brainchild of people who haven't faced suicide attempts themselves.)🤦🏻♂️
Human. Male. Possibly. Don't be a divider. My characters' backgrounds are written like instruction manuals rather than stories. My opinion and preferences don't mean you're wrong. I am 99.7603% convinced that the digital dice are messing with me. I roll high when nobody's looking and low when anyone else can see.🎲 “It's a bit early to be thinking about an epitaph. No?” will be my epitaph.
I would seriously consider reporting those researchers to their university. This smacks of Phrenology, which is the debunked theory that the shape of one's head or bumps on one's head determine personality.
Sure, as a fun thing it can be just that, entertaining, but if they are claiming this to be some sort of actual legitimate research, well.... that's another story.
Kotath, this research is legitimate - it's been through an ethics committee and everything. But perhaps you've missed my descriptions, or not read the supporting statements. As all the support material says, they're not trying to create an AI that can make accurate judgments of people's attributes, but rather to examine the assumptions standard biometric facial recognition algorithms make by translating them into an entertainment context, and asking people how the results match up to their expectations.
As I understand it, the AI used in this project - and its predecessor the Biometric Mirror - uses some standard baseline data and algorithms common to facial recognition systems, but makes its own decisions, evolving through machine learning. Those baselines are indeed partially inspired by phrenology, as mentioned by one of the researchers on Twitter. So the point of the research is to find out how people feel about this, and how much they know or can infer from the results of a scan like this.
I'm no expert in machine learning, but from my understanding there must be a feedback loop for it to learn. You need to tell it if it is correct or not, so that it can incorporate that into future results.
Asking how sometime feels about the result doesn't really sound like a reasonable way to determine if it was successful, unless your goal is to present the stats which make the person feel the best.
I don't have time to read much about the project, but what I have read about it on here doesn't inspire me with much confidence. In fact, the only thing I can see so far is yet more proof that phrenology, and any judgement of a person's personality etc based on appearance etc, is tosh.
I was under the impression that the project was (a) mostly for fun and/or (b) a demonstration of how AI and/or phrenology *doesn’t* work, and how fantasy archetypes are not always good. I hope the goal isn’t “train the AI to be more accurate,” but “understand the sociological impact of AI.” If it’s not...well, then we have a problem.
For an interesting if cheeky look at a similar (and much more catastrophic) AI facial recognition issue, check out this video: https://youtu.be/Ok5sKLXqynQ
Interesting, but I don’t think there’s much value in this, it seems to just associate most people with chaotic evil barbarian and the general trend for stats looks like it gives most people high strength and constitution, low intelligence and dexterity and somewhat varying charisma and wisdom, tried this with my group and we all got pretty similar results, the same alignment and class (CE barbarian), with maybe 2-3 points of differences across all stats.
I tried a few different facial poses: neutral, making a weird shape with my mouth, smiling, and mouth and eyes both wide open. Every single time it gave me scores around 15 in str and con, with low int and wis, and CE barbarian. For reference, I'm a 210ish lb. male high school English teacher who hates going outside.
I was curious whether or not this could be used with an actual piece of art, instead of your physical picture, so I used an image on my phone, and held my phone up to the camera. Turns out it works just fine.
To be fair to the program, I also used my real face.
Apparently, I have a bent for being Chaotic Evil, or so I thought, until I did another one, with a pair of glasses on and a weird, looking of the rim stare.
It would be interesting to see what it gave me, after I have getting rid of the beard, been the spar and everything, and looking a bit younger. I might try that actually, as I do clean up rather well. Right now I look a bit scruffy after spending so long stuck in the house.
Rollback Post to RevisionRollBack
A caffeinated nerd who has played TTRPGs or a number of years and is very much a fantasy adventure geek.
To post a comment, please login or register a new account.
Also it seems that a regular resting face is almost always CE barb.
"h"
Kotath, this research is legitimate - it's been through an ethics committee and everything. But perhaps you've missed my descriptions, or not read the supporting statements. As all the support material says, they're not trying to create an AI that can make accurate judgments of people's attributes, but rather to examine the assumptions standard biometric facial recognition algorithms make by translating them into an entertainment context, and asking people how the results match up to their expectations.
As I understand it, the AI used in this project - and its predecessor the Biometric Mirror - uses some standard baseline data and algorithms common to facial recognition systems, but makes its own decisions, evolving through machine learning. Those baselines are indeed partially inspired by phrenology, as mentioned by one of the researchers on Twitter. So the point of the research is to find out how people feel about this, and how much they know or can infer from the results of a scan like this.
I would like to point out that an ethics committee does not legitimise research; they simply review it to ensure that it doesn't violate ethical guidelines. You can have completely illegitimate or frivolous research that gets ethics approval.
The goal seems to be a bit of a nebulous and self reinforcing. Algorithms don't make assumptions; they instead inherit biases programmed into them by the developer. It seems that the initial bias here is phrenological principles. Asking how people react to this also seems kinda nebulous; without knowing what the mapping intent is, how are people supposed to react? If the AI is programmed to map stoic to barbarian and cheerful to bard for example, then people could react, saying "Well I thought I was more cheerful than stoic but the AI disagrees. Maybe the AI is wrong?"
As it stands, the AI is blackbox associating visual traits with D&D parameters, thus meaning there's no way to actually really respond other than "Huh, I guess that's what it says." There doesn't appear to be any feedback loop for the machine to learn off other than it looks at faces and generates an output. It just feels like a crapshoot without any feedback, transparency in function, or rational
Find my D&D Beyond articles here
Due to a alarming trend in the past decade, I've been saying for months now that you cannot replace psychiatry with a computer. The "mind" isn't a logical puzzle with plausible physical representations but an emotional entity of intangible ideas.
These "compute your personality" things are gimmicks. Never take them seriously. Have fun with them! Treat them as the silly things that they are.
(Why have I spoken out against digital psychology for only months if the trend's been going on for a decade? Because a few months ago, someone came up with the bright idea to prevent suicide by using "deep learning" AI to bring back famous suicide victims virtually. I figured it's the brainchild of people who haven't faced suicide attempts themselves.)🤦🏻♂️
Human. Male. Possibly. Don't be a divider.
My characters' backgrounds are written like instruction manuals rather than stories. My opinion and preferences don't mean you're wrong.
I am 99.7603% convinced that the digital dice are messing with me. I roll high when nobody's looking and low when anyone else can see.🎲
“It's a bit early to be thinking about an epitaph. No?” will be my epitaph.
I'm no expert in machine learning, but from my understanding there must be a feedback loop for it to learn. You need to tell it if it is correct or not, so that it can incorporate that into future results.
Asking how sometime feels about the result doesn't really sound like a reasonable way to determine if it was successful, unless your goal is to present the stats which make the person feel the best.
I don't have time to read much about the project, but what I have read about it on here doesn't inspire me with much confidence. In fact, the only thing I can see so far is yet more proof that phrenology, and any judgement of a person's personality etc based on appearance etc, is tosh.
I was under the impression that the project was (a) mostly for fun and/or (b) a demonstration of how AI and/or phrenology *doesn’t* work, and how fantasy archetypes are not always good. I hope the goal isn’t “train the AI to be more accurate,” but “understand the sociological impact of AI.” If it’s not...well, then we have a problem.
For an interesting if cheeky look at a similar (and much more catastrophic) AI facial recognition issue, check out this video: https://youtu.be/Ok5sKLXqynQ
Wizard (Gandalf) of the Tolkien Club
Interesting, but I don’t think there’s much value in this, it seems to just associate most people with chaotic evil barbarian and the general trend for stats looks like it gives most people high strength and constitution, low intelligence and dexterity and somewhat varying charisma and wisdom, tried this with my group and we all got pretty similar results, the same alignment and class (CE barbarian), with maybe 2-3 points of differences across all stats.
I tried a few different facial poses: neutral, making a weird shape with my mouth, smiling, and mouth and eyes both wide open. Every single time it gave me scores around 15 in str and con, with low int and wis, and CE barbarian. For reference, I'm a 210ish lb. male high school English teacher who hates going outside.
I was curious whether or not this could be used with an actual piece of art, instead of your physical picture, so I used an image on my phone, and held my phone up to the camera. Turns out it works just fine.
To be fair to the program, I also used my real face.
Apparently, I have a bent for being Chaotic Evil, or so I thought, until I did another one, with a pair of glasses on and a weird, looking of the rim stare.
It would be interesting to see what it gave me, after I have getting rid of the beard, been the spar and everything, and looking a bit younger. I might try that actually, as I do clean up rather well. Right now I look a bit scruffy after spending so long stuck in the house.
A caffeinated nerd who has played TTRPGs or a number of years and is very much a fantasy adventure geek.