Avenue wide open for an ethnoguessr style homoguessr game, either using averaged faces, or using actual random photos and comparing human scores to ai scores
Humans have high sensitivity (face in the clouds!) but poor specificity (men can imitate women fairly easily) because it was never evolutionarily that useful.
The same AI profiling which thought black faces were apes shows what poor sensitivity will do, and confusing grooming with sexuality will be humorous over time.
You need to present more negative contexts (x-rays coding data on images triggers diagnosis), it’s not uniformly working to say the least 🌞
as probative in a given individual case, worthless. 99% wouldn't be sufficient, and the AI isn't even close. But, ooh, scoring far above random chance...I suggest making it into a wagering game, people will bet on anything nowadays
From a footnote here “No single skeletal feature definitively determines racial classification; rather, combinations of features that appear more frequently in one group allow for accurate identification.”
On some level, people are often confident that they can identify racial backgrounds/makeup in others. Long ago, NatGeo had an issue on race. Scientists at that time concluded that there were as many differences between individuals of one “race” as there were between features said to represent one race versus another.
In fact, things like epicanthic eye folds weren’t a determinant factor that many think they are. What one thing can be said? Ear wax. Sticky vs. crumbly. Caucasoid types have sticky and Asians have crumbly. This as a sole dividing feature may have changed since I read that and full genomes have been sequenced.
On a social impact level, race still impacts many, if not most, lives. I wonder if there are fewer impacts as more people identify as bi-racial or multi-racial and as many grow up exposed to people of many ethnic groups. Younger people seem less hung up on it.
AI for lie detection sounds extraordinarily important.
Avenue wide open for an ethnoguessr style homoguessr game, either using averaged faces, or using actual random photos and comparing human scores to ai scores
How about an app for that?
There won’t be one app, there will be many apps. You have to be clear what you are looking for.
Humans have high sensitivity (face in the clouds!) but poor specificity (men can imitate women fairly easily) because it was never evolutionarily that useful.
The same AI profiling which thought black faces were apes shows what poor sensitivity will do, and confusing grooming with sexuality will be humorous over time.
You need to present more negative contexts (x-rays coding data on images triggers diagnosis), it’s not uniformly working to say the least 🌞
> Men can imitate women fairly easily
Imitate yes, pass as, no. Only a select few twinks can really pull it off, and even then they're pretty obvious when next to a female.
100% agreement.
Mimic, imitate but not pass, not persistently.
Once you’ve seen it you can’t unsee it.
Is there any literature on detecting Aspergers?
Yes
https://pmc.ncbi.nlm.nih.gov/articles/PMC10856091/
Scary and possibly evil
Hmmm, this kind of technology could be used to determine whether a prisoner gets parole or not based on the likelihood of re-offending.
I feel like there's a tv show I saw that criticizes exactly this.
That's partly why I said parole. It would most certainly be a terrible idea to use it to determine the outcome of trials.
Anyone run an automated version you can upload your own pictures to?
as probative in a given individual case, worthless. 99% wouldn't be sufficient, and the AI isn't even close. But, ooh, scoring far above random chance...I suggest making it into a wagering game, people will bet on anything nowadays
The Victorians were right! Phrenology is real!
Wonder if this could be integrated into a “goggle glass” or other system for us to take everywhere
From a footnote here “No single skeletal feature definitively determines racial classification; rather, combinations of features that appear more frequently in one group allow for accurate identification.”
On some level, people are often confident that they can identify racial backgrounds/makeup in others. Long ago, NatGeo had an issue on race. Scientists at that time concluded that there were as many differences between individuals of one “race” as there were between features said to represent one race versus another.
In fact, things like epicanthic eye folds weren’t a determinant factor that many think they are. What one thing can be said? Ear wax. Sticky vs. crumbly. Caucasoid types have sticky and Asians have crumbly. This as a sole dividing feature may have changed since I read that and full genomes have been sequenced.
On a social impact level, race still impacts many, if not most, lives. I wonder if there are fewer impacts as more people identify as bi-racial or multi-racial and as many grow up exposed to people of many ethnic groups. Younger people seem less hung up on it.
How to Judge People by What They Look Like is a fantastic book. I recommend it to anyone who enjoyed this article.