AI can predict who will be criminals based on facial features

[SurvivorNVL]

It'd be better to just document everyone in a country based on their IQ's, and then be capable of determining likelihood of criminality from there.  Generally the majority of crimes are committed by people with an average of 84 or lower.  Majority of serial killers are between 116-124 points and have chemical imbalances resulting in copper deficiency or heightened copper/iron counts.  They could likely get a better prediction based on peoples chemistry and IQ ratios rather than ctOS3.0.

[Bananasplit_00]

Psyco Pass much? thats litteraly all i can think of here... also if you havent already go watch Psyco Pass, its not for everyone but it has some shit that makes you think

[Bananasplit_00]

4 hours ago, mvitkun said:

Psycho-Pass IRL

my thoughts EXACTLY

[Bananasplit_00]

8 minutes ago, SurvivorNVL said:

It'd be better to just document everyone in a country based on their IQ's, and then be capable of determining likelihood of criminality from there.  Generally the majority of crimes are committed by people with an average of 84 or lower.  Majority of serial killers are between 116-124 points and have chemical imbalances resulting in copper deficiency or heightened copper/iron counts.  They could likely get a better prediction based on peoples chemistry and IQ ratios rather than ctOS3.0.

that + this AI stuff is basically Psyco Pass... i dont want Psyco Pass to become reality because 1: that shit is scary 2: i would be super screwed basically the second the system would go online lol

[SurvivorNVL]

Just now, Bananasplit_00 said:

that + this AI stuff is basically Psyco Pass... i dont want Psyco Pass to become reality because 1: that shit is scary 2: i would be super screwed basically the second the system would go online lol

Oh, in the wrong hands this kind of system would probably nuke me off the map for my political, economic, and social beliefs.

[Bananasplit_00]

1 minute ago, SurvivorNVL said:

Oh, in the wrong hands this kind of system would probably nuke me off the map for my political, economic, and social beliefs.

yah this is quite litteraly no proper freedom anymore, you cant think what you want, you cant say what you want, you cant rise against an opressive state by civil disobedience seeing as you would get plucked of by the system before you can act....

[SurvivorNVL]

Just now, Bananasplit_00 said:

yah this is quite litteraly no proper freedom anymore, you cant think what you want, you cant say what you want, you cant rise against an opressive state by civil disobedience seeing as you would get plucked of by the system before you can act....

Though if you work for the system, it'd be a fun day.  

[sazrocks]

inb4 Captain America The Winter Soldier senario.

[Bananasplit_00]

Just now, SurvivorNVL said:

Though if you work for the system, it'd be a fun day.  

the one who controlls the system has imense control, they can do whatever they want, add exeptions for certain people ect... this is a system that should NEVER be made, its fun to speculate but this is something that NEVER should exist

[DrMikeNZ]

25 minutes ago, Ryan_Vickers said:

unless it actually turns out to be reliable, in which case, that would be very interesting and would have implications that warrant further investigation  

It would not be possible to determine if it is reliable without determining if each individual on the planet is a criminal or not. And if you can do that, you no longer need the algorithm.

 

17 minutes ago, SurvivorNVL said:

It'd be better to just document everyone in a country based on their IQ's, and then be capable of determining likelihood of criminality from there.  Generally the majority of crimes are committed by people with an average of 84 or lower.  Majority of serial killers are between 116-124 points and have chemical imbalances resulting in copper deficiency or heightened copper/iron counts.  They could likely get a better prediction based on peoples chemistry and IQ ratios rather than ctOS3.0.

IQ tests are inaccurate and useless. Even if a majority of people who commit crimes fell into certain IQ categories it does not make the opposite true.

[Bensemus]

lol screw Skynet, theses people are on their way to creating the Sibyl System. 

[Energycore]

There is correlation between these, sure, but that doesn't mean that suddenly every person with those facial characteristics is a (potential) criminal. In fact assuming so would fill our jails with innocents.

[goodtofufriday]

[Sky Daddy]

I reverse image searched myself and i get 99% mugshots...

[SeraphicWings]

Hey, this is a bit too soon for Psycho Pass to be real 

[NumLock21]

Minority report gone wrong

[vanished]

1 hour ago, DrMikeNZ said:

It would not be possible to determine if it is reliable without determining if each individual on the planet is a criminal or not. And if you can do that, you no longer need the algorithm.

Well, that would be needed to verify if it was 100% reliable, but I suspect that who ever is planning on using this is comfortable with a lower threshold, and so if they can, for example, get 98% reliability on a sample size of 10,000 or something, that would probably be fine.

 

Is that OK?  Absolutely not, I think it would be insane.  I just think there's probably some government out there crazy enough to do it though...

[Wolther]

Justice departments no longer on defense, but now on offense.. Interesting.

[patrickjp93]

Okay, until we see the implementation, we cannot say there is a bias built in by the programmer. We have generalized pattern seeking and pattern matching formulae. If you train an algorithm based on a non-biased, large sample of criminals with an even distribution on race, then the AI will be objective, because the algorithm is objective.

 

TLDR: the article trying to discredit this doesn't actually discredit the technique. It just makes a shallow attempt to disparage it.

[handymanshandle]

Sure a computer may not have emotions of be able to have bias but programmers can/do have those things.

So I mean the program could be made to take certain features commonly found in a specific race and say that they are criminals.

[patrickjp93]

1 minute ago, wcreek said:

Sure a computer may not have emotions of be able to have bias but programmers can/do have those things.

So I mean the program could be made to take certain features commonly found in a specific race and say that they are criminals.

I doubt that's the implementation. It's easier to write a naive, generalized pattern former and matcher than it is to inject preconceived notions.

[DrMikeNZ]

21 minutes ago, Ryan_Vickers said:

Well, that would be needed to verify if it was 100% reliable, but I suspect that who ever is planning on using this is comfortable with a lower threshold, and so if they can, for example, get 98% reliability on a sample size of 10,000 or something, that would probably be fine.

 

Is that OK?  Absolutely not, I think it would be insane.  I just think there's probably some government out there crazy enough to do it though...

2% false positive rate would result in millions of people being persecuted.

While individuals can be compiled together to obtain statistical information on a group, statistics cannot be used to assess or predict individuals. Such a system could never be used to determine criminality.

 

Definitely insane. Nearly as insane as drawing a correlation between childhood consumption of confectionery and convictions of violent crimes as an adult.

[vanished]

16 minutes ago, DrMikeNZ said:

2% false positive rate would result in millions of people being persecuted.

While individuals can be compiled together to obtain statistical information on a group, statistics cannot be used to assess or predict individuals. Such a system could never be used to determine criminality.

 

Definitely insane. Nearly as insane as drawing a correlation between childhood consumption of confectionery and convictions of violent crimes as an adult.

I just pulled those numbers out of the air so don't worry too much about that

All I'm saying is, if they can get it close to 100%, which (if there is a correlation to be found) is possible (in theory), then it wouldn't take a sample the size of the world to do it, and some government might just implement this as a tool.  Scary notion, imo...

[patrickjp93]

19 minutes ago, DrMikeNZ said:

2% false positive rate would result in millions of people being persecuted.

While individuals can be compiled together to obtain statistical information on a group, statistics cannot be used to assess or predict individuals. Such a system could never be used to determine criminality.

 

Definitely insane. Nearly as insane as drawing a correlation between childhood consumption of confectionery and convictions of violent crimes as an adult.

I don't think it would lead to persecution so much as it would just make surveillance more sensitive and targeted. Obviously, no one should be harassed or accused until they actually commit a crime,  but you can't have good proactivity in law enforcement without some guidelines.

[Arty]

2 hours ago, SurvivorNVL said:

It'd be better to just document everyone in a country based on their IQ's, and then be capable of determining likelihood of criminality from there.  Generally the majority of crimes are committed by people with an average of 84 or lower.  Majority of serial killers are between 116-124 points and have chemical imbalances resulting in copper deficiency or heightened copper/iron counts.  They could likely get a better prediction based on peoples chemistry and IQ ratios rather than ctOS3.0.

Tbh most under IQ 90 shouldn't be allowed procreate....but anyyyyywayyyyyyy, its hard to say someone iq of 116-124 is a killer cause they could just be smart.