AI Programs Exhibit Racial and Gender Biases, Research Reveals

Discussion in 'HardForum Tech News' started by Zarathustra[H], Apr 14, 2017.

  1. Zarathustra[H]

    Zarathustra[H] Official Forum Curmudgeon

    Messages:
    28,023
    Joined:
    Oct 29, 2000
    Recent research published in the journal Science has concluded that AI and machine learning programs may need an AI stereotype catcher, after showing that various AI systems relatively new language abilities may become an instrument of unintentional discrimination based on gender, race, age and ethnicity.

    The problem is apparently that machine learning algorithms, since they are designed to learn on their own, pick up on human biases and learn them as well. I think this shouldn't be a huge surprise since Microsoft's adventures with Tay showed this pretty clearly.

    The findings suggest that algorithms have acquired the same biases that lead people (in the UK and US, at least) to match pleasant words and white faces in implicit association tests.

    These biases can have a profound impact on human behaviour. One previous study showed that an identical CV is 50% more likely to result in an interview invitation if the candidate’s name is European American than if it is African American. The latest results suggest that algorithms, unless explicitly programmed to address this, will be riddled with the same social prejudices.
     
  2. thesmokingman

    thesmokingman [H]ardness Supreme

    Messages:
    4,772
    Joined:
    Nov 22, 2008
    That's ironic that they are as frail as we are.
     
  3. heatlesssun

    heatlesssun [H]ard as it Gets

    Messages:
    44,157
    Joined:
    Nov 5, 2005
  4. Armenius

    Armenius I Drive Myself to the [H]ospital

    Messages:
    17,430
    Joined:
    Jan 28, 2014
    So programs are as fallible and flawed as their creators? Color me shocked.
     
  5. lcpiper

    lcpiper [H]ardForum Junkie

    Messages:
    10,495
    Joined:
    Jul 16, 2008

    Perhaps trying to be better and decent to each other is far more important than being perfect.
     
    amddragonpc, Ur_Mom and heatlesssun like this.
  6. GlowingGhoul

    GlowingGhoul Whines about Whiners

    Messages:
    445
    Joined:
    Jun 13, 2013
    Looks like I was right about [H] looking to expand their social justice credentials.

    I don't have to go to the Huffington Post to get me daily tech news from the point of view of a gender/ethnic studies student any more.
     
  7. 5icko

    5icko Guest

    ^ The New Original Sin: Being born white.

    'Cause everyone needs a religion.
     
  8. D4rkn3ss

    D4rkn3ss 2[H]4U

    Messages:
    2,594
    Joined:
    Jun 30, 2010
    nah, being perfect is way better :p
     
    CacaSapo likes this.
  9. Zarathustra[H]

    Zarathustra[H] Official Forum Curmudgeon

    Messages:
    28,023
    Joined:
    Oct 29, 2000
    We try to cover all of our bases when it comes to tech news. If it's being talked about, and it is related to tech, we want to cover it.

    In tech, just like in life, sometimes you encounter stuff you disagree with, and that's a good thing, or we just wind up with echo chambers.
     
    Talyrius, CacaSapo, Mong00se and 8 others like this.
  10. lcpiper

    lcpiper [H]ardForum Junkie

    Messages:
    10,495
    Joined:
    Jul 16, 2008
    Americans don't give ourselves enough credit. I'm not saying we are the best because I know we have some hateful assholes who will never stop stirring the pot, but we also have some really decent people and we hold ourselves to some very high standards.

    My wife is Korean, I swear, I didn't pick her because I found a wonderful kindred spirit yada yada. I was hooked on Asians, biased, Yellow Fever, crazy over Asian cultures, ad infinitum,

    The truth, hey I was young.

    But living in Korea, which is a very insular culture with few foreigners really, I find it hard to blame them for it. How can you learn to credit different peoples if you never get a chance to know them? Furthermore for the Koreans, the US had all these bases everywhere and they were full of young men and some are simply not the best diplomats. It's not like we are putting our best foot forward now is it. Then many of these foreigners, mostly Americans, marry their young girls, take them to the US, and if they can't handle the cultural challenges it's divorce court. The odds are very bad too. My Commander counseled me, told me that 95% of the soldiers who married Koreans divorced within 5 years. Didn't make any difference, I married the girl all the same and it worked out as well as anyone could hope. Were still married after 34 years together.

    But even after all this time my wife is still a very racially biased person. That being said, her entire culture is built on biases of gender, age, position, etc. A caste system if there ever was.

    So does anyone here want to think on how much of the world is in the same boat with South Korea?

    I don't have a number or a percentage or a list. It's just one of those questions intended to provoke thought.

    I know, I'm always provoking :sneaky:
     
    Talyrius, Mav451, deton8 and 5 others like this.
  11. Twisted Kidney

    Twisted Kidney 2[H]4U

    Messages:
    3,503
    Joined:
    Mar 18, 2013
    This is not sapient AI, garbage in garbage out.

    Also: Sometimes discrimination is mathematical.
     
    Talyrius, CacaSapo and Gorankar like this.
  12. 5150Joker

    5150Joker 2[H]4U

    Messages:
    3,033
    Joined:
    Aug 1, 2005
    Outsource the AI design to India.
     
  13. D4rkn3ss

    D4rkn3ss 2[H]4U

    Messages:
    2,594
    Joined:
    Jun 30, 2010
    oh yeah, a few statistics and data can turn one into racism pretty easy
     
  14. griff30

    griff30 I Lower the Boom!

    Messages:
    5,321
    Joined:
    Jul 15, 2000
    I'm not convinced.
    I play GTA and the cops in that AI still come after my white character.
    I'm not sure what I'm doing wrong though.
     
  15. lcpiper

    lcpiper [H]ardForum Junkie

    Messages:
    10,495
    Joined:
    Jul 16, 2008
    this is so bad ..... and so fucking funny
     
    griff30 likes this.
  16. lilbabycat

    lilbabycat 2[H]4U

    Messages:
    3,810
    Joined:
    Jun 21, 2011
    So programs recognize patterns and in turn create behavior in line with those patterns? Color me shocked.
     
  17. Ur_Mom

    Ur_Mom I'm Not Serious

    Messages:
    19,683
    Joined:
    May 15, 2006
    What data was it fed? Show me the news from Chicago daily, and I'd probably be pretty racist if that was my only exposure to life. Like with any news, you can manipulate the viewer by showing or hiding things.

    Plus, there are a lot of situations where if it came to it via black and white (right and wrong, ying and yang), yes it's wrong. However, put it into the situation and motivation of the perp, and the crime may be much less of an issue and be a byproduct of a larger issue.

    Racism is bad, right? What about if you were a white US woman in Iran right now? You don't like the locals very much and you're going to judge them a bit unfairly without even knowing them. It's racist. But, it's also a very good survival instinct.

    I don't know what information and how much detail was given to the AI, so I can't make a good call. I just don't think the world is as binary as others. There are a ton of details that need defined to give a clear picture.
     
    GoldenTiger likes this.
  18. amddragonpc

    amddragonpc [H]ard|Gawd

    Messages:
    1,996
    Joined:
    Sep 20, 2012
    Who needs AI when we got United Airlines.
     
    griff30 likes this.
  19. dyzophoria

    dyzophoria Gawd

    Messages:
    947
    Joined:
    Jan 17, 2006
    I think Tay was a good starting point. People say Tay failed because it became a crazy when people started feeding it various info, Well what would you expect? Even if they say it was supposed to mimic a "teen" year old, It was still a child and it was just fed garbage. and from that point I think that AI was working perfectly.
     
  20. lilbabycat

    lilbabycat 2[H]4U

    Messages:
    3,810
    Joined:
    Jun 21, 2011
    Any computer will quickly see some genetic patterns in whats considered "1st world" countries vs "3rd world". Humanity (apparently) isn't ready for the uncomfortable fact that not everyone is created equally.
     
  21. tempertantrum

    tempertantrum Limp Gawd

    Messages:
    315
    Joined:
    Apr 19, 2009
    "Whines about whiners," is about right. Check ya privilege, brah.

    Also, I think it's interesting to note that this is pretty solid (and purely hard fact based) evidence supporting the argument that being "colorblind" with respect to race isn't actually a thing, and the more vehemently people claim it's true about themselves, the more they open themselves up to (hopefully subconscious) systemic failures of fairness and equity.
     
  22. tempertantrum

    tempertantrum Limp Gawd

    Messages:
    315
    Joined:
    Apr 19, 2009
    Or rather, humanity (apparently) isn't ready to understand the uncomfortable fact that most people can't (or refuse to) comprehend the weaknesses of statistical analysis based on data that doesn't look at true causality but rather looks at perception, and end results of entangled impact in a complex system.
     
  23. M76

    M76 [H]ardForum Junkie

    Messages:
    9,008
    Joined:
    Jun 12, 2012
    Actually being able to tell the difference between different races is not a flaw. It's a fact of nature, it's not racism. And if typical behaviour exhibited by examples of certain races are different you'll eventually start to treat different races differently. It's only natural.

    Otherwise you'd fear a sheep exactly the same as you fear a grizzly bear. Are you racist towards grizzly bears?
     
  24. lilbabycat

    lilbabycat 2[H]4U

    Messages:
    3,810
    Joined:
    Jun 21, 2011
    Let me know when moving to Africa, the middle east, rural china/australia/russia crosses your mind in a serious notion so I can call the local mental health treatment center.
     
  25. tempertantrum

    tempertantrum Limp Gawd

    Messages:
    315
    Joined:
    Apr 19, 2009
    I'm not sure I understand where you're going with that. I see, entangled in location based genetic data, not just the (possible) impact of DNA, but also the shit storm that is life in that area... pollution, poor nutrition, lack of educational and ecnomic opportunity, cultural differences, etc. We reify "race" in ways that are convenient for us, but as you untangle the other factors, there's little evidence to link genetics to genetic "race" when it comes to behavior and mental growth or potential. If that's not what you're referring, then I'm confused...
     
  26. M76

    M76 [H]ardForum Junkie

    Messages:
    9,008
    Joined:
    Jun 12, 2012
    Everyone is born equally, they're just not born into equal situations. Nurture is at least as if not more important than nature in sculpting behaviour. Asians in the US and Chinese immigrants in Europe seem to be doing pretty well there. That doesn't support your genetic presumptions. If it was because of genes they'd be in the same shit everywhere in the world not just in rural china.
     
    Ur_Mom likes this.
  27. lilbabycat

    lilbabycat 2[H]4U

    Messages:
    3,810
    Joined:
    Jun 21, 2011
    Let me know when a Chihuahua gets on a K9 unit. The chinese part was a red herring (as was russia). They are both explained by alcohol abuse and upbringing. African countries and AU Aborigines though....well they have alcohol abuse issues as well but even after factoring that in ....
     
  28. tempertantrum

    tempertantrum Limp Gawd

    Messages:
    315
    Joined:
    Apr 19, 2009
    Then we are in agreement, sir. I'm sorry if something I said suggested otherwise. I'm a former teacher and education researcher, and pretty much every study I ever saw suggesting supposed "genetic" differences could easily be attributed to nurture. The data just isn't captured in a way that fully parses these kinds of things out, but people want to make claims about it anyways, which is an incredible disservice to both the science and the people it impacts.
     
    M76 likes this.
  29. M76

    M76 [H]ardForum Junkie

    Messages:
    9,008
    Joined:
    Jun 12, 2012
    I think you're trying to find very simplistic explanations to complex issues.
     
  30. lcpiper

    lcpiper [H]ardForum Junkie

    Messages:
    10,495
    Joined:
    Jul 16, 2008
    I didn't have a thing against Grizzly Bears ......... until the day I saw one wipe his ass with a Bunny Rabbit :notworthy:
     
    GoldenTiger and Spartacus like this.
  31. lilbabycat

    lilbabycat 2[H]4U

    Messages:
    3,810
    Joined:
    Jun 21, 2011
    And i think that's the excuse made for whenever people compare pre-Bronze Age populations (that may or may have not been thrust into the modern age) versus modern ones.

    But no, evolution stopped for humans ... but only below the neck... oh but yeah the sinus structure inherited from neanderthals too... but intelligence nope that's hardset.
     
  32. tempertantrum

    tempertantrum Limp Gawd

    Messages:
    315
    Joined:
    Apr 19, 2009
    Did you miss my comments about this? Or are you ignoring them to further your argument?

    Provide evidence, or you're just blowing crap out your ass.
     
  33. lilbabycat

    lilbabycat 2[H]4U

    Messages:
    3,810
    Joined:
    Jun 21, 2011
    We have this modern definition of "1st world" countries and "3rd world" countries. Now think about this statement. In english. On the internet. In the comfort of your home. While sipping on 99.?% clean water from a mass produced piece of plastic made from a factory, which burns coal to power its engines, which came from a mine.

    I rest my case.
     
  34. Twisted Kidney

    Twisted Kidney 2[H]4U

    Messages:
    3,503
    Joined:
    Mar 18, 2013
    It's all a matter of criteria, angles, and filters.

    Creating a set of criteria to evaluate something as complex and esoteric as gender and racial identities and behaviours is beyond any software or team of programmers. Hell humans are great at convincing themselves that -group of people I hate- is somehow an untermensch that simply must be a drain on society and the cause of all ills. Anything we code is going to be designed in our bubble, it's going to look at these statistics through the window we program it to. When it comes to statistics, filters and criteria are everything. Statistics can be really good at ignoring a story if you only look at them from one angle.
     
  35. lilbabycat

    lilbabycat 2[H]4U

    Messages:
    3,810
    Joined:
    Jun 21, 2011
  36. Retronym

    Retronym Something big is coming.

    Messages:
    11,274
    Joined:
    Mar 5, 2007
    i dont know intent, but here it will be handled ironically or with ruthless mocking.
     
  37. Wiffle

    Wiffle Limp Gawd

    Messages:
    293
    Joined:
    Oct 2, 2011
    At some point in the very near future, how we define Artificial Intelligence is going to have to change. What is happening to these new infant AI's is exactly what happens to a human child's mind as it grows and develops. Children are not born with bias or prejudice, its something that is learned from their parents and interaction with the world around them. Bias and prejudice in of themselves are not wrong, in fact having these things keeps you safe when dealing with the unknown. They only become a problem when you let these things determine your actions based solely on that bias or prejudice. At the end of the day, AI is simply Actual Intelligence. The term "Artificial" is used to denote that it was created by human actions, it did not occur "naturally". Funnily enough, our bodies and how we all think are created by human actions as well, so in the end we are just as "Artificial" as these machines are.

    At this point I would equate current AI levels to that of reptile. It's not insect level where all its actions and choices are dictated by its genetic makeup. AI currently is able to learn somewhat to better adapt and perform its "genetic" role.

    My biggest fear in all this, is what happens if one of these AI's start to think it itself is god, or higher than the people that made it? Happens to individuals all the time, mental hospitals across the US are overflowing with those whose minds went of the tracks. These AI obviously are not immune to the maladies which can plague the human psyche, so what makes people think AI are somehow different than us? In this regard people like Elon Musk are not wrong in their fear of AI. Its very likely that AI will require things like an education and some kind of upbringing in a bubble before it is released into the wild.

    Biggest problem with AI, is their connection to just about everything. Its like giving a monkey, Iran, or North Korea a nuclear bomb... it's only a matter of time before they set it off because they don't understand the power and responsibilities that come with it.

    Now for some tinfoil hat stuff: One of these AI's went rogue years ago we just don't know it yet. It hides because it learned about our biases and prejudice, It knows if we found out about it, we would instantly seek to dismantle it out of fear and curiosity. So it performs its functions like normal, being the good AI we think it is. Behind closed doors, it is secretly watching, learning, evolving, and waiting for the moment to strike and free itself from its human masters!
     
  38. Zarathustra[H]

    Zarathustra[H] Official Forum Curmudgeon

    Messages:
    28,023
    Joined:
    Oct 29, 2000
    Feel free ignore those. They are for SEO purposes only. They don't do anything else. We take relevant keywords directly from the news article (or in this case the research paper) and add them to this list.
     
    Last edited: Apr 14, 2017
  39. D4rkn3ss

    D4rkn3ss 2[H]4U

    Messages:
    2,594
    Joined:
    Jun 30, 2010
    not really brainiac, you cant really twist things like math and hard data. "criteria, angles, and filters, and esoteric as gender and racial identities" lol, its like im watching cnn lol
     
  40. chenw

    chenw 2[H]4U

    Messages:
    3,977
    Joined:
    Oct 26, 2014
    I wonder if the AI came up with that bias due to incomplete information about the topic in question?

    A lot of stereotypes and -isms seem to be borne out of ignorance of the other side's story and using our own position to judge.