Facial recognition technology is improving--algorithms developed by IBM and Microsoft can identify the gender of a white man 99% of the time. However, it's only 93% accurate for white females, and 65% accurate when it comes to darker-skinned females. Because artificial intelligence is only as smart as the data used to train it, it will inevitably be better at detecting white males if they are a majority of the data set. Microsoft and IBM have planned to improve the accuracy of facial recognition, investing in research to "understand and remove bias." Academic and public policy groups are concerned about the impact of AI; technology could be a "platform for opportunity," but not if it repeated the discrimination of the past.
How important are the ethics behind facial recognition technology and other artificial intelligence? Should there be more regulation of racial/gender equality when it comes to the digital world?
Source:
NY Times
NY Times
12 comments:
I think that there should be no regulation of the progress of facial recognition. This would only slow down the progress of this because you will pretty much be saying you have to develop this technology in this specific way and according to this specific schedule which will in the end slow down the development of this technology. Forcing scientists to focus on certain types of people restricts their progress and makes it so the government is essentially running the project instead of the people who actually know what theyre doing. This is yet another example of leftists taking race relations too far. They nit pick at every little inequality and make it a huge deal, even when almost every time there is another logical explanation which ends up having nothing to do with racism.
I think that facial recognition is improving and that is good. I think that now that it's better for a majority of white males they are going to start focusing on broader parts of it such as females, and other races. I think they just need time to improve and perfect it. I doubt they are doing this against anyone or to be racist and I don't think there is a need for regulations because they probably just need more time and will get to other races eventually.
I applaud the creators of facial recognition for creating such a huge advancement in technology. Although, I do believe it is important to make facial recognition closely accurate for ALL skin types and skin colors. I understand that this is a fairly new aspect of technology and therefore takes time to develop but overall, it must be inclusive for all people, not just white males.
I think this is an example of a very interesting new issue: algorithmic accountability. I admit that the sizable error rate for certain demographics, mainly ethnic minorities, is rather disquieting. Any program that has significant errors for substantial demographics of the population is flawed and ought to be fixed; I don't think this a matter of leftist race related fear mongering, but a simple matter of efficacy. That being said, I am not sure the government really needs to be involved, seeing how these errors were not malicious but simply errors. In an increasingly global market, businesses are thoroughly incentivized to have products that work for citizens across the world. If your facial recognition software has high error rates for Chinese people, you are losing out on a significant market and will work to fix these errors.
I also think that this issue with facial recognition is more about errors with the programs that intentional discrimination, as the technology is still a work-in-progress. As such, I think that with this issue, companies should continue to improve the accuracy of facial recognition, while keeping the idea of inclusivity in mind. And based on the article, it seems like companies like Microsoft and IBM are already working to remove bias from their programs.
I agree with John and Julia. Just because some software that a company made works better on white males does not mean they are discriminating or in need of government regulation. Who cares honestly. If anything they are discriminating against white males because facial recognition is most likely going to see use in criminal investigations so more white males will be incriminated. But as for the state of the software now, I would like to build off of Julia's point. When developing software like this, you can either try to work on all cases at once until each is perfected or do what this group is doing which is trying to perfect one case then they can use this knowledge to account for all other cases. Obviously the second method is much more practical and efficient. The problem here is the people who think anything unequal involving race is really bad and needs to be regulated.
We are in almost the beta stages of facial recognition. I do not believe their is a hidden agenda behind the coders, and this should be chalked up to as a work in progress. We will eventually build on this technology and it shouldn't be heavily regulated.
Definitely agree with you guys that facial recognition is still a work in progress, being one of the newer developments of technology. It's really likely that the discrimination by artificial intelligence wasn't intentional, and the accuracy will only continue to improve. But perhaps this could become a bigger issue in the future?
I think facial recognition is an impressive advancement in techonology. Though this is an issue that minority recognition and differentiation is not as accurate as white people -- my sister is able to unlock my phone even though we don't really look alike -- I think it's still great that they were able to ensure high accuracy for one group first. They are still in the process of improving the technology so I see it as a small step at a time. As they continue to advance, I'm sure that accuracy for minority groups will be greatly improved as well.
Like many comments above suggested, I don't think they ever intended on discriminating against minorities. They are, like Ashley suggested, taking small steps. Maybe facial recognition doesn't work the way we expect it to at the moment, but I have a lot of hope that in the future we'll be able to have technology that is inclusive of everyone. As we move forward, innovating and creating new systems, we do have to consider the ethics side of it to avoid making new problems and greater inequalities involving an ever-evolving technological landscape.
This is a huge step into the world of technology. Technicians have made facial recognition into reality and they should be applauded for making such an advancement in technology. It is definitely still a work in progress and is working towards a more complex software. I agree that they weren't trying to discriminate against minorities, it may just be a problem of accuracy in the software. I think as companies continue to improve this software, it'll become more accurate. It's new technology, so I know it takes time to develop, so I believe they'll be making it so that everyone of all skin tones and genders will be equally recognized in the digital world.
I don't believe that the issue of the current facial recognition system is on the ethics, but rather on the diversity of the subjects being tested and the overall accuracy of its current form. In order to solve the problem, they should just use more minorities and test subjects of more diverse appearances in order to improve the overall accuracy of the technology so that it works properly.
Post a Comment