A research was carried out on visual artificial intelligence of Microsoft, IBM and Google. Wunderman Thompson’s Data Group showed photos of people wearing protective masks to well-known visual AI systems. The results are slightly different.. So, how does artificial intelligence evaluate masked photos?
Artificial intelligence is still a developing technology.. Some critics state that artificial intelligence is a work that will destroy humanity’s belief in itself. Therefore, sometimes artificial intelligence studies can create a question mark in mind.. Some tech companies are also pulling out of this space. IBM also announced that it was withdrawing from the face recognition business.
With the entry of COVID-19 into our lives, many of our life habits have changed and we have new habits.. Wearing a mask is one of them.. Different results came out in the research on what artificial intelligence evaluates masks, which are a part of our lives recently.. For the research, 256 images of people of all genders with different qualities were used.. These images were tested on Google Cloud Vision, IBM’s Watson Visual Recognition, and Microsoft Azure’s Cognitive Services Computer Vision.
The probability of men being recognized among mask wearers, although no system is perfect at detecting masks. was twice as likely. So while these systems could somehow recognize masked men, what did they think was on women’s faces? How does artificial intelligence evaluate masked photos? Here are the results:
- Google’s AI has determined that 28% of images of women have their mouths covered with tape. 8% identified their masks as facial hair.
- Microsoft Computer Vision thought that 40% of women used fashion accessories in their images. 14% identified the mask on their face as a lipstick.
- IBM’s Watson Visual Recognition produced different results.. She identified 23% of the images of women as wearing gag (gagging). The other 23% were sure that they were wearing chains or that it was a restriction.
The results are somewhat disturbing. Where does artificial intelligence get these ideas from? Researchers have suggested that artificial intelligence machines come up with ideas in this way inspired by the victims of violence or silenced women in the dark areas of the internet.. Although these are not taught individually to artificial intelligence, they can learn them because they have their own learning.
As a result
Artificial intelligence is not always used for good works.. With the development of technology over time, it has become easy to access all kinds of information.. Good or bad, we can reach whatever we want to achieve.. That means it’s up to us what to choose.. In other words, it cannot be claimed that artificial intelligence is completely bad or harmful.
People can use technology for different purposes, just like everything else.. In the past days, a research was carried out on crimes that could have dangerous consequences, in which artificial intelligence was used.. You can find more details on this issue from the news Deepfake Selected as the Most Serious Artificial Intelligence Supported Crime Threat.