Tech giant IBM has released a trove of data that contain 1 million images of faces taken from a Flickr dataset with 100 million photos and videos. The company researchers think these specific data will assist developers in training their Artificial Intelligence-enabled facial recognition systems to recognize faces more fairly and precisely.
IBM head scientist, John Smith stressed the importance of variety in datasets for facial recognition systems to mirror real-world diversity and lessen the rate of mistake in matching a face to a person. Smith said that several high-flying datasets utilized in the field are too narrow and fall short in coverage and balance. The data does not reflect the faces we see in the world. Albeit, Experts have warned on the potential for AI to be biased but, research has illustrated that facial recognition technology is much more adroit at making out the faces of white males than it is with minorities. IBM itself has been criticized over its facial recognition system. In a paper published last year by MIT researcher Joy Buolamwini, where IBM Watson’s visual recognition platform found an almost 35 percent error rate when it came to recognizing darker-skinned females, and a less than 1 percent error rate for lighter-skinned males. Studies like this have amplified the concerns over the use of facial recognition in some sectors like law enforcement and the AI-powered racial profiling.
The facial recognition technology today is testing and utilizing all over the world. For example, the Metropolitan Police in the U.K. is testing facial recognition, while Chinese AI firm SenseTime helps local authorities in spotting crime suspects with the use of its facial recognition system.