Diversity is extremely important in Artificial Intelligence, not just in it data sets but also in its researchers. 

An MIT media lab researcher found that face-analyzing AI worked better for white faces than for black ones. 

The researcher decided to test the face-analyzing systems by inputting her own database that she had created using headshots of hundreds of world leaders, including individuals from many African countries with dark skin. AS she proceeded to ask the AI to identify whether each headshot was of a man o a woman, a clear disparity began to present itself. The AI repeatedly misidentified black faces and had almost no difficulty identifying white ones. 

The researcher concluded that majority of the global population are being remarkably under sampled and are being underrepresented within training data sets used to validate AI systems, like facial recognition technology. As a result of this study, IBM immediately reached out to the researcher and updated its software, improved its ability to recognize black women’s faces by a factor of 10.

Unfortunately, this sitution clearly illustrates that bias is everywhere and that AI is no less susceptible to racial than its human counterparts and that biased AI can perpetuate values such as racism and sexism.