The research has been welcomed by major AI companies like Google and Amazon, who are reportedly reviewing the paper. However, the creator of the Monk Skin Tone Scale, a commonly used measure, defended his system, arguing that it does take undertones and hue into account and that its simplicity is a strength, as it avoids the inconsistencies associated with having more categories.
Key takeaways:
- Sony AI has published research suggesting that AI algorithms should consider red and yellow skin hues in addition to lightness or darkness of skin tones to reduce bias.
- Current scales, such as the Monk Skin Tone Scale and the Fitzpatrick scale, primarily focus on the lightness or darkness of skin tone, potentially missing biases against certain ethnic groups.
- Sony's research found that common image datasets overrepresent people with skin that’s lighter and redder in color, and underrepresent darker, yellower skin, leading to inaccuracies in AI systems.
- Sony proposes an automated approach based on the CIELAB color standard, which would also eschew the manual categorization approach used with the Monk scale. Google and Amazon are reportedly reviewing the paper.