Mori Magazine

The Power and Limits of AI in Historical Photo Analysis  

Family history research often relies on old photographs to make connections between ancestors. New AI technologies show promise in unlocking more insights from these old images, but also have some key limitations. In this post, we'll explore what AI can currently do with historical photos and where there is still room for improvement.

One exciting capability is using AI facial recognition to identify ancestors in unlabelled photographs. By comparing facial features across a dataset of images, algorithms can spot similarities and suggest possible matches. This can help researchers confirm the identity of ancestors and make new family connections. However, accuracy depends greatly on having enough photos of each individual for comparison. Images from certain eras and cultures remain a challenge.

AI programs are also now able to convincingly colorize black and white photos. This can bring new life and detail to faded old family portraits. However, current AI colorization mainly applies generic colors without incorporating historical context. It may get skin tones or clothing hues wrong in ways that seem unrealistic. Better training data and techniques are needed to colorize photos accurately based on time period and location.

In summary, AI has made impressive steps in unlocking family history insights from visual sources. But human expertise is still essential for making nuanced connections and checking AI suggestions against other evidence. As the technology improves, AI will become an increasingly powerful assistant for ancestors research, while likely still requiring human supervision for the best results.