THE AI CONUNDRUM

Book supplement · 2023-05-04

Facial Recognition, Privacy and Protecting Children

An archived post from the original book website, presented in its historical context.

"Privacy is not always inherently good for children," notes Sarah Gardner from Thorn. She provides a compelling argument for why privacy is a continuum, and we need to re-consider how we protect children and their rights not to be abused. 

It is an important video to watch and consider how encryption of Facebook messenger and other services represents a trade-off of the safety and protection of children, and protection of the privacy of the children who have been violated as a future adult.

Watch video on YouTube

Read the story here.

She envisions a world where we protect privacy and protect children. 

In our view, artificial intelligence can play a role in independently monitoring the encrypted space with a governance model that keeps it from prying human eyes, unless the AI surfaces child sexual abuse materials (CSAM). It seems a reasonable balance of privacy and safety. 

Thorn is known for their work with Spotlight, an AI that looks at advertisements, which could contain children. They match the images to missing children database, and when the AI thinks it has a match, a the image can be forwarded to parents or guardians to confirm the match. If it is a match, a child has a chance to be reconnected and end the exploitation. When it isn't a match, the parent can say it isn't their child, and continue to look. This is a different risk profile than using AI as the sole basis to make an arrest, which we have noted in the book is a dangerous use of AI. From our perspective, the question is, "what if the AI is wrong?" In the case of Spotlight and much of Thorn's work, there is a human in the loop and the consequence of AI being wrong is minimal.  

To learn more about Thorn, please read Thorn CEO Julie Curdua vision for the year.