“There’s so much passion for using AI in ways that are beneficial,” said Anna Bethke, Head of Intel’s AI4 Social Good project at Intel’s AI Products Group, in our latest episode of The New Stack Makers podcast.
Bethke’s team partners with companies and social justice organizations to give them access to Intel’s knowledge of deep learning skills and techniques, making it easier for them to mine social media and other data to filter out the noise.
One of her first partnerships is with the National Center for Missing and Exploited Children (NCMEC). Its mission is to find children who are being trafficked or in danger of being kidnapped and alert the local law enforcement so they can take action.
NCMEC gets hundreds of thousands of alerts a day from social media like Facebook and Twitter, which was completely unmanageable. Leveraging the expertise of the people at NCMEC, Bethke built algorithms using keywords and phrases to flag possible social media posts, identify where the post originated, and prioritize the danger level. The data from the social media sites is set to cull every possibility and AI is the perfect tool for separating an awkward social post from a real threat.
There are three points of information NCMEC needs. First, it needs to identify and prioritize which posts represent the most danger to the child. Second, it needs to identify where the post is originating, which is often complicated by the use of multiple IP addresses and IP address obscuring. Once the location is found, the legal jurisdiction needs to be identified so the information can be sent to the proper authorities (e.g., police or sheriff department). But how do you find that information in the flood of data?
Enter the AI4 Social Good team. It built an on-premises, scaleable AI system that is completely self-contained. The data is so sensitive, that none of it is in the cloud or was even transferred to any Intel system for testing.
The project also built-in ways to search the dark web, which is where a lot of the child trafficking takes place. It added the dark web data, along with the data gathered from social media into NCMEC’s AI system.
Another issue is determining the age of the person in the photo, which has a lot to do with what actions law enforcement can take. Facial recognition, especially on a global scale is still in its infancy and Bethke noted that most of the AI facial recognition training programs only have white adult males in their training guides. This part of the program is evolving and combines facial recognition with the text of the post to determine a possible age.
In this process, Bethke wanted to make it easier for others at Intel and beyond to get involved. Her AI Developer Program is now so popular that projects are happening without her even knowing it. She recently heard of an Intel developer who wrote a phone app to identify Tibetan currency so that blind people can know how much money they have.
Listen in to hear more details of NCMEC project and other programs the AI4 Social Good has in progress.
In this Edition:
- 3:37: Working with the National Center for Missing and Exploited Children and how technology and data is used to aid in combatting child trafficking.
- 6:43: Working with Facebook
- 15:00: Exploring a good use case of police body cameras that the Indian police force was able to find hundreds of reported missing children in different orphanages due to facial recognition.
- 17:08: Other projects.
- 19:00: How many projects AI4 Social Good is working on.
- 23:42: How to get involved.