Ethics of data privacy

- [Instructor] We are back on the Azure AI demo screen. The ethics in the world of H-AI are constantly evolving. We touched on human-centered AI and the data sets we use to build our autonomous vehicles. Now let's take a look at what building with data privacy in mind might look like in H-AI. Instead of physically walking through an aisle and looking for shipments, which can be time consuming as well as dangerous, depending on the size of the products, warehouse robots are there to safely take our place. They can map every aisle and shelf, but they also capture people who are working and moving. This is the reason it's highly critical that data privacy is at the forefront of our thoughts when we are using H-AI. Let's look at an example. You've seen this picture before. It's a picture in a warehouse, and we are doing object detection using Azure AI, and it has put bounding boxes and identified four people in here. We discussed about degree of confidence. That's not what this demo is about. But look at this. We see Azure saying it's picked up four people, but, it's capturing people's faces, features. It shows what they're wearing. It shows headgear such as helmet. It is showing a robotic arm at the back. I'm not worried about the object here. I'm thinking about the people and there's lot of wires and cables and one person is at the laptop. There's expressions on each of these faces, which looks different. They all seem to be looking at the same thing. So there is so much information of people. So when an autonomous agentic AI or even a simple object detection model looks at a scenery or an environment, it sees a person. That's what it is saying it is seeing a person, but it is seeing more than just the person because it is able to see the expressions and their outfit and a lot more information. In some cases, we zoom in and we can even see the name tag of that person. So it is important for us to understand what we need to white out if we are giving a picture. In some data sets, when you go to download, especially from Europe where they follow GDPR very, very diligently, you can see the faces and expression of people are white out because we can do facial recognition and identify these people, and that is against GDPR compliance. So I want you to think about ethics as the right thing to do, but it's also about following the data privacy laws of the state or country and for GDPR compliance. So if this is a real person and you're saving their picture and this is going to the cloud and you're saying, "Hey, I'm just going to let the AI identify pictures of people in here", then you'll have to think about the GDPR compliance of permission of using their faces and storing them. And again, you have to decide are you using that in the cloud or in the Edge? If this is a face and a person's features that are stored right on the Edge and Edge AI is running inference and making decisions and moving on, or is this information stored? Think about this. If this image is just being stored as data and it is on the Edge and running inference and being able to make some predictions, you might think this is an H-AI problem in some workflow, but, if this data is stored with the timestamp, it is informing who was here, what were they looking at, whether it is... It could be a company offsite. It could be something else. So there's a lot more information that goes with the location and timestamp associated with the person that you are not even thinking as H-AI use case because that's not the use case. So those are the things you should be thinking about. What are you storing? What is the data privacy of it? And what other information is that data or that AI inference results carrying in it that you might want to be thinking about for data privacy and also for ethics compliance? We'll do a demo challenge next, and I want you to think about this and decide, okay, what kind of things are in this image and what am I omitting? What am I keeping? Let's do a fun challenge next.