Q: Name three applications of Computer Vision you can recognize on your phone.
A: Face unlock, Google Lens, and Snapchat filters.
Q: Take a photo of your desk and ask ChatGPT to identify or name all the objects.
A: When I uploaded a photo of my desk, ChatGPT guessed all the items perfectly. It even mentioned the brands of the items like my mouse and even my laptop.
Q: How many steps are involved in this identification process?
A: Usually three steps:
- The image is broken into pixels.
- The computer finds patterns and features like edges or shapes.
- It compares them to what it already knows and names the object.
Q: Can you identify any challenges for ChatGPT?
A: Yes, it can make mistakes if the lighting is poor, objects overlap, or items look similar. For example, it might confuse a phone with a calculator or not see something that is half hidden.
Q: Ask ChatGPT to highlight or circle a specific item on your desk. Describe the results and possible challenges.
A: When I ask it to circle something, it tries to, but sometimes it doesn’t mark the exact spot or may circle the wrong item. It’s hard for it if there are many similar-looking things or shadows.
Q: Several applications involve decision-making processes that rely on Computer Vision. Name some applications that you believe are not trustworthy or safe.
A: Self-driving cars that still make visual mistakes, facial recognition used without consent, and fake deepfake videos are not fully safe or trustworthy yet.