COMPUTER-USE AND BROWSER AGENTS

LEARNING GOALS Most agents you have studied act through tools with neat, structured doors, such as a function that takes a customer number and returns an order. But a great deal of the world has no such door. Many programs and websites were built for human eyes and human hands, with buttons, menus, and forms. In this chapter, you will learn about agents that control a screen, mouse, and keyboard, agents that browse and act on the web, and the special safety concerns that arise when an agent can take actions in the wild. THE REMOTE CONTROL PICTURE Imagine you must operate a machine that has no computer connection. The only way to use it is to stand in front of it, look at its display, press its buttons, and read what happens. Now imagine you cannot be there, so you ask a helper with a video camera and a robotic finger to do it for you. The helper looks at the display, decides which button to press, and presses it. That is a computer-use agent. It does not call a special door, because it uses the same front door that a human uses.