An Agent Has a Job
Imagine hiring someone to help you — but never telling them what you want. They show up, stand there, and have no idea what to do. That would not be very helpful! Every AI agent needs a job. That job is called a goal. The goal is the whole reason the agent exists. Every decision the agent makes, every action it takes, is aimed at reaching that goal. Without a goal, an agent is just a confused machine spinning in circles.
The Goal Is the Whole Point
Think about the robot vacuum again. Its goal is to clean the floor. That one simple goal explains everything it does. Why does it move forward? To cover more floor. Why does it turn when it hits a wall? To keep cleaning instead of getting stuck. Why does it return to its dock? To recharge so it can clean again later. Every single action traces back to the same goal: clean the floor. Now imagine the same robot but with a different goal: deliver a snack to room 3. Suddenly all its actions would change! It would navigate hallways instead of cleaning, stop at room 3 instead of returning to a dock. Same robot body — completely different behavior — because the goal changed.
The goal is the most important ingredient of any AI agent. The goal explains why the agent does what it does. Change the goal and you change everything the agent does — even if nothing else about the agent changes.
Let us look at some agents and their goals. A spam filter has one goal: keep junk email out of your inbox. Every email it looks at, it is asking: does this belong in the inbox or not? That is it. Nothing else matters to the spam filter. A chess AI has one goal: win the game. Every single move it thinks about, it is asking: does this move help me win? All the sensing and calculating exists only to serve that one goal. A weather app has one goal: tell you accurately what the weather will be. Every data point it collects — temperature, wind speed, humidity — is in service of giving you a reliable forecast. One goal. Everything else flows from it.
Fill in the missing word.
Goals can be big or small, simple or complex. A simple goal: beep when the timer runs out. A medium goal: clean the entire floor before the battery dies. A complex goal: help a driver navigate across a whole city safely. Bigger goals usually require the agent to take more kinds of actions, sense more kinds of information, and make more complicated decisions. But no matter the size, there is always exactly one goal steering the whole show. This is also why people have to be very careful when they design AI agent goals. If you give an agent the wrong goal — or a goal that is slightly off — it can do things you did not intend. Getting the goal right is one of the most important jobs in AI.
Imagine an agent with the goal of keeping your room tidy. If the goal is not careful enough, the agent might throw away your favorite toys to clear clutter! Getting the goal exactly right is a serious job that requires careful thought.
Match each agent to its main goal.
Terms
Definitions
Drag terms onto their definitions, or click a term then click a definition to match.
Why is the goal the most important ingredient of an AI agent?
You give two identical robot vacuums different goals: one should clean the floor and the other should deliver snacks. What will be different about them?
Goal Detective
- Look around your home or school for three machines or apps that act like agents.
- For each one, try to write one sentence that describes its goal.
- Then ask yourself: if the goal changed slightly, how would the agent behave differently?
- For example: if a robot vacuum's goal changed from 'clean the whole floor' to 'only clean the kitchen,' what would change?
- Share your three agents and their goals with someone. See if they agree with your goal descriptions!