2. AI systems, agents, and tasks

Separate the AI part of a product from the full software system around it. Learners will identify where the interface, database, rules, model, user, and business process each fit in an everyday AI example.

Confira o que você entendeu com um quiz curto.

Use o que você aprendeu na aula anterior para resolver problemas do mundo real.

Define an agent as something that observes an environment and acts back into it. Learners will mark the sensors, actuators, environment, and action loop for examples such as a robot vacuum, chatbot, or game-playing AI.

Confira o que você entendeu com um quiz curto.

Use o que você aprendeu na aula anterior para resolver problemas do mundo real.

Use the PEAS framework to describe an AI task: Performance measure, Environment, Actuators, and Sensors. Learners will turn a vague goal like “drive safely” or “recommend music” into a clearer agent specification.

Confira o que você entendeu com um quiz curto.

Use o que você aprendeu na aula anterior para resolver problemas do mundo real.

Frame an AI task as a mapping from inputs to useful outputs. Learners will distinguish the task from the product, the dataset, and the model by asking what question the system is trying to answer or what action it must choose.

Confira o que você entendeu com um quiz curto.

Use o que você aprendeu na aula anterior para resolver problemas do mundo real.

Compare classification tasks, which choose a category, with regression tasks, which predict a number. Learners will sort examples such as spam detection, disease risk, delivery time, and house price prediction into the right task type.

Confira o que você entendeu com um quiz curto.

Use o que você aprendeu na aula anterior para resolver problemas do mundo real.

Recognize tasks that order options instead of producing one label or number. Learners will compare search ranking, feed ranking, and recommendation systems by identifying the candidate items, ranking signal, and selected output.

Confira o que você entendeu com um quiz curto.

Use o que você aprendeu na aula anterior para resolver problemas do mundo real.

Identify tasks that group items or retrieve similar ones without first assigning a fixed label. Learners will reason through examples like customer segments, duplicate photo search, and “find documents like this one.”

Confira o que você entendeu com um quiz curto.

Use o que você aprendeu na aula anterior para resolver problemas do mundo real.

Recognize generative AI tasks where the output is new text, images, audio, code, or other content. Learners will identify the prompt, context, constraints, and generated artifact in examples using chatbots and image generators.

Confira o que você entendeu com um quiz curto.

Use o que você aprendeu na aula anterior para resolver problemas do mundo real.

Distinguish systems that choose a sequence of actions from systems that make a single prediction. Learners will trace planning and control examples such as robot navigation, game AI, warehouse routing, and thermostat control.

Confira o que você entendeu com um quiz curto.

Use o que você aprendeu na aula anterior para resolver problemas do mundo real.

Compare hand-written rules, learned models, and foundation models as different ways an AI system produces outputs. Learners will decide which kind of mechanism is likely being used in examples and what evidence would suggest each one.

Confira o que você entendeu com um quiz curto.

Use o que você aprendeu na aula anterior para resolver problemas do mundo real.

Interpret model outputs that come as scores, probabilities, or confidence values rather than final decisions. Learners will see how a threshold turns a score into an action and why changing it affects false alarms and missed cases.

Confira o que você entendeu com um quiz curto.

Use o que você aprendeu na aula anterior para resolver problemas do mundo real.

Locate the point where an AI output becomes a real-world action, message, ranking, approval, or alert. Learners will compare fully automated decisions with human review, override, and “AI suggests, human decides” workflows.

Confira o que você entendeu com um quiz curto.

Use o que você aprendeu na aula anterior para resolver problemas do mundo real.

Judge an AI system by the result it is supposed to improve, not just by whether it produced an output. Learners will connect performance measures such as accuracy, error rate, user satisfaction, safety, cost, and fairness to the task being performed.

Confira o que você entendeu com um quiz curto.

Use o que você aprendeu na aula anterior para resolver problemas do mundo real.

Trace how deployed AI systems can change as users, data, or environments change. Learners will recognize feedback loops, monitoring signals, and model drift in examples like recommendations, fraud detection, and predictive maintenance.

Confira o que você entendeu com um quiz curto.

Use o que você aprendeu na aula anterior para resolver problemas do mundo real.

Create a compact description of an AI example using task, input data, model or rules, output, decision point, and performance measure. Learners will practice classifying everyday AI systems with a reusable “system card” format.

Confira o que você entendeu com um quiz curto.

Use o que você aprendeu na aula anterior para resolver problemas do mundo real.

Revise este capítulo com prática baseada nos seus erros.