Innovating with Google Cloud Artificial Intelligence practice questions

Artificial intelligence is the broad field of building computer systems that can carry out tasks that would normally require human intelligence.

The board member wants the broadest definition, which is AI as a field: machines performing tasks that need human-like intelligence. The other options are real but narrower things: supervised ML (learning from labelled examples), generative AI (creating content) and business intelligence (reporting on data).

1. Notice the stem asks for the broadest sense of AI.
2. Recognise that ML and generative AI are subsets of AI.
3. Recognise that dashboards are business intelligence, not AI.
4. Choose the description that covers the whole field.

Exam tip: When asked for the broadest definition, AI means machines doing tasks that need human-like intelligence.

Generative AI is a type of AI, built on machine learning, that creates new content such as text, images, audio and code from a prompt.

The deciding detail is that the software must create new descriptions and images. Creating content is what makes AI generative. Predictive ML outputs a prediction, business intelligence outputs reports on existing data, and rule-based automation can only follow fixed logic.

1. Find what the software must output: new text and images.
2. Match creating new content to generative AI.
3. Check that the other options produce predictions, reports or fixed actions.
4. Choose generative AI.

Exam tip: If the output is new content, the technology is generative AI.

Predictive ML and generative AI do different jobs: predictive models output a score or label, while generative AI creates new content such as text.

The bank already has prediction. The new capability must therefore be something only generative AI does: creating content, here a tailored letter. Segmenting applicants is tempting because it is also AI, but it is analytical ML. The probability score duplicates the existing model, and the dashboard is business intelligence.

1. Note what the existing model already does: predict default.
2. Ask what generative AI does that prediction does not: create content.
3. Look for the option that produces new content.
4. Reject the analytical ML and reporting options.

Exam tip: Predictive ML answers 'what will happen?'; generative AI produces something new, such as a letter or an image.

Business intelligence (BI) is the practice of collecting business data and presenting it in reports and dashboards so people can monitor performance and decide what to do.

Dashboards that track sales against targets are the standard output of business intelligence. The AI options are tempting because they sound modern, but none is needed: there is nothing to predict, generate or act on, only performance to report.

1. Identify the output: dashboards of past sales against targets.
2. Recognise reporting on business performance as business intelligence.
3. Check that no prediction, content or action is required.
4. Choose business intelligence.

Exam tip: Dashboards and reports that track how the business is performing are business intelligence.

Data analytics examines data to find patterns and answer questions; business intelligence packages business data into recurring reports and dashboards that leaders use to track performance.

The analyst is answering a 'why' question by exploring and testing data, which is data analytics. The colleague delivers a standing report for decision makers, which is business intelligence. The swapped option is the near-twin, and the machine learning options fail because no model is being trained.

1. Classify the analyst's work: exploring data to explain a change.
2. Classify the colleague's work: a recurring report for leaders.
3. Map exploration to analytics and recurring reporting to BI.
4. Reject options that swap the roles or claim a model is being trained.

Exam tip: Analytics asks and answers questions of the data; BI delivers the regular reports and dashboards leaders run the business on.

Agentic AI describes systems that pursue a goal on a user's behalf: they reason, plan multi-step work and take actions through tools and connected systems.

The difference is action. A chatbot responds; an agent plans and acts, for example by looking up a record and updating it. Multilingual answers are still just answers, a scripted decision tree is rule-based automation, and reporting on questions is analytics.

1. Identify what a basic chatbot does: it answers.
2. Ask what an agent adds: planning and taking actions toward a goal.
3. Reject options that a chatbot or a script can already do.
4. Choose the option that involves acting in other systems.

Exam tip: Chatbots answer; agents plan and act to get a goal done.