7 AI Limitations and Responsible Use Free Online FlashCards

What kinds of tasks can AI systems perform?

AI can recognize patterns, classify information, make predictions, recommend actions, and generate text or images.

Why might an AI system struggle in a new setting?

Performance depends on the task, training data, design, and conditions. A system that performs well on familiar examples may be less reliable when the data or setting changes.

What is an AI hallucination?

A hallucination, sometimes called a confabulation, is fluent AI output that contains false details or invented sources.

Does confident, polished AI output prove that it is correct?

No. A polished or confident answer is not proof of correctness; important claims should be checked against reliable evidence.

How can teams respond when AI results may be unreliable?

Test it in the setting where it will be used, track uncertainty and errors, and provide a way to pause or fall back to another process when results are unreliable.

Where can bias enter an AI system?

Bias can enter through training data, labels, design choices, or the way people use a system.

How could past hiring records bias a hiring model?

A hiring model trained on past hiring records may reproduce patterns in those records.

Why is overall accuracy not enough to assess fairness?

Evaluate performance for relevant groups and involve affected people. A single overall accuracy score may hide unequal errors.

What are core practices for protecting privacy in AI use?

Collect only the information needed, protect access, and consider privacy throughout design and use.

When is it appropriate to enter confidential information into an AI tool?

Avoid entering confidential information unless the tool’s data practices and your authorization make it appropriate.

How should safeguards for AI risks be chosen?

Match safeguards to the possible harm: test before deployment, monitor for problems, limit access where needed, and provide ways to override, repair, or stop the system.

What should people be told about AI in decisions affecting them?

People should know when AI meaningfully contributes to a decision that affects them.