7 Limitations and Responsible AI Free Online FlashCards

How can an AI system make errors confidently?

AI can produce plausible-sounding answers or predictions that are inaccurate.

Why can complex AI models be difficult to interpret?

Complex models may make it difficult to identify why a particular output was produced.

Why does AI use not remove human responsibility for outcomes?

Data quality, objectives, and human choices affect AI outcomes; using AI does not remove human responsibility.

What is bias in an AI system, and where can it come from?

Bias is a systematic skew in data, system behavior, or outcomes. It can arise from under-representation, historical discrimination, inaccurate labels, or development choices.

What privacy risks can arise from AI use?

Risks include collecting more personal data than necessary, exposing it through weak safeguards, or inferring private details from other information.

How do AI safety and security differ?

Safety focuses on preventing harmful outcomes; security helps protect systems and data from unauthorized access or manipulation.

Why might AI perform poorly in unfamiliar or changing conditions?

Its performance may drop when real-world data differs from its training data or changes over time.

What can a correlation fail to tell us about an outcome?

A pattern or correlation does not necessarily explain why something happened or what action is appropriate.

How can neutral information contribute to discriminatory AI outcomes?

A seemingly neutral input can act as a proxy for a sensitive characteristic and contribute to unequal or discriminatory results.

What practices can help reduce harmful bias in AI?

Check data quality and representation, test performance across relevant groups, involve varied perspectives, and monitor outcomes after deployment.

Why can one fairness score not prove that an AI system is fair?

Fairness depends on the setting, and fairness measures can conflict. A single score or test cannot establish fairness in every respect.

What does transparency about an AI system involve?

Transparency provides appropriate information about where and why AI is used, its important limitations, and who is responsible for it.