AI for Everyone: Week 4 Quiz

What are the jobs that AI is most likely to displace over the next several years?

Jobs that comprise primarily of routine, repetitive work

Say you are building an AI system to help make diagnoses from X-ray scans. Which of the following statements about explainability of AI do you agree with?

Lack of explainability can hamper users' willingness to trust and adopt an AI system

What is the Goldilocks Rule of AI?

One shouldn't be too optimistic or too pessimistic about AI technology

What are the current limitations of AI technology? (Select all that apply)

✅Explainability is hard ✅AI technology can discriminate ✅AI technology can be biased ✅AI technology is susceptible to adversarial attacks

Congratulations! You deserve a pat on the back for finishing this course.

True

If a developing economy has a strong and thriving coffee bean manufacturing industry (or some other vertical industry), then it has an advantage in applying AI to coffee bean manufacturing (or other vertical industry).

True

Using current AI technology, if a machine learning system learns from text that reflects unhealthy biases/stereotypes, then the resulting AI software may also exhibit similarly unhealthy biases/stereotypes.

True

Using current AI technology, if a machine learning system learns only from text that is completely neutral and does not reflect any gender biases, then we would expect it to exhibit no, or at most minimal, gender bias.

True

Which of these are examples of adversarial attacks on an AI system? (Select all that apply)

✅Subtly changing an image to make an AI system mistakenly recognize a dog as a cat. ✅Subtly modifying an audio clip to make a speech recognition system think someone said "Yes, authorized" when they actually said "No, reject." ✅Adding a sticker to a stop sign to make an AI system fail to detect it.

Which of these are good practices for addressing bias in AI? (Select all that apply)

✅Systematic auditing processes to check for bias ✅Technical solution such as "zeroing out" bias ✅Using more inclusive/less biased data