AI and Hallucinations

Al-powered object detection could bring relief to __________ impaired people, most of whom live in developing countries.

• Visually

• Hearing

Al could also analyze mobile phone __________ to distinguish between a benign skin lesion and melanoma, potentially helping millions of rural dwellers who do not have access to dermatologists.

• Images

• Scanner

Al-enabled chatbots for education and skill development to improve access and quality of _____________.

• Education

• Living

What error did Mr. Schwartz make during

the case?

• He looked up the wrong cases.

• b) He did not fact-check ChatGPT’s responses himself.

• c) He used ChatGPT to help his work.

• d) He submitted his brief to the wrong judge.

Approximately how often does the current version of ChatGPT hallucinate?

• a) 5% of the time.

• b) 15% of the time.

• c) 20% of the time.

• d) 50% of the time.

After hallucinating, ChatGPT will typically:

• a) Admit it is wrong.

• b) Always ask the user to check its answers.

• c) Provide useful links so users can check information.

• d) Insist it is correct.

Mr. Schwartz admits he ‘outsourced’ his work. Using clues from the text, what is the best definition for ‘outsourced’?

• a) To do work then forget to check answers.

• b) To get something/one else to do work for you.

• c) To provide unreliable sources of information.

• d) To be busy and stressed.

Define Chatbot hallucination in your own words

How is hallucination different from lying?

If you were teaching someone to use ChatGPT for the first time, what would you recommend they do to prevent any negative effects from chatbot hallucination? How could you use Mr Schwartz’s case as an example of what not to do?