Ethics of Generative AI Use in Medical Practise and as a Student

Question 1

AEREOLE PE - 9

Answer 1

1. AI stupidity
2. Evil AI and existential threats
3. Risk and Liability
4. Energy use and environmental impact
5. Opacity
6. Loss of jobs
7. Ethics of AI use in studies
8. Perpetuating inequality and bias
9. Effects on human behaviour and intelligence/skill

Question 2

Opacity - 2

Answer 2

• generatve AI does not always follow the code and is able to operate semi autonomously without human supervision or intervention
• Its processes are unexplainable

Question 3

T/F AI users are entitled to explanations

Answer 3

True

Question 4

EU guidelines require

Answer 4

• that the decisions made by the AI system can be undrstood and traced by human beings
• A certain degree of transparency should be required for high risk AI systems

Question 5

Weapons of math destruction -3

Answer 5

1. Opacity
2. Damage
3. Scale

Question 6

Transparency and opacity both carry moral weight

Answer 6

When they say both carry moral weight, it means both transparency and opacity have ethical importance, but it doesn’t mean both are good. Opacity can have a negative moral weight, leading to unfairness or hidden harm. Transparency, on the other hand, is generally seen as positive because it allows for fairness and accountability.

Question 7

AI stupidity - 3

Answer 7

1. Generative AI does not always know when its outputs are wrong.
2. It mimics human language and behaviour
3. When it doesnt find the right aswer it invents the answer and presents it as corrrect

Question 8

Automation bias

Answer 8

When people trust answers from computers (software programmes) over those of persons

Question 9

Concerns about automation bias - 2

Answer 9

1. there can be mistakes in how computers are programmed thus affecting the data they access
2. Processing may appear perfect while is not

Question 10

LLM

Answer 10

• large language models
• have a tone of authority and are “confidently wrong”

Question 11

hopes of AI

Answer 11

• that it might help overcome social inequities
• make decisions that are less biased that those of human beings (bc it is not influenced by emotion)

Question 12

Bias in AI - 2

Answer 12

1. genertaive AI inherits many of our human biases
2. New generation generative AI is trained on data which contains biased decisions and reflects historical social inequities (things from the past then get carried over)

Question 13

How does AI keep dividing and perpetuate inequality -2

Answer 13

1. when it gives advice or solutions, it does so through the lens of the priveledged
2. might recommend effective interventions that ignore the need for social justice

Question 14

Risk and liability

Answer 14

• how data is processed is opaque to the user, so we cannot know if it is secure or not

Question 15

Can personal data of an individual be linked or added to AI

Answer 15

No, that is a volation of the POPI ACT

Question 16

Is all text produced by an LLM plagiarism?

Answer 16

he doesnt think so

Question 17

5 scenarios where using AI is unethical

Answer 17

1.
2.
3.
4.
5.

Question 18

Faculty of HS guidelines on AI use by students

Question 19

Potential for AI to harm without malice

Question 20

Potential for AI to be abused for harmful purposes

Question 21

Potential for AI to become intentionally malicious

Question 22

Possibility for an existential threat to humanity