Artificial Intelligence: Ethics in AI Research

Artificial Intelligence
Business Disruptive Technology

Drag and Drop

AI machines types are classified into ​

Placeholder

main categories based on their capabilities and complexity in decision-making and awareness.

Drag these tiles and drop them in the correct blank above

four

one

two

three

Reactive Machines

Limited Memory Machines

Theory of Mind Machines

Self-Aware Machines

Reactive Machines

Dropdown

They don’t store memories or ​

Select answer

and therefore cannot use ​

Select answer

information to influence future decisions.

Multiple Choice

Which AI system was designed to respond to moves in chess but doesn’t learn or adapt beyond its programming?

• IBM Blue Sky

• IBM Deep Thought

• IBM Deep Blue

• IBM Smart Move

Scenario for Reflection

Imagine a customer service chatbot used by a large e-commerce company. The chatbot is programmed to respond to common customer inquiries, such as order status, returns, and basic product questions. It follows a set of pre-defined scripts based on keywords in customer messages but doesn’t learn or adapt from previous interactions.

Open Ended

Reflection Questions: How could this "reactive" chatbot be both helpful and limiting for the company?

Word Cloud

In what situations might customers feel frustrated by the chatbot’s limited responses? (Please type keywords or phrases that come to mind)

• Unanswered questions

• Lack of personalization

• Repetitive answers

• Confusing responses

• Complex queries

• Limited understanding

Poll

How might the company benefit if the chatbot could learn and adapt from past interactions, rather than staying strictly reactive?

• Improved customer satisfaction due to more personalized responses

• Faster resolution of customer issues with fewer repeat inquiries

• Increased sales from better product recommendations

• Reduced workload for customer service agents

• All of the above

Drag and Drop

Reactive machines are purpose built for singular or specialised task using ​

Placeholder

machine learning models.

Drag these tiles and drop them in the correct blank above

supervised

unsupervised

deep learning

Limited Memory Machines

Multiple Choice

Which of the following best describes the concept of Limited Memory AI?

• AI systems that can store data temporarily and use past experiences to inform future actions.

• AI systems that respond to inputs but do not learn from past experiences.

• AI systems capable of understanding human emotions and intentions.

• AI systems that are aware of their own existence and can make decisions based on self-awareness.

Limited memory AI systems can learn from past data and use that information to make decisions, but they don’t retain the information over long periods.

Limited Memory Machines

Dropdown

Customers who frequently buy books on a specific topic or watch a certain genre are given ​

Select answer

recommendations. System uses relevant past data to adjust future suggestions, but does not rely on ​

Select answer

.

Drag and Drop

Self-Driving Cars: These vehicles, like those from Tesla or Waymo, use ​

Placeholder

and ​

Placeholder

to make real-time decisions (e.g., stopping at a red light or adjusting speed based on traffic).

Drag these tiles and drop them in the correct blank above

sensors

past driving data

manual controls

driver intuition

immediate feedback

real-time data

Word Cloud

What are the key benefits or utilities of self-driving cars in the future of transportation?

Theory of Mind

Draw

Draw an example of a situation where a machine with Theory of Mind could understand human emotions and respond appropriately.

Multiple Choice

What capability would an AI with "Theory of Mind" possess?

• It could store large amounts of data permanently.

• It has the capability to carry out tasks entirely on its own, without requiring any human guidance or involvement.

• It could understand human emotions and intentions to enable social interactions.

• It could process data faster than any human.

Drag and Drop

AI with Theory of Mind would, in theory, have the ability to understand human ​

Placeholder

, allowing for social interactions.

Drag these tiles and drop them in the correct blank above

emotions; intentions

memory; data

speed; efficiency

calculations; predictions

Open Ended

What might be some potential benefits and challenges of AI understanding emotions and intentions?

Multiple Choice

What is the nationality granted to Sophia the robot, making her the first robot to receive citizenship?

• United Arab Emirates

• Singapore

• Saudi Arabia

• Japan

Multiple Choice

What role has Sophia the robot played in the realm of AI?

• She is a musician

• She is an advocate for AI ethics

• She is a social media influencer

• She is a computer programmer

Sophia has been involved in discussions on AI ethics and has participated in numerous conferences and public events to raise awareness about the responsible development of artificial intelligence.

Multiple Choice

How can Theory of Mind machines assist in elderly care?

• By providing physical care and medical assistance without any emotional interaction.

• By offering financial advice and managing the elderly person's finances.

• By focusing only on cognitive activities, without considering emotional or social needs.

• By recognizing emotions, providing companionship, and adapting routines based on mood or behaviour.

Dropdown

While ​

Select answer

machines are still considered hypothetical, AI with this capability would need a much broader understanding of human interactions and motivations, thus crossing into the ​

Select answer

scope of general AI.

Self-Aware Machines

Self-Aware AI in Business Applications:

• Its development is still speculative but holds significant potential for business applications.

• While still in early stages, the integration of self-aware AI into business systems could drive innovation, efficiency, and customer-centric services.

Multiple Choice

What is the concept of Singularity in AI research?

• A point where AI exceeds human intelligence and can self-improve autonomously.

• A system in which AI and human intelligence merge to create a hybrid consciousness.

• The moment when AI begins to understand human emotions and intentions.

• A situation where AI systems are limited by their creators and cannot adapt beyond programming.

Multiple Choice

Which sector has made the most significant advancements in developing self-aware machines, particularly in autonomous systems capable of making tactical decisions?

• Entertainment

• Healthcare

• Retail

• Military Defense

Multiple Choice

What is a key characteristic of advanced AI in hypothetical autonomous military systems?

• The ability to understand their role in conflict and make moral decisions based on that.

• The capacity to evaluate tactical situations without considering the broader consequences.

• The limitation to only perform basic tasks without any strategic decision-making ability.

• The reliance on human operators for every tactical decision in combat scenarios.

Multiple Select

ChatGPT is an example of ------.

• Reactive AI

• Limited Memory AI

• Theory of Mind

• Self-Awareness

Match

Select the correct answer as either Supervised Learning, Unsupervised Learning, or Deep Learning:

• The system is trained using labelled examples ("spam" or "not spam") to make predictions.

• The clustering algorithm groups data without labelled examples.

• The facial recognition system uses neural networks.

• The AI uses labelled features such as house size and location to predict prices.

Supervised Learning

Unsupervised Learning

Deep Learning

Supervised Learning

• Supervised Learning

• Unsupervised Learning

• Deep Learning

• Supervised Learning

Dropdown

​

Select answer

potentially have a "​

Select answer

," allowing them to perform a wide variety of tasks that require abstract reasoning and ethical considerations.

Ethical Concerns in AI Research

Word Cloud

What are some ethical concerns associated with AI research?

Ethics in AI Research

• Singularity

• Moral Status

• Algorithm

• Robo-Ethics

• Economics

Dropdown

Singularity: As AI systems become more advanced, their ability to act ​

Select answer

could lead to unintended catastrophic outcomes (e.g., a self-aware AI pursuing goals in ways ​

Select answer

).

Drag and Drop

Moral Status is about Moral Accountability: If an AI makes harmful decisions, who is responsible? The ​

Placeholder

, ​

Placeholder

, or ​

Placeholder

?

Drag these tiles and drop them in the correct blank above

creator

the machine

the user

the government

the internet

society

Drag and Drop

Algorithm: The “​

Placeholder

” nature of many algorithms makes it difficult to determine how decisions are made, raising concerns about fairness and accountability.

Drag these tiles and drop them in the correct blank above

black-box

transparent

predictable

linear

white-box

The "black box" nature of algorithms

An example of the "black box" nature of algorithms is seen in facial recognition systems. These AI systems process facial features to make identifications, but how they reach a conclusion is often unclear.

For instance, if a system makes a mistake, it’s hard to know why it happened, which raises concerns about accountability and biases, especially in sensitive areas like law enforcement or hiring.

Multiple Choice

How can human biases impact AI algorithms?

• AI algorithms are always neutral and free from human biases

• Human biases can unintentionally be fed into AI algorithms, influencing their decisions

• AI systems that replicate human biases might reinforce discriminatory or flawed patterns in decision-making

• AI algorithms can eliminate all human biases without intervention

Poll

Should machines be allowed to make life-and-death decisions in warfare, given the ethical concerns around accountability?

• Yes, if it leads to faster and more efficient decision-making.

• No, human oversight should always be required.

• Maybe, under strict ethical guidelines and regulations.

• Unsure, need more information on the risks and benefits.

Multiple Choice

Economic Impacts: Which of the following statements best describes the economic concerns related to AI and automation?

• AI systems primarily enhance employment opportunities by creating new job sectors.

• AI automation may displace workers in sectors like manufacturing and retail.

• AI has no significant impact on wealth distribution, as it benefits all social classes equally.

• Economic inequality could decrease as AI systems allow for more equal access to technology and resources across society.