Unit 6: Artificial Intelligence

Question 1

Artificial Intelligence (AI)

Answer 1

Machines that can perform tasks requiring human-like intelligence

Question 2

2 Subfields of AI
-> Machine Learning
-> Deep Learning

Answer 2

Machine Learning = AI learns from data.

Deep Learning = AI uses artificial neural networks (inspired by human brain).

Example:
ML is like teaching a child with specific instructions: Child learns patterns from the examples you give them.
=> ML in online store suggests products based on your purchase trends (e.g. recommending manga bc you previously bought manga)

DL: App that recognises plant species from photos. DL automatically identify the leaves’ colour, shape from pixels, and learns features without specific instructions.

Question 3

2 Main Types of AI
-> Narrow AI
-> General AI

Answer 3

Narrow AI = good at 1 thing only
=> All AI today is Narrow AI

General AI
* Not real yet (theoretical)
* Human-like intelligence
* Can solve any problem

Question 4

Why would it matter to differentiate between types of AI for businesses? (2)

Answer 4

1) AI ≠ one thing
* Different AI have different abilities, limits & costs
* Clear understanding enables responsible & effective use

2) Understanding AI is useful because it…
…helps decide if a project is technically possible
…makes communication with technical teams clearer
…assess ethics, laws, risks & limits

Question 5

Architecture of Deep Neural Networks: Artificial Neural Networks (ANNs)

Answer 5

ANNs = “human brain”
* Made of connected artificial neurons (nodes)
* Turn input → output

Neural Network Structure:
* Input layer → receives raw data (image, text, numbers)
* Hidden layers → process data & find patterns
* Output layer → final answer

🧠 “Deep” = many hidden layers
🧠 More layers → detect more complex patterns
🧠 Deep ≠ intelligent, just more layers

Question 6

AI uses Decision-Making Through Probability (5 Steps)

I,P,PM, P, D

Core idea: AI does not “know” things — it calculates probabilities.

Answer 6

1) Input
* AI gets data (image, text, sound)
* Data is turned into numerics (pixels)

2) Processing (Neural Network)
* Data passes through layers
* Neurons weight important features & filter patterns

3) Pattern Matching

4) Probability
* AI assigns a probability (0–1) to each possible result

5) Decision
* AI chooses the most likely outcome

Question 7

Probability in AI vs. Prediction in AI

Answer 7

Probability
= “confidence level” (How likely is answer correct, based on what AI learned)

Prediction
= “best guess” based on probability.

🧩 Example:
AI says an image is 96.8% likely to be a dog.
That doesn’t mean it knows it’s a dog — it just estimates based on past data.

Question 8

3 Limitations of Decision-Making with AI

Answer 8

1) No human explanation/reasoning

2) Confidence can be wrong
* When data is misleading or biased

3) Context matters
* AI works best in the area it was trained in
* Performs shitty in new or different situations

Question 9

Strategic Lessons for Business: Rule-Based AI “DEEP BLUE”

=> 3 Business Lessons

Answer 9

IBM’s Deep Blue vs. Garry Kasparov
- Pre-programmed rules
- Specialized only for chess
- 200 million board positions per second
- Using 480 custom-designed chips

Business Lessons:
1) Good for structured tasks: Rule-based AI good in predictable, well-defined environments.

2) AI executes known rules well. No adaption to new situations (= No innovation)

Good for: Tax Software or other rule-driven business processes.

Question 10

Strategic Lessons for Business: Learning-Based AI “AlphaGo”

=> 3 Business Lessons

Answer 10

Google’s AlphaGo vs. Go Champions
- Go = cannot be solved by brute-force calculation
- Learned from millions of past games and adapted strategies
- Excelled in complex, dynamic, unstructured environments

Business Lessons:
1) Adaptability: Learning-based AI handles complex, unpredictable situations.

2) Data-driven: AI improves as it processes more data, enabling smarter strategies.

3) Innovation: AI can discover patterns or strategies humans might not see.

Good for: Customer recommendation systems, dynamic pricing.

Question 11

Opportunities: How AI Creates Value

A,DDDM,P,GA

Answer 11

1. Automation:
   AI handles repetitive & rule-based tasks better than humans.
   e.g.: Chatbots handling customer queries 24/7

=> Strategic value: ↓costs↓ & employees more time for higher-value work.

2. Data-Driven Decision-Making:
   Predicting market demand or stock fluctuations by identifying massive datasets

=> Strategic value: Planning accuracy, ↓risk↓ & faster responses to change.

3. Personalization:
   e.g.: Personalized advertising in e-commerce (Amazon, Zalando)

=> Strategic value: customer relationships + loyalty.

4. Generative AI:
   => Strategic value: Opens new markets and product categories.

Question 12

Challenges and Risks: What AI Still Struggles With (5)

Answer 12

1) Security & Privacy Risks:
e.g.: Deepfakes, data breaches or misinfo.

2) No Transparency = AI = “Black Box”

3) Job Displacement = AI takes over lower-skilled labor

4) Overdependence & Human De-Skill:
If decisions rely too heavily on AI, employees lose critical thinking skills.

5) Bias:
AI mirrors biases in training data.
Solution: Diverse training data, bias audits, and human oversight.

Question 13

The EU AI Act (2024–2026)
-> Unacceptable Risk

Answer 13

Unacceptable Risk -> Social scoring, manipulative AI, mass surveillance -> Prohibited

Question 14

The EU AI Act (2024–2026): Penalty

Answer 14

Up to €30 million or 6% of global turnover for severe violations.

Question 15

The EU AI Act (2024–2026)
-> High Risk

Answer 15

High Risk -> CV screening, credit scoring, education, healthcare
-> Strict testing, documentation, human oversight

Question 16

The EU AI Act (2024–2026)
-> Limited Risk

Answer 16

Limited Risk -> Chatbots, deepfakes
-> Transparency required

Question 17

The EU AI Act (2024–2026)
-> Minimal Risk

Answer 17

Minimal Risk -> Spam filters, game AIs, recommender systems
-> No major restrictions