AI characteristic: the ability to learn and adapt

Name:

Date:

Class:

Quick Questions

Shade in the A or B circle for each question.

1. Which statement about the ability to learn and adapt is correct?

A An AI system changes its own behaviour as it gains new data or experience

B An AI system only changes when a programmer rewrites its code

2. A spam filter gets better at spotting junk email as users mark more messages as spam. What is it showing?

A Following a fixed rule that never changes

B Learning from new data

3. Do all AI systems have to learn and adapt?

A No, some AI systems can learn but it is not required

B Yes, a system cannot be AI unless it learns

4. Which is an example of adapting rules or data?

A A system updates its own rules after seeing the results of past decisions

B A system gives the same answer every time whatever happens

5. Which type of AI is most closely linked to the ability to learn and adapt?

A A fixed program with no changes

B Machine learning

Match the Pairs

Draw a line connecting each item on the left to its match on the right.

Left column		Right column
1learn and adapt		ALearning and adapting means the system changes its own behaviour using new data or experience. A person editing the rules is not the system adapting itself.
2machine learning		Bnew data -> update rules -> better results
3A spam filter learns from new emails that users mark as spam and updates its own rules, so it gets better at spotting spam		CThe type of AI that can learn and adapt by changing its own rules or data as it receives new data
4An expert system uses a fixed rule base and only changes when a person edits it, so it does not have to learn and adapt		DAn AI system changes its own behaviour as it gains new data or experience, so it improves over time
5All AI systems must learn and adapt, or they are not AI		ELearning and adapting is not required. Some AI systems, such as expert systems, do not change themselves, but can still be AI.
6A programmer edits the rules by hand, so the system has learned and adapted		Flearning = optional

Definitions

learn and adapt

Fill in the Blanks

Complete each line of code by filling in the missing parts.

_____ newData data ← data + newData

IF prediction ≠ actual ____ rules ← update(rules, newData) ENDIF

// fixed expert system learns ← _____ isAI ← TRUE

FOR count ← 1 __ 100 model ← update(model, newData) NEXT count

storedData ← storedData + _______

Spot the Error

Shade in the letter of the line that contains an error.

Question 1

A WHILE newData EXISTS

B prediction ← guess(rules)

C err ← actual - prediction

D rules ← rules

E ENDWHILE

Question 2

A INPUT answer

B IF answer = correct THEN

C weight ← weight - 1

D ENDIF

Question 3

A INPUT word

B IF word IN vocabulary THEN

C ADD word TO vocabulary

D ENDIF

Question 4

A newData ← TRUE

B IF newData AND canLearn THEN

C OUTPUT 'No change'

D ENDIF

Question 5

A OUTPUT newData

B model ← learn(model, newData)

C OUTPUT predict(model)

Answer Key

AI characteristic: the ability to learn and adapt

Section A: Quick Questions

A — An AI system changes its own behaviour as it gains new data or experience

B — Learning from new data

A — No, some AI systems can learn but it is not required

A — A system updates its own rules after seeing the results of past decisions

B — Machine learning

Section B: Match the Pairs

learn and adapt → An AI system changes its own behaviour as it gains new data or experience, so it improves over time

machine learning → The type of AI that can learn and adapt by changing its own rules or data as it receives new data

A spam filter learns from new emails that users mark as spam and updates its own rules, so it gets better at spotting spam → new data -> update rules -> better results

An expert system uses a fixed rule base and only changes when a person edits it, so it does not have to learn and adapt → learning = optional

All AI systems must learn and adapt, or they are not AI → Learning and adapting is not required. Some AI systems, such as expert systems, do not change themselves, but can still be AI.

A programmer edits the rules by hand, so the system has learned and adapted → Learning and adapting means the system changes its own behaviour using new data or experience. A person editing the rules is not the system adapting itself.

Section C: Definitions

learn and adapt — An AI characteristic where a system changes its own behaviour as it gains new data or experience, so it improves over time. Some AI systems can do this, but it is not required.

Section D: Fill in the Blanks

INPUT newData
data ← data + newData

IF prediction ≠ actual THEN
   rules ← update(rules, newData)
ENDIF

// fixed expert system
learns ← FALSE
isAI ← TRUE

FOR count ← 1 TO 100
   model ← update(model, newData)
NEXT count

storedData ← storedData + newData

Section E: Spot the Error

D — A system that learns and adapts must change its own rules using the error from each prediction. Here the rules are copied back unchanged, so nothing is learned. It should be rules ← adjust(rules, err).

C — When the answer is correct, the system should reinforce what worked by increasing the weight. Subtracting 1 makes it worse with experience. It should be weight ← weight + 1.

B — The system should only learn a word it does not already know. With IN, it would add words that are already in the vocabulary and never learn new ones. It should be IF word NOT IN vocabulary THEN.

C — When new data arrives and the system can learn, it should update its model. Outputting 'No change' means it does not adapt. It should be model ← retrain(model).

A — A learning system must take in new data as input so it can learn from it. OUTPUT sends data out instead. It should be INPUT newData.