The maths behind AI (made simple) quiz

Question 1

What is the "mean" (average) of a list of numbers, and why is it useful in AI?

Accepted Answer

Add all numbers and divide by how many there are

Answer

The largest number in the list

Answer

The middle number after sorting

Answer

The difference between biggest and smallest

Question 2

An artificial neuron computes a "weighted sum". If inputs are 2 and 3 with weights 0.5 and 1.0, what is the weighted sum?

Accepted Answer

4

Answer

5

Answer

6

Answer

2.5

Question 3

In the line equation y = mx + c, what does the letter m represent?

Accepted Answer

The slope (steepness) of the line

Answer

The point where the line crosses the y-axis

Answer

A random number

Answer

The total number of points

Question 4

What is a vector in the simple maths used by AI?

Accepted Answer

An ordered list of numbers

Answer

A single yes-or-no value

Answer

A picture file

Answer

A type of computer chip

Question 5

In probability, what range of values can the probability of an event take?

Accepted Answer

Between 0 and 1

Answer

Between -1 and 1

Answer

Any number at all

Answer

Between 0 and 100 only

Question 6

A model labels 100 images and gets 90 correct. What is its accuracy?

Accepted Answer

90 percent

Answer

9 percent

Answer

10 percent

Answer

100 percent

Question 7

What does the "error" or "loss" of an AI model measure?

Accepted Answer

How far predictions are from the correct answers

Answer

How fast the computer runs

Answer

How many inputs there are

Answer

The colour of the data

Question 8

Gradient descent is often compared to walking downhill. What is the model trying to reach?

Accepted Answer

The lowest point, where error is smallest

Answer

The highest point of the hill

Answer

A random spot

Answer

The starting point again

Question 9

What does the "learning rate" control during training?

Accepted Answer

The size of each adjustment step

Answer

The number of images used

Answer

The speed of the internet

Answer

The colour of the output

Question 10

Why do we often "normalise" or scale numbers (for example to the range 0 to 1) before feeding them to an AI model?

Accepted Answer

So features with big values do not unfairly dominate

Answer

To delete the data

Answer

To make the file smaller on disk

Answer

To change the colours

Question 11

How can a computer measure the straight-line distance between two points (x1,y1) and (x2,y2)?

Accepted Answer

square root of ((x2-x1)^2 + (y2-y1)^2)

Answer

(x2-x1) + (y2-y1)

Answer

(x2 x y2) - (x1 x y1)

Answer

x2 - x1 only

Question 12

The k-Nearest Neighbors method classifies a new point by looking at the k closest points. What maths idea does it mainly rely on?

Accepted Answer

Distance between points

Answer

Multiplication tables

Answer

The alphabet order

Answer

Random guessing

Question 13

A sigmoid function squashes any number into a small range. What range is that?

Accepted Answer

Between 0 and 1

Answer

Between -100 and 100

Answer

Between 1 and 10

Answer

Whole numbers only

Question 14

A model outputs a probability of 0.8 that an email is spam, and we use a threshold of 0.5. What is the decision?

Accepted Answer

Spam, because 0.8 is above 0.5

Answer

Not spam, because 0.8 is below 0.5

Answer

Cannot be decided

Answer

Always not spam

Question 15

When a model gives probabilities for several classes (for example cat, dog, bird), what should all those probabilities add up to?

Accepted Answer

1

Answer

100

Answer

0

Answer

The number of classes

Question 16

If a model's class probabilities are cat 0.7, dog 0.2, bird 0.1, which class does "argmax" choose?

Accepted Answer

Cat

Answer

Dog

Answer

Bird

Answer

None of them

Question 17

What does it mean when a model is "overfitting"?

Accepted Answer

It memorises training data but fails on new data

Answer

It runs too slowly

Answer

It uses too little memory

Answer

It always gives random answers

Question 18

Why do we split our data into a "training set" and a "test set"?

Accepted Answer

To honestly test the model on data it has not seen

Answer

To make training take longer

Answer

To delete half the data

Answer

To use more electricity

Question 19

Two students score the same average on a test, but one set of scores is spread out and the other is close together. Which maths idea describes this spread?

Accepted Answer

Variance (or standard deviation)

Answer

The mean

Answer

The largest score

Answer

The number of students

Question 20

If 3 out of 4 predictions are correct, what is this as a percentage?

Accepted Answer

75 percent

Answer

34 percent

Answer

25 percent

Answer

43 percent

Question 21

In AI maths, what is a matrix?

Accepted Answer

A table of numbers in rows and columns

Answer

A single number

Answer

A type of graph line

Answer

A computer program

Question 22

What is "one-hot encoding" and why is it used?

Accepted Answer

Turning a category into a list with a single 1 and the rest 0s

Answer

Making text all uppercase

Answer

Deleting duplicate rows

Answer

Adding random noise

Question 23

In a neuron, what is the job of the "bias" term that is added to the weighted sum?

Accepted Answer

It shifts the output up or down for flexibility

Answer

It deletes inputs

Answer

It makes training slower on purpose

Answer

It counts the inputs

Question 24

If it has rained on 8 of the last 10 cloudy days, what is the estimated probability of rain given that it is cloudy?

Accepted Answer

0.8 (80 percent)

Answer

0.2 (20 percent)

Answer

0.5 (50 percent)

Answer

0.08 (8 percent)

Question 25

Two variables are "positively correlated" when one tends to go up as the other goes up. Which example shows positive correlation?

Accepted Answer

More hours studied, higher exam score

Answer

More rain, fewer umbrellas sold

Answer

Taller person, smaller shoe size

Answer

Older car, faster top speed

Question 26

A piece of data doubles every step: 1, 2, 4, 8, 16... What kind of growth is this?

Accepted Answer

Exponential growth

Answer

Linear growth

Answer

No growth

Answer

Shrinking

Question 27

A logarithm answers the question "how many times do I multiply the base by itself to get a number?". What is log base 2 of 8?

Accepted Answer

3

Answer

4

Answer

8

Answer

2

Question 28

Why are the weights of a neural network usually set to small random numbers before training begins?

Accepted Answer

So different neurons can learn different things

Answer

Because random numbers are always correct

Answer

To slow the computer down

Answer

Because zero is not a real number

Question 29

How many different 2-letter codes can you make using the 26 letters of the alphabet, if letters can repeat?

Accepted Answer

676

Answer

52

Answer

26

Answer

100

Question 30

Why do AI systems sometimes round numbers, such as showing 0.6667 as 0.67?

Accepted Answer

To keep numbers short and readable

Answer

To make answers wrong on purpose

Answer

Because computers cannot add

Answer

To use more memory

Test yourself