IB Questionbank

User interface language: English | Español

Date	November 2019	Marks available	2	Reference code	19N.1.SL.TZ0.4
Level	SL	Paper	1	Time zone	no time zone
Command term	Outline	Question number	4	Adapted from	N/A

Question

Artificial neural networks are changing surveillance

Currently, security cameras record activities in supermarkets and a security guard watches the camera footage in real time. If the guard sees something suspicious, action can be taken. The company AI Security Innovations is looking at ways to further develop this technology using artificial intelligence (AI) to automate the process.

The field of AI is developing rapidly and is being transformed by machine learning using artificial neural networks (ANNs). AI Security Innovations wants to link the cameras to an AI system that can distinguish between something innocent, such as a child playing with a toy gun, and a crime, such as shoplifting (see Figure 2). The AI security camera system will be designed to make the decision about whether to intervene or not. This could include “locking down” the premises to secure all exits so that a suspect cannot escape, or sending an alert in real time to the police.

Figure 2: An example of a security camera

[Source: https://pixabay.com/photos/video-camera-surveillance-camera-3121655/]

These cameras will be part of extremely sophisticated detection systems based on pattern recognition. They will be able to detect humans, rapidly separate authorized personnel from intruders, and match faces from multiple camera sources in order to track people moving from location to location. Researchers are even exploring systems that can detect the presence of concealed guns based on the way an individual walks.

Identify two types of artificial intelligence (AI).

[2]

a.i.

Identify two types of machine learning.

[2]

a.ii.

Outline how pattern recognition works.

[2]

a.iii.

Identify two advantages of using artificial neural networks (ANNs).

[2]

b.i.

Identify two disadvantages of using artificial neural networks (ANNs).

[2]

b.ii.

Identify two characteristics of deep learning.

[2]

b.iii.

To what extent should pattern recognition in AI systems be trusted to make decisions about sending real-time alerts to the police?

[8]

Markscheme

Answers may include:

Strong
Full
General
Weak
Narrow
Domain specified

Award [1] for identifying each type of AI up to maximum of [2].

a.i.

Answers may include:

Supervised
Unsupervised
Reinforced learning

Award [1] for identifying each type of machine learning up to maximum of [2].

a.ii.

Answers may include:

Images are stored in a database.
A new image is input.
Defining features of the new image are identified.
The new image is compared to the images in a database.
If the image approximates to one of the images in the database, it is recognized.

Award [1] for identifying each aspect of how pattern recognition functions up to maximum of [2].

a.iii.

Answers may include:

Can be trained to learn from past decisions.
Can implement a task that a linear program cannot.
Does not need to be reprogrammed.

Award [1] for each advantage of artificial neural networks identified up to maximum of [2].

b.i.

Answers may include:

Large processing requirements.
Can appear black box in nature.
The most accurate results may have arisen by trial and error.

Award [1] for each advantage of artificial neural networks identified up to maximum of [2].

b.ii.

Answers may include:

Includes multiple layers (usually more than three).
Requires very large processing requirements.

Award [1] for each characteristic of deep learning identified up to maximum of [2].

b.iii.

Answers may include:

Reasons why pattern recognition in AI systems should be trusted to make decisions about sending real-time alerts to police:

It can analyse large amounts of data, so police decisions will be based on greater information (systems).
It can provide more information to police to assist them in carrying out their job (systems).
Standard response to dangers can eliminate human error, such as panic responses or ill-judged responses / increase reliability and consistency.
There is less chance of not noticing a dangerous situation, whereas humans can get distracted (systems).
The AI system can react faster than a human to suspicious circumstances.

Reasons why pattern recognition in AI systems should not be trusted to make decisions about sending real-time alerts to police:

Unlike humans, the AI system cannot make decisions based on ethical criteria.
Creating rules that take into account all of the possible ethical dilemmas may not be possible and cannot always determine right from wrong (systems).
If unsupervised learning is used, the AI may self-learn and arrive at decisions that may not be appropriate (systems, algorithms).
There may be inherent biases in the algorithms.
The AI could become unreliable due to a glitch in the system.
There might be a new situation never anticipated by the AI system.
The AI system may be seen as a form of monitoring, concerns about the loss of privacy.
There is the problem of accountability. At what point can the AI system, or the programmers, etc., be held accountable for an error?

In part (c) of this question it is expected there will be a balance between the terminology related to digital systems and the terminology related to social and ethical impacts.

Keywords: policies, laws, regulations, automation, reliability, privacy, monitoring, surveillance, trust, transparency, accountability, algorithms, bias, change, power, systems, values, ethics

Refer to SL/HL paper 1, part c markbands when awarding marks. These can be found under the "Your tests" tab > supplemental materials > Digital society markbands and guidance document.