Big picture
Emerging technologies combine computation, data, connectivity and automation in new applications. A weather warning system might collect sensor readings, store them remotely and analyse patterns before notifying farmers. Understanding the components is more useful than memorising fashionable product names.
The technologies in this chapter overlap but are not interchangeable. Artificial intelligence is not the same as a robot, cloud computing is not simply a website, and a blockchain cannot guarantee that entered information is truthful. For each application, connect the requirement, the mechanism, the benefit and a realistic limitation.
Learning goals
Distinguish AI, machine learning and rule-based automation.
Match learning types to supplied training information.
Explain IoT, robotics, edge and cloud roles.
Compare cloud services, AR, VR and blockchain.
Evaluate benefits with privacy, bias and reliability limits.
Core ideas
1. Artificial intelligence and machine learning
Artificial intelligence, or AI, concerns systems performing tasks associated with intelligent behaviour, such as recognising patterns, interpreting language or selecting actions. Machine learning, or ML, is an approach in which models learn useful patterns from data rather than having every decision rule explicitly specified.
A hand-written rule that flags temperature above a threshold is automation, but it does not necessarily involve learning. A model trained on labelled examples of healthy and diseased leaves uses data-driven learning. Its predictions can be wrong when training data are unrepresentative or conditions differ. Distinguish training from use: fitting a model uses examples to determine its behaviour, while inference applies the learned model to a new input. More data alone do not guarantee better predictions if the labels are unreliable.
2. Learning types and pattern-recognition tasks
Supervised learning uses examples with known target labels or values. Classifying a message as spam uses categories, while predicting rainfall amount is a numeric prediction task. Unsupervised learning finds structure without supplied target labels, such as grouping similar reading preferences. Reinforcement learning studies action selection using feedback or rewards from interaction.
Natural language processing works with human language, while computer vision works with visual information. A speech interface also involves recognising audio and interpreting language. Match the description to the task rather than the device's appearance: a camera connected to a laptop is not automatically an AI system. For an exam scenario, identify what is being predicted or grouped, what training information exists and how a result would be checked before being relied upon.
3. Robotics and automation
A robot is a programmable physical system that can sense and act in its environment. Sensors supply observations, a controller determines actions and actuators produce movement or other physical effects. AI can support a robot's decisions, but a robot can also follow fixed instructions without machine learning.
An automated greenhouse can use a moisture sensor and a pump. If the controller switches the pump on below a threshold, the system is an example of sensor-based automation. A learned system might predict irrigation needs from several observations instead. Physical automation needs safeguards because a wrong output can affect equipment or people. A sensible answer includes feedback and failure handling rather than merely claiming that removing human involvement always improves safety.
4. Internet of Things and edge processing
The Internet of Things, or IoT, connects physical objects that sense, exchange data or act through networked systems. A typical flow is sensor, local controller or gateway, network, storage or analysis, and an action or dashboard. Connectivity makes remote monitoring possible but creates security and privacy responsibilities.
Edge computing processes data near its source rather than sending every observation to a distant service first. A flood sensor could trigger a local warning even when external connectivity is interrupted. This can reduce latency and bandwidth demands, while central services may support broader analysis. Edge and cloud can cooperate. Discuss power, connectivity, authentication, updates and sensor calibration: an inaccurate reading remains inaccurate even if transmitted quickly and displayed attractively.
5. Cloud computing and service models
Cloud computing provides computing resources over a network with on-demand access and managed scaling. It can include processing, storage and applications. A service's remote location alone does not explain every cloud feature, and internet access is still a practical dependency for many uses.
Model | Main service supplied | Example responsibility |
|---|---|---|
IaaS | Virtual computing infrastructure | Customer manages much of the software stack |
PaaS | Managed application platform | Customer develops and deploys application code |
SaaS | Ready-to-use application | Customer uses and configures the application |
An online document editor is a SaaS-style example; renting a virtual machine is IaaS-style. Compare what the provider manages and what the user must manage. The models do not imply that customers have no security responsibilities. Access control, data handling and appropriate configuration still matter.
6. Deployment models and big data
Cloud deployment models include public, private, community and hybrid arrangements. Public services are offered to multiple customers, private infrastructure serves a particular organisation, and hybrid designs combine environments. Public cloud does not mean every customer's stored files are publicly readable; permissions remain essential.
Big data describes datasets and processing challenges whose scale or characteristics call for suitable distributed tools and approaches. The commonly taught Vs include volume, velocity and variety; veracity and value add attention to quality and usefulness. A small messy sheet is not automatically big data, and a huge collection is not automatically useful. For transport observations, rapid arrival and mixed sensor formats may matter as much as total size. Identify the actual challenge instead of counting fashionable terms.
7. AR, VR, blockchain and distributed records
Augmented reality, or AR, overlays digital information on a view of the physical world. Virtual reality, or VR, creates an immersive simulated environment. A phone showing labels over a real machine illustrates AR; a headset-based simulated workshop illustrates VR. Both need accessible alternatives and attention to user comfort.
A blockchain is a distributed record structure in which blocks are linked using cryptographic techniques, with agreement mechanisms governing additions. It can make certain alterations evident and support shared records, but it does not prove that an entered real-world fact was correct. A false shipment description can be preserved consistently. Blockchain and cryptocurrency are related in some applications but are not synonyms. Evaluate the trust problem before proposing a distributed ledger as a universal solution.
8. Ethical evaluation and a complete scenario
A responsible technology proposal asks what data are needed, whose data are used, whether errors affect groups differently and who can challenge a decision. Bias can arise from data selection, labelling or system design. Human review is especially important when decisions affect education or access to essential opportunities.
Consider a school greenhouse: IoT sensors collect moisture, edge logic handles immediate pump control, cloud storage retains readings, and an ML model might later predict demand. A dashboard presents trends, while a robot is unnecessary unless physical movement or manipulation is required. Include a manual override, calibrated sensors, restricted access and a plan for network failure. This explanation assigns a specific role to each technology and admits limits. An exam answer earns clarity by justifying a small suitable set of components, rather than listing every technology in the chapter without a purpose.
A reliable approach to questions
Read the command word before choosing an answer. A definition needs the distinguishing feature; a comparison needs matching criteria on both sides; an application needs a reason linked to the situation.
Extract the supplied evidence and assumptions. Keep a small working table rather than relying on memory. If information is absent, state the limit instead of inventing a value or claiming certainty.
Work through the relevant rule one step at a time. For a program, record the changing state; for a scenario, connect the requirement to the feature that meets it. Explain why the closest alternative fails.
Check the result against the original question, including boundary cases, units and wording. Present a concise final answer followed by enough reasoning for another student to reproduce your conclusion.
Connections that matter
Computer fundamentals supply sensors, processors and storage.
Digital society examines bias, privacy and access.
Data projects test whether observations support useful conclusions.
Common mistakes
Equating every automated rule with machine learning: learning requires data-driven model fitting.
Calling public cloud publicly readable storage: access permissions are separate.
Claiming a blockchain guarantees truthful inputs: stored falsehoods remain falsehoods.
Adding every technology to a proposal: justify each role from a requirement.
Revision checklist
I can distinguish AI, machine learning and rule-based automation.
I can match learning types to supplied training information.
I can explain IoT, robotics, edge and cloud roles.
I can compare cloud services, AR, VR and blockchain.
I can evaluate benefits with privacy, bias and reliability limits.