Recent Trends in Technology

This unit is the one that dates fastest. The machines are the same as in Unit 2; what has changed is where the computing happens (someone else's data centre), how much data it chews (petabytes), how it is programmed (learned from data rather than coded by hand) and what it is attached to (your watch, a tractor, a hospital). The figures quoted below are stated as of 2025–26 — quote the year whenever you quote a number in the exam, because the number will be wrong by the time you are in college.

7.1 Artificial Intelligence (AI) and Robotics

The term was coined at the Dartmouth conference of 1956 (John McCarthy). Milestones since: Deep Blue beat the world chess champion Kasparov (1997), AlphaGo beat the Go champion Lee Sedol (2016), the transformer architecture was published (2017), and ChatGPT put generative AI into ordinary hands (November 2022). Today's headline systems are large language models (LLMs) — networks with billions of parameters trained on enormous text and image collections, which generate new text, code or pictures rather than only classifying old ones. Since 2024 the industry has moved on to multimodal models (text + image + audio in one model) and agentic AI, where a model is allowed to use tools and take multi-step actions on a user's behalf.

AI is classified by capability into narrow AI (ANI — good at one task; every system in use today, from Google Translate to a face unlock), general AI (AGI — human-level across all tasks; does not yet exist) and super AI (beyond human ability; hypothetical). Its working branches are machine learning (supervised, unsupervised and reinforcement learning), natural language processing, computer vision, expert systems, speech recognition and robotics.

Robotics is the branch of engineering and computer science that designs, builds and programs robots. A robot is a programmable machine that senses its environment, processes what it senses and acts on the physical world. The word comes from Karel Čapek's 1920 play R.U.R. (Czech robota, forced labour); Asimov's three laws of robotics date from 1942; the first industrial robot, Unimate, went to work at General Motors in 1961.

Robot type	Where used	Example
Industrial arm	welding, painting, assembly lines	car factories
Mobile / AGV	moving goods in a warehouse	Amazon warehouse robots
Humanoid / service	reception, teaching, hospitality	the robot waiter built by Paaila Technology for a Kathmandu restaurant (2018)
Medical	minimally invasive surgery	da Vinci surgical system
Aerial (drone/UAV)	survey, mapping, agriculture, delivery	drone mapping after the 2015 Gorkha earthquake
Exploration	places humans cannot go	Mars rovers, mine and reactor inspection

Applications of AI you can safely list in an exam: medical diagnosis from X-rays, weather and flood forecasting, speech recognition and translation, recommendation systems, fraud detection in banking, self-driving vehicles, agriculture (pest detection from leaf photographs), and chatbots for customer service.

Advantages: works 24 hours without fatigue, handles jobs that are dangerous or repetitive, very fast on large data, fewer human errors, consistent decisions. Disadvantages: high cost of setup and training, job displacement, no common sense or emotion, results can be biased because the training data was biased, poor accountability when it gets things wrong, and a real privacy cost.

7.2 Cloud Computing

The NIST definition lists five essential characteristics: on-demand self-service, broad network access, resource pooling (multi-tenancy), rapid elasticity, and measured service (pay-per-use metering). The enabling technology underneath is virtualisation — one physical server split into many virtual machines or containers.

Service model	What you rent	You manage	Examples
IaaS — Infrastructure as a Service	virtual machines, storage, network	OS, runtime, application, data	AWS EC2, Google Compute Engine, Azure VMs
PaaS — Platform as a Service	a ready platform to deploy code on	application and data only	Google App Engine, Heroku, Azure App Service
SaaS — Software as a Service	finished software in a browser	your data only	Gmail, Google Docs, Zoom, Microsoft 365

Advantages. No capital cost for servers; scale up or down in minutes; access from anywhere; automatic backup and disaster recovery; the provider patches and secures the hardware; a small Nepali startup can rent the same infrastructure as a multinational.

Disadvantages. Useless without a reliable internet connection; recurring cost that never ends; data is physically in someone else's country, which raises legal and privacy questions; vendor lock-in makes moving away expensive; a provider outage takes your service down with it.

Industry trackers put the 2025 global cloud-infrastructure market at roughly AWS 30 %, Microsoft Azure 22 %, Google Cloud 12 %, with the rest split among Alibaba, Oracle, IBM and others. In Nepal, government systems are hosted in the Government Integrated Data Centre run by the National Information Technology Centre, and local providers now sell cloud servers hosted inside Nepal, which keeps data under Nepali law.

7.3 Big Data

The standard answer is the five Vs (the first three are Doug Laney's 2001 list; veracity and value were added later):

V	Meaning	Example
Volume	the sheer quantity — terabytes to zettabytes	a telecom's call detail records for a year
Velocity	the speed at which data arrives and must be processed	Fonepay transactions, sensor streams, tweets
Variety	structured, semi-structured and unstructured together	tables + JSON logs + photos + voice
Veracity	uncertainty — how much of it is noisy, biased or wrong	user-entered addresses, GPS drift
Value	the usable insight extracted from all of it	which route to add buses to

One zettabyte is 10^{21} bytes. Estimates put the data created worldwide at around 180 ZB for 2025, up from about 64 ZB in 2020 — roughly a doubling every two to three years.

Big data is handled by distributed tools instead of one big machine: Hadoop (the HDFS file system plus MapReduce processing), Apache Spark (in-memory, much faster), and NoSQL databases such as MongoDB and Cassandra that drop the strict relational schema in exchange for scale.

Applications: telecom network planning and churn prediction, bank fraud detection, retail recommendation, disease outbreak tracking, weather and landslide prediction from satellite data, traffic management, and analysing remittance flows — for a country where remittance is a huge share of GDP, that last one matters.

7.4 Virtual Reality

A VR system needs four things: a display worn on the head (HMD) with one image per eye to create stereo depth; tracking of the head and hands so the view updates as you move; an interaction device (controllers, gloves, treadmill); and a graphics engine fast enough to redraw both eyes 90 times a second — if the picture lags behind the head, the user feels sick within minutes.

VR is classified as non-immersive (a 3-D world on an ordinary screen, such as a flight-simulator game), semi-immersive (a large projected screen or CAVE, used in driving and flight training) and fully immersive (head-mounted display with tracking). The current consumer devices are Meta Quest 3 (2023) and Apple Vision Pro (2024); in the browser, the WebXR standard lets a web page open a VR view without installing anything.

Field	Use of VR
Education	virtual laboratories and field trips; dissect a frog without a frog
Medicine	surgical training, therapy for phobias and PTSD, rehabilitation
Engineering and architecture	walk through a building before it is built
Military and aviation	flight and combat simulators — cheap failure
Entertainment	games, 360° films, virtual concerts
Heritage and tourism	a virtual walk through Pashupatinath or a reconstructed Kathmandu Durbar Square before 2015

Advantages: learning by doing without danger or cost; mistakes are free; distance disappears. Disadvantages: expensive hardware; motion sickness; eye strain and isolation with long use; content is costly to produce.

7.5 e-commerce, e-medicine and e-governance

e-commerce is the buying and selling of goods and services, and the transfer of money and data, over electronic networks. Its models are named by who is at each end:

An e-commerce system needs a product catalogue, a shopping cart, a payment gateway and a delivery chain. In Nepal payment runs through eSewa, Khalti, IME Pay, connectIPS and Fonepay QR, and the legal basis for digital signatures and electronic records is the Electronic Transactions Act, 2063 BS (2008). Advantages: open 24 × 7, no shop rent, national reach from a Jhapa village, easy price comparison. Disadvantages: you cannot touch the goods, delivery in the hills is slow and costly, online fraud and card theft, and disputes are hard to settle.

e-medicine (telemedicine) is the delivery of health care at a distance using information and communication technology: teleconsultation by video, transmission of X-rays and ECGs to a specialist (tele-radiology), remote monitoring of patients with wearable sensors, electronic health records, and online pharmacy. For Nepal the argument is geography — a specialist in Kathmandu can advise a health post in Humla in minutes instead of a two-day walk plus a flight, and Nepal's government telemedicine programme has linked district hospitals to central hospitals for over a decade. Limits: no physical examination, needs reliable bandwidth and power, patient-privacy risk, and legal uncertainty about who is responsible for a remote diagnosis.

e-governance is the use of ICT by government to deliver services and information to citizens, businesses and its own offices. Its four interaction types are G2C (government to citizen — online passport application, vehicle-tax payment, the Nagarik App), G2B (government to business — online company registration, PAN/VAT filing with the Inland Revenue Department), G2G (between offices — the national ID database shared across ministries) and G2E (government to employee — payroll, transfers, records). Nepal's programme is set out in the Digital Nepal Framework (2019), and the Government Integrated Data Centre hosts the systems. Benefits: transparency and less corruption, fewer visits to offices, faster service, records that cannot quietly disappear. Barriers: the digital divide (electricity, internet and literacy), cyber-security, resistance inside offices, and the cost of keeping systems running after the donor project ends.

7.6 Mobile Computing

It has three parts: mobile hardware (smartphone, tablet, laptop, wearable), mobile software (Android, iOS, apps, mobile browsers) and mobile communication (the cellular network, Wi-Fi, Bluetooth, satellite, and the protocols and handover rules that keep a call alive as you move between cells).

In Nepal, 4G reaches most district headquarters, Nepal Telecom began 5G trials in 2023, and the number of active mobile SIMs has for some years exceeded the population — the Nepal Telecommunications Authority reports mobile penetration above 100 %, which means many people carry two SIMs, not that everyone has a phone.

Advantages: work from anywhere, instant communication, location-based services, cheap for a country that never built copper telephone lines. Limitations: battery life, small screen and keyboard, patchy coverage in the mountains, lower bandwidth and higher interference than cable, and security — an open Wi-Fi is trivially easy to eavesdrop on.

7.7 Internet of Things (IoT)

The term was coined by Kevin Ashton in 1999. An IoT system is usually drawn as four layers:

Common IoT protocols are MQTT and CoAP (light messaging for small devices), Zigbee, Bluetooth Low Energy and LoRaWAN (long range, tiny power), and NB-IoT over the cellular network. Because every thing needs its own address, IoT is a major reason for moving to IPv6.

Applications: smart homes (lights, locks, cameras), smart agriculture (soil and weather sensors, drip irrigation), health wearables, smart meters, vehicle tracking, air-quality monitoring in the Kathmandu valley, glacial-lake and river level monitoring for flood warning, and industrial machine monitoring.

Advantages: automation, real-time monitoring, energy and water saved, early warning, better data for decisions. Challenges: weak security (cheap devices shipped with default passwords have been hijacked into botnets), privacy, no single standard, power supply in the field, and the cost of maintaining thousands of scattered devices.

Chapter summary

• AI builds machines that do jobs needing intelligence (coined 1956); machine learning ⊂ AI, deep learning ⊂ ML, generative AI/LLMs ⊂ deep learning. Types: narrow (all real systems today), general, super.
• A robot senses, decides and acts: sensors → controller → actuators → end effector, with feedback. Types: industrial, mobile, humanoid, medical, drone, exploration.
• Cloud computing rents computing over the internet, pay-per-use. Service models IaaS, PaaS, SaaS; deployment models public, private, community, hybrid; five NIST characteristics; it trades capital cost for dependence on a provider and a link.
• Big data = volume, velocity, variety, veracity, value; handled by Hadoop, Spark and NoSQL rather than one server. World data creation was about 181 ZB in 2025.
• VR replaces sensory input with a computer-generated 3-D world; needs an HMD, tracking, interaction devices and a fast engine; classified non-immersive, semi-immersive, fully immersive; AR adds to the real world, MR anchors to it.
• e-commerce models: B2B, B2C, C2B, C2C, supported by payment gateways and, in Nepal, the Electronic Transactions Act 2063. e-medicine delivers care at a distance. e-governance interacts as G2C, G2B, G2G, G2E.
• Mobile computing = mobile hardware + software + wireless communication; generations 1G to 5G, peak rates from 2.4 kbps to about 10 Gbps, latency falling from ~50 ms (4G) to ~5 ms (5G).
• IoT (Kevin Ashton, 1999) connects physical things through four layers — perception, network, processing, application — using MQTT, Zigbee, LoRaWAN and IPv6; its biggest weakness is security.

Practice questions

Group A — Multiple choice (1 mark each)

1. Gmail is an example of which cloud service model? [1]aIaaSbPaaScSaaSdDaaS
2. Which of the following is not one of the five Vs of big data? [1]avolumebvelocitycvirtualisationdveracity
3. The term "Internet of Things" was coined by [1]aKevin AshtonbTim Berners-LeecJohn McCarthydDoug Laney
4. Selling a used bicycle to another person through Hamrobazar is an example of [1]aB2BbB2CcC2CdG2C
5. A technology that adds computer-generated objects to a view of the real world is called [1]avirtual realitybaugmented realitycartificial intelligencedcloud computing
6. Paying vehicle tax through a government website is an example of [1]aG2GbG2BcG2CdG2E

Answers to Group A

1. (c) — finished software used in a browser; you manage only your data.

2. (c) — virtualisation is a cloud technology, not a V of big data.

3. (a) — Kevin Ashton used the phrase in 1999 while working on RFID.

4. (c) — consumer to consumer, through a marketplace.

5. (b) — AR adds to reality; VR replaces it.

6. (c) — a government service delivered to a citizen.

Group B — Short answer (5 marks each)

1. Define cloud computing. Explain its three service models with one example of each. [5]
2. What is big data? Explain the five Vs with one example of each. [5]
3. Define IoT and explain its four-layer architecture with a suitable example. [5]
4. What is e-governance? Explain its four types with a Nepali example of each. [5]
5. Differentiate between virtual reality and augmented reality, and state any three applications of VR. [5]
6. A college server costs Rs 6,00,000 and lasts 5 years; running it costs Rs 3,500 per month in electricity and Rs 2,500 per month in administration. An equivalent cloud server costs Rs 12,000 per month. Which is cheaper, and by how much per year? [5]

Answers to Group B

1.–5. Definitions and tables from §7.2, §7.3, §7.7, §7.5 and §7.4; one mark for the definition, one per model/V/layer/type with its example.

6. Hardware per month = 6{,}00{,}000 / 60 = \text{Rs }10{,}000. On-premises total = 10{,}000 + 3{,}500 + 2{,}500 = \text{Rs }16{,}000 per month. The cloud at Rs 12,000 is cheaper by Rs 4,000 per month, i.e. 4{,}000 \times 12 = \textbf{Rs } 48{,}000 per year. (Mention also the non-monetary factors: no load-shedding risk, but a permanent dependence on the internet link.)

Group C — Long answer (8 marks each)

1. Define artificial intelligence and robotics. Draw the block diagram of a robot and explain its components. State any four applications and any three disadvantages of AI. [8]
2. Define cloud computing and explain its service and deployment models. Compare cloud computing with maintaining your own server on cost, scalability, security and reliability, using a numerical example. [8]

Answers to Group C

1. Definitions (2), block diagram sensors → controller → actuators → end effector with the feedback loop and power supply (2), explanation of each component (2), four applications (1), three disadvantages (1) — all from §7.1.

2. Definition with the five NIST characteristics (2), IaaS/PaaS/SaaS with examples (2), public/private/community/hybrid (2), comparison (2): cost — on-premises is a large capital cost plus running cost while cloud is operational only, as computed in Worked example 7.1 where the cloud saved Rs 2,876 a month; scalability — minutes in the cloud against weeks to buy hardware; security — the provider has better physical security and patching, but your data sits under another country's law; reliability — providers publish 99.9 %+ availability, better than a single office server on an unreliable supply, but an outage or a cut internet link stops everything at once.