AI105 WPR1 STUDY

Artificial intelligence (AI)

Computer systems that perform tasks associated with human intelligence, such as recognizing patterns, understanding language, and making decisions.

Generative AI (GenAI)

AI that creates new content, such as text, images, audio, video, or code, based on patterns learned from training data.

Large language model (LLM)

A model trained on large amounts of language data that generates text by predicting tokens based on context.

How are AI, GenAI, and LLMs related?

AI is the broad category; GenAI is AI that generates content; LLMs are a type of model commonly used for generating language.

Machine learning (ML)

A branch of AI in which systems learn patterns from data instead of relying only on explicitly programmed rules.

Artificial narrow intelligence

AI designed to perform specific tasks; it does not have general human-level intelligence across all domains.

Artificial general intelligence (AGI)

A proposed form of AI that could perform a broad range of intellectual tasks at a level comparable to humans.

Artificial superintelligence

A proposed form of AI that would greatly exceed human intellectual abilities across a broad range of tasks.

What elements can make a prompt effective?

Task, context, example, persona, format, and tone.

Task in a prompt

A clear statement of what the AI should accomplish.

Context in a prompt

Background information, constraints, and relevant details the AI needs.

Example in a prompt

A sample showing the desired kind of answer or output.

Persona in a prompt

The role or perspective the AI should adopt, such as a tutor or editor.

Format in a prompt

The requested structure of the output, such as a table, paragraph, or list.

Tone in a prompt

The requested style of communication, such as professional, friendly, or concise.

Other considerations when constructing a prompt

Specify the audience, scope, constraints, and success criteria; avoid ambiguity and unnecessary sensitive information.

AI hallucination

An AI-generated statement that is false or unsupported but presented as if it were accurate.

Examples of AI hallucinations

Invented citations, nonexistent sources, fabricated quotations, incorrect dates, or unsupported claims about a webpage.

Four steps of the AI105 Problem-Solving Process

Define, Design, Create, Validate.

Step 1: Define

Identify the problem, intended audience, desired outcome, requirements, constraints, and criteria for success.

Step 2: Design

Plan an approach and choose appropriate tools, resources, and methods.

Step 3: Create

Carry out the plan and produce the intended solution or product.

Step 4: Validate

Check whether the result is accurate, appropriate, and meets the requirements; revise as needed.

How does Define differ across tasks?

Research emphasizes the question and evidence needed; communication emphasizes audience and purpose; technical work emphasizes required behavior and constraints.

How does Validate differ across tasks?

Research requires checking claims and sources; communication requires checking facts, tone, and audience fit; technical work requires testing functionality.

How should risk affect validation?

Higher-risk tasks require stronger checking, more reliable evidence, and greater human oversight.

Advanced search techniques

Use exact phrases, specific keywords, domain restrictions, excluded terms, file-type filters, and date filters.

Quotation marks in a search

Search for an exact phrase, such as "artificial intelligence".

site: search operator

Restricts results to a particular website or domain, such as site:westpoint.edu.

Minus sign in a search

Excludes a term, such as jaguar -car.

filetype: search operator

Restricts results to a file type, such as filetype:pdf.

How does a search engine process information?

It searches an index of webpages and ranks results based on relevance and other signals.

How does a search engine deliver results?

It usually provides links and snippets so the user can inspect the original sources.

How does a standard LLM query process information?

It generates a response using learned patterns and the provided context; access to current sources depends on its available tools.

How does a standard LLM query deliver results?

It usually provides a direct synthesized answer that may contain errors or unsupported claims.

How does LLM Deep Research process information?

It generally performs multiple searches, reads sources, and synthesizes findings through a longer research process.

How does LLM Deep Research deliver results?

It generally provides a more detailed report with citations, which still require validation.

When is a search engine useful?

When you want original sources, direct links, or control over which sources you inspect.

When is a standard LLM query useful?

For quick explanations, brainstorming, drafting, or working with information you provide.

When is LLM Deep Research useful?

For complex questions that require gathering and comparing information from multiple sources.

Digital footprint

The trail of data created by a person's digital activities.

Examples of a digital footprint

Social media posts, searches, location records, purchases, account registrations, and website activity.

Active versus passive digital footprints

Active footprints come from information deliberately shared; passive footprints come from data collected through activities or tracking.

Deepfake

AI-generated or manipulated media that makes someone appear to say or do something they did not.

Possible ramifications of a deepfake

Fraud, misinformation, reputational damage, impersonation, and reduced trust in authentic media.

Digital privacy

Control over how personal digital information is collected, accessed, used, and shared.

Principles behind phishing emails

Attackers impersonate trusted sources and exploit urgency, fear, curiosity, or authority to persuade recipients to reveal information or take unsafe actions.

Strategic communication

Purposeful communication designed for a specific audience to achieve a desired outcome.

Strategic communication and Define

Identify the purpose, audience, key message, desired response, and constraints.

Strategic communication and Design

Choose the message structure, tone, medium, and delivery approach.

Strategic communication and Create

Produce the written, visual, audio, or video message.

Strategic communication and Validate

Check accuracy, clarity, tone, audience fit, and whether the message supports the intended outcome.

Input devices

Hardware used to send data or commands into a computer, such as keyboards, microphones, and cameras.

Output devices

Hardware that presents information from a computer, such as monitors, speakers, and printers.

RAM

Random-access memory: fast, temporary working storage for programs and data currently in use; its contents are normally lost when power is removed.

Hard drive

Persistent storage that retains files and programs when power is removed; SSDs also provide persistent storage.

Peripheral devices

Additional hardware connected to a computer, such as printers, external drives, keyboards, and webcams.

CPU

Central processing unit: executes instructions, performs calculations, and coordinates computing operations.

Binary

A base-2 number system using only 0 and 1.

Bit

A binary digit that can have a value of 0 or 1.

Byte

A group of 8 bits.

File extension

A filename ending, such as .pdf or .xlsx, that identifies the expected file type and helps the operating system choose an application.

Does changing a file extension convert the file?

No. Renaming the extension does not change the underlying data or actual file format.

Computer network

A group of connected devices that exchange data and share resources.

Ethernet

A networking technology commonly used for wired local connections.

Wi-Fi

A wireless networking technology commonly used to connect devices to a local network.

Bluetooth

A wireless technology commonly used for short-range connections between devices.

Cellular network

A wireless network using cell towers to provide connectivity over broad geographic areas.

Bandwidth

The data-carrying capacity of a connection, commonly measured in bits per second.

Latency

The delay involved in transmitting data or receiving a response.

Range

The distance over which a connection can operate effectively.

Bandwidth versus latency

Bandwidth concerns how much data can be transferred per second; latency concerns how long transmission or a response takes.

Network packet

A small unit of transmitted data containing part of a message plus addressing and control information.

How are data transmitted in packets?

Data are divided into packets, forwarded across network links, and reassembled by the receiving system; packets may follow different routes.

Relative spreadsheet reference

A cell reference that adjusts when a formula is copied, such as A1.

Absolute spreadsheet reference

A cell reference that stays fixed when a formula is copied, such as $A$1.

When should relative references be used?

When a copied formula should use different cells in each row or column.

When should absolute references be used?

When formulas should keep referring to a fixed cell, such as a tax rate or conversion factor.

Mixed spreadsheet reference

A reference that fixes either the column or row, such as $A1 or A$1.

Internet

The global network of interconnected networks that supports services such as the web, email, and file transfer.

World Wide Web

A system of linked webpages and resources accessed over the Internet using web browsers.

Internet versus World Wide Web

The Internet is the network infrastructure; the web is one service that operates on it.

Domain Name System (DNS)

A service that translates domain names, such as westpoint.edu, into IP addresses used to locate network resources.

Uniform Resource Locator (URL)

The address of a resource, often including a scheme, domain name, and path.

HTML

Defines the structure and content of a webpage.

CSS

Controls the appearance and layout of a webpage.

JavaScript

Adds programming logic, dynamic behavior, and interactivity to webpages.

Surface Web

Web content that standard search engines can generally index and find.

Deep Web

Web content not indexed by standard search engines, such as private email, account pages, and restricted databases.

Dark Web

A portion of the Deep Web accessed through specialized software or configurations, such as Tor.

Deep Web versus Dark Web

The Deep Web includes all nonindexed web content; the Dark Web is a smaller subset requiring specialized access.

Client-server model

A client requests a service or resource, and a server processes the request and provides a response.

Client-server model in web browsing

A browser acts as a client and requests webpages or other resources from web servers.

Client-server model in cloud storage

A client application uploads, retrieves, or synchronizes files stored on remote servers.

Client-server model in online applications

The client provides a user interface while servers may process requests, run application logic, and store data.

Physical layer of cyberspace

The physical infrastructure, including computers, cables, routers, servers, and radio equipment.

Logical layer of cyberspace

The software, protocols, addressing, and logical connections that allow systems to communicate.

Cyber-persona layer of cyberspace

Digital identities and representations, such as accounts, usernames, and email addresses.

Cyber threat

A potential cause of harm to a system, network, or information.

Threat actor

A person, group, or organization that carries out or attempts a harmful cyber action.