Large language models learn patterns of words by analyzing billions of examples from text across the internet.
AI chat systems can summarize long documents into concise points faster than most humans can read them.
Generative AI can create stories, music, code, and images by predicting the most likely sequence of words or pixels.
LLMs do not think as humans do; they generate responses based on probabilities, not conscious reasoning.
Transformer architecture, the backbone of modern AI, processes input in parallel with layers called attention heads.
Attention mechanisms allow AI to weigh the importance of each word in context, making sense of long passages.
LLMs use embeddings to convert words into numerical vectors in high-dimensional space.
AI can translate between languages by mapping meaning across different word embeddings.
Fine-tuning allows a model to specialize on tasks like legal writing, customer service, or medical research.
Reinforcement learning with human feedback helps align AI responses with user preferences.
Prompt engineering is the art of phrasing instructions to guide AI toward useful results.
Few-shot prompting shows an AI how to answer by giving examples inside the input itself.
Zero-shot prompting asks an AI to do something with no prior examples, testing its adaptability.
Chain-of-thought prompting encourages step-by-step reasoning before producing an answer.
Role prompting can instruct AI to act as a teacher, coder, or storyteller depending on user goals.
LLMs can be misled with adversarial prompts, producing harmful or nonsensical outputs.
Data quality is critical; biased or false training data can produce biased or false results in AI.
Hallucination occurs when AI invents plausible-sounding but incorrect information.
AI can power chatbots that handle customer service 24/7 with natural conversation.
Researchers use AI to assist in drug discovery by predicting protein structures and chemical interactions.
In education, AI can generate personalized study guides and practice questions tailored to each student.
Artists use AI to brainstorm ideas, generate drafts, or remix existing styles.
Coders use AI to debug, explain, or write functions in multiple programming languages.
LLMs can act as translators between technical jargon and plain language for wider audiences.
AI can analyze medical images to spot early signs of disease that doctors might miss.
AI-driven fraud detection scans millions of transactions for unusual patterns in real time.
Military and government agencies explore AI for intelligence analysis and logistics planning.
Overuse of AI in creative work can lead to homogenized, less original cultural output.
Unchecked reliance on AI in decision-making can amplify systemic bias.
LLMs cannot guarantee privacy; text provided in prompts may be logged or reused for training.
AI systems consume significant electricity, raising environmental concerns.
Smaller local models can run on laptops or Raspberry Pi boards for privacy and offline use.
Open-source LLMs allow hobbyists and researchers to adapt AI without corporate control.
Alignment research aims to keep AI behavior consistent with human values and ethics.
Prompt injection is a new attack vector where malicious input changes how AI behaves.
AI companions and chatbots can provide comfort, but may create dependency or isolation.
AI image generators can create convincing forgeries, raising issues of trust in media.
AI can help design efficient solar panels, batteries, and renewable energy systems.
Some companies use LLMs to automate legal contract review, saving time but risking subtle errors.
Fact-checking with AI requires multiple sources; one model alone may hallucinate false confidence.
AI is not self-aware; claims of consciousness are projections by humans onto machine output.
AI ethics frameworks stress transparency, accountability, and fairness in deployment.
AI tools can summarize meetings, emails, and chats, making teams more efficient.
Healthcare uses of AI include predictive models for patient risk and hospital resource planning.
AI can generate synthetic data to protect privacy while training other machine learning models.
Misinformation spread by AI deepfakes can destabilize trust in institutions.
Future protocols may give AIs digital signatures so generated content is traceable.
A well-crafted system prompt sets the boundaries of an AI assistant’s behavior.
LLMs function like autocomplete on steroids, building responses one token at a time.
The biggest danger of AI is not superintelligence but misaligned human use and misuse.
AI can analyze satellite data to track deforestation, urban growth, and climate change.
Robotics combined with AI enables machines to adapt to unpredictable environments.
Speech recognition models allow AI to transcribe and translate conversations instantly.
AI-generated music can mimic genres or invent entirely new sonic landscapes.
A small tweak in a prompt can radically change the tone of an AI’s answer.
Vector databases store embeddings that let LLMs perform semantic search over massive archives.
Retrieval-augmented generation combines AI with external knowledge sources to boost accuracy.
LLMs can simulate dialogue between historical figures for education and entertainment.
AI can read and draft computer code across dozens of programming languages.
Content moderation systems use AI to filter harmful text, images, and video at scale.
When AIs are trained on their own outputs, feedback loops of error can emerge.
AI can analyze scientific papers to find hidden connections between research fields.
Synthetic voices powered by AI can now mimic human tone and inflection convincingly.
Misinformation amplified by AI chatbots could influence elections if unchecked.
Companies embed LLMs into productivity suites to automate email, reports, and slides.
AI can personalize therapy chatbots that practice cognitive behavioral techniques.
Smaller specialized models are more efficient than massive general-purpose ones.
Self-improving code assistants can speed up software development cycles.
AI can flag suspicious insider trading patterns on stock exchanges.
Neural networks train on GPUs and TPUs, chips optimized for parallel computation.
Tokenization is how words and phrases are broken into chunks for AI to process.
AI often mirrors the style and tone of the user’s prompt, making framing critical.
LLMs can explain math step by step, but may miscalculate without verification.
The boundary between human writing and AI-generated text is increasingly hard to detect.
AI literacy is becoming a core skill, like reading or basic computer knowledge.
AI-powered design tools can prototype websites, logos, and layouts in seconds.
Chatbots can reduce loneliness but may also encourage people to avoid human contact.
AI video generators raise concerns about actors’ rights and digital likeness.
AI can convert handwriting to text and generate searchable archives of notebooks.
The context window of a model determines how much information it can consider at once.
Guardrails like content filters and moderation policies are layered onto commercial AI systems.
Some AIs can now take in multiple inputs like text, images, and audio at the same time.
The arms race between AI safety research and adversarial misuse is ongoing.
AI assistants may appear confident but can be disastrously wrong in high-stakes tasks.
AI can reconstruct missing text from damaged historical manuscripts.
Adversarial images can fool AI vision systems with patterns invisible to humans.
Open-weight models allow full inspection of architecture and training data sources.
The rise of AI watermarking may help track generated content across the web.
Large-scale training datasets often include billions of tokens scraped from online sources.
The cost of training frontier LLMs can exceed hundreds of millions of dollars.
AI can draft entire lesson plans for teachers customized to student levels.
Privacy-focused LLMs run offline, ensuring sensitive data never leaves the device.
AI systems increasingly interact with APIs, letting them book flights or pull live data.
The greatest power of AI is augmentation, not replacement, of human creativity.
Computers – Python3
Python code is case sensitive
Indentation is used to define blocks of code
Use the print function to display output
Variables do not need explicit type declarations
Strings are written inside single or double quotes
Use triple quotes for multiline strings
The input function reads text from the user
Comments start with a hash symbol
Use meaningful variable names for clarity
Integers and floats are two basic numeric types
Division with a single slash returns a float
Division with two slashes returns an integer result
The percent sign gives the remainder of a division
Comparison operators return True or False
Boolean values are True and False with capitals
The if statement controls conditional execution
Elif means else if for multiple conditions
Else handles all remaining cases
Loops repeat blocks of code
The for loop iterates over a sequence
The while loop repeats while a condition is True
Break exits a loop early
Continue skips the rest of the current loop iteration
Functions are defined with the def keyword
Return sends a value back from a function
Parameters allow you to pass values into functions
Default parameter values can be specified
Use docstrings to explain what a function does
Lists hold ordered collections of items
List indices start at zero
Negative indices count from the end
Slicing selects parts of a list
Tuples are like lists but immutable
Dictionaries store key value pairs
Sets store unique unordered items
Use len to get the length of a sequence
Import statements bring in modules and libraries
Use as to give a module a short alias
Exceptions are errors detected at runtime
Use try and except to handle exceptions
Classes define new object types
Use self to access instance variables inside a class
Use init to set up an object when created
Inheritance lets one class build on another
Use pass as a placeholder for empty code blocks
PEP 8 is the style guide for Python code
Read error messages carefully they often explain the problem
Practice by writing small scripts and experiments
Use online documentation and help functions to learn syntax
Build projects to connect concepts into real skills
Computers – Terminal Commands
pwd prints the current working directory
ls lists the files and folders in the current directory
cd changes the current directory
cd .. moves up one directory level
mkdir creates a new directory
rmdir removes an empty directory
touch creates an empty file or updates a files timestamp
cp copies files or directories
mv moves or renames files or directories
rm removes files or directories
cat displays the contents of a file
less shows file contents one screen at a time
head shows the first lines of a file
tail shows the last lines of a file
tail dash f follows a file as it grows
echo prints text or variable values to the screen
man shows the manual page for a command
which shows the path of an executable command
find searches for files in a directory tree
grep searches inside files for matching text
history shows previously used commands
clear clears the terminal screen
df shows disk space usage
du shows the size of files and directories
top shows running processes in real time
ps shows currently running processes
kill sends a signal to stop a process
chmod changes file permissions
chown changes file ownership
nano opens a simple text editor
vi opens the vi text editor
ssh connects securely to another computer
scp securely copies files between computers
wget downloads files from the web
curl transfers data from or to a server
tar archives and extracts files
gzip compresses files
gunzip decompresses files
zip compresses files into a zip archive
unzip extracts files from a zip archive
ping tests network connection to a host
ifconfig shows or configures network interfaces
whoami prints the current logged in user
uname shows system information
date displays or sets the system date and time
uptime shows how long the system has been running
exit closes the terminal session
sudo runs a command as superuser
apt get installs or removes software on Debian based systems
Computer Science 1
Define an algorithm. An algorithm is a step by step set of instructions to solve a problem.
Explain a program. A program is a collection of algorithms written in code for a computer to run.
Describe data. Data is information stored and processed by a computer.
How small is a bit. A bit is the smallest unit of data and can be zero or one.
How many bits form a byte. Eight bits form a byte which can store a character or small number.
Name the language of computers. The language of computers is binary which uses only zeros and ones.
Give an example of hardware. Hardware is the physical part of a computer like the CPU memory and disk.
Give an example of software. Software is the set of instructions that run on hardware.
Where are instructions executed. Instructions are executed in the CPU the brain of the computer.
How does memory work. Memory stores data and instructions temporarily while the CPU works.
What does storage do. Storage devices hold data permanently until erased.
Define an operating system. An operating system manages hardware and software resources.
How are files organized. A file system organizes data into files and directories.
Name two input devices. Keyboards and mice are input devices that send information to the computer.
Name two output devices. Monitors and printers are output devices that show results from the computer.
How are computers connected. Computers are connected through networks to share data.
Explain the internet. The internet is a global network of connected computers.
What do protocols control. Protocols are rules that govern how data is sent across a network.
Define a database. A database organizes large amounts of structured data.
How does a compiler work. A compiler translates code into machine instructions.
How does an interpreter work. An interpreter executes code line by line without compiling.
Name two high level languages. Python and Java are examples of high level languages closer to human language.
Name a low level language. Assembly is a low level language close to machine code.
What qualities make a strong algorithm. A strong algorithm is correct efficient and easy to understand.
How do we measure algorithm speed. Algorithm speed is measured with time complexity.
How do we measure algorithm memory use. Algorithm memory use is measured with space complexity.
Define a data structure. A data structure organizes data for efficient access and modification.
What is an array used for. An array stores a fixed sized collection of items of the same type.
How is a list different from an array. A list is an ordered collection that can grow or shrink.
How does a stack store data. A stack uses last in first out access.
How does a queue store data. A queue uses first in first out access.
What shape describes a tree. A tree organizes data in a branching hierarchy.
How does a graph represent information. A graph stores connections or relationships between objects.
Explain recursion. Recursion is when a function calls itself to solve smaller problems.
Explain iteration. Iteration repeats steps until a condition is met.
What does abstraction achieve. Abstraction hides complex details and shows only essentials.
What does encapsulation achieve. Encapsulation bundles data and methods together in objects.
How does inheritance help programming. Inheritance lets new classes reuse code from old classes.
How does polymorphism help programming. Polymorphism lets one interface work with many implementations.
Define cybersecurity. Cybersecurity protects systems against unauthorized access and attacks.
Define encryption. Encryption scrambles data so only authorized users can read it.
Name a field of AI. Artificial intelligence makes machines perform tasks that mimic human thinking.
How does machine learning work. Machine learning improves performance by finding patterns in data.
What is Big O used for. Big O notation describes algorithm efficiency.
How does parallel computing help. Parallel computing splits tasks across multiple processors.
How does cloud computing work. Cloud computing delivers computing services over the internet.
Summarize computer science. Computer science is the study of problem solving using logic data and computation.
Where are algorithms used daily. Algorithms are used in searching sorting and routing.
How are errors handled in code. Errors are handled with exceptions try blocks and catch blocks.
Why is debugging important. Debugging finds and fixes errors in code to make programs work.
What is source code. Source code is human readable instructions written by a programmer.
What is machine code. Machine code is binary instructions executed by the CPU.
What is an API. An API is an interface that lets programs communicate with each other.
What is open source software. Open source software is code shared freely for anyone to use and modify.
How does version control help. Version control tracks code changes and allows collaboration.
Give an example of version control. Git is a version control system used worldwide.
What is an algorithm’s input. An input is the data provided to start the algorithm.
What is an algorithm’s output. An output is the result produced by the algorithm.
Why use pseudocode. Pseudocode explains algorithms in plain language without syntax.
How does sorting help. Sorting arranges data for faster searching and analysis.
Name a sorting algorithm. Bubble sort insertion sort and quicksort are sorting algorithms.
What is a search algorithm. A search algorithm finds items in a dataset.
Compare linear and binary search. Linear search checks each item one by one binary search splits the data in halves.
What is hashing. Hashing maps data to fixed size values for quick lookup.
What is a cache. A cache stores frequently used data for fast access.
Why is modularity useful. Modularity breaks a program into smaller reusable parts.
What is concurrency. Concurrency allows multiple tasks to make progress at once.
What is virtualization. Virtualization lets one computer run multiple simulated machines.
What is an operating system kernel. The kernel is the core of an operating system that manages resources.
What is a packet. A packet is a small unit of data sent over a network.
How do firewalls protect computers. Firewalls block or allow traffic based on rules.
What is DNS. DNS translates human readable website names into IP addresses.
