Introduction To AI And LLMs
Artificial intelligence is a family of methods that let computers learn patterns from data and use them to make predictions or generate content. Large language models (LLMs) are one branch of that family, trained on huge amounts of text to produce fluent language on request.
What Artificial Intelligence Means
Artificial intelligence, usually shortened to AI, is the field of building computer systems that perform tasks we normally associate with human thinking. Examples include recognizing a face in a photo, translating a sentence, recommending a film, or answering a question.
The word covers a wide range of techniques. Some are simple rule lists written by people. Others learn from examples. When people talk about AI today, they usually mean the second kind, called machine learning.
It helps to separate two ideas. Narrow AI does one task well, such as spelling correction or route planning. General AI would handle any intellectual task a person can. Every system in use today is narrow, even though some, like modern chat assistants, cover a surprisingly broad range of tasks.
A helpful way to picture the field is as a set of nested circles. AI is the largest circle. Machine learning sits inside it, deep learning sits inside machine learning, and large language models sit inside deep learning. Each inner circle is a more specific technique that contributes to the one around it.
Machine Learning: Programs That Learn From Examples
In traditional programming, a person writes the rules. In machine learning, a person provides examples and the computer works out the rules itself. Imagine teaching a program to recognize spam email. Instead of writing every rule by hand, you show it thousands of emails labeled spam or not spam. The program finds the patterns that separate them.
The result of this process is called a model. A model is a large collection of numbers, called parameters, that together encode what the program has learned. Training is the process of adjusting those numbers until the model's answers match the examples well. Inference is the later step of using the finished model on new input.
Three common learning styles are worth knowing:
- Supervised learning: the examples come with correct answers (labels), and the model learns to reproduce them.
- Unsupervised learning: the examples have no labels, and the model finds structure such as groups of similar items.
- Reinforcement learning: the model tries actions and receives rewards or penalties, and it learns which actions earn the most reward over time.
Neural Networks And Deep Learning
A neural network is a model built from many small units called neurons, arranged in layers. Each neuron takes numbers in, multiplies them by adjustable weights, adds them up, and passes a result on. The idea was loosely inspired by the brain, but a neural network is a mathematical structure, and it works very differently from biological tissue.
Deep learning means using neural networks with many layers. Early layers pick out simple features, such as edges in an image. Later layers combine them into richer ideas, such as an eye, a face, and then a particular person.
Networks learn through an algorithm called backpropagation, combined with gradient descent. In plain terms, the network makes a guess, measures how wrong it was, and nudges every weight a little in the direction that reduces the error. Repeating this millions of times gradually produces a useful model.
Deep learning became dominant in the 2010s for three reasons: much more data, much faster hardware (especially graphics processors, GPUs), and better training techniques.
Training a large network takes a great deal of computing power and electricity, and the finished model is then copied and used by millions of people. This is why a handful of organizations with large data centers build the biggest models, while many others adapt or run smaller open models on their own hardware.
What A Large Language Model Is
A large language model, or LLM, is a neural network trained on a very large collection of text, including books, articles, websites, and code. Its core skill is predicting the next piece of text given the text so far.
Models read text in small units called tokens. A token can be a whole short word, part of a longer word, or a punctuation mark. When you type a prompt, the model converts it to tokens, calculates a probability for every possible next token, picks one, adds it to the text, and repeats. A full answer is built one token at a time.
Training on this simple prediction task, at very large scale, produces broad abilities: summarizing, translating, explaining, drafting, and writing code. The word large refers to the number of parameters, which for modern models ranges from billions upward, and to the amount of training text.
Most chat assistants add a second stage after the first. Human feedback and further training teach the model to follow instructions, stay helpful, and decline harmful requests.
A model's behavior also depends on settings chosen at generation time. A setting called temperature controls how adventurous the word choices are. A low temperature gives steady, predictable answers, and a higher one gives more varied wording, which can help with creative writing.
The Transformer: The Design Behind Modern LLMs
Almost every current LLM uses an architecture called the transformer, introduced in the 2017 research paper "Attention Is All You Need" by researchers at Google. Its key idea is attention, a mechanism that lets the model look at all the words in a passage at once and decide which ones matter most for understanding each other word.
Consider the sentence: "The trophy did not fit in the suitcase because it was too big." To know what "it" refers to, the model must connect "it" with "trophy". Attention gives the network a direct way to make that connection, even across long distances in the text.
Transformers also train efficiently on modern hardware, because they process many words in parallel. That efficiency made it practical to scale models up to billions of parameters and beyond.
The amount of text a model can consider at once is called its context window. Everything you type, plus the conversation so far, must fit inside it.
What LLMs Do Well And Where They Need Care
LLMs handle language tasks with strong skill. They draft and edit text, summarize long documents, translate between languages, answer questions, explain concepts at different levels, brainstorm, and write and debug code. Many people use them as a patient tutor or a first-draft partner.
Because an LLM generates likely text rather than looking up verified facts, it can produce statements that sound confident and turn out to be wrong. This is called a hallucination. Good habits reduce the risk:
Models also differ in size and cost. Larger models usually handle harder tasks better and run more slowly and expensively. Smaller models respond quickly and can run on a laptop or phone, which suits simple tasks and private data that should stay on the device.
- Check important facts, numbers, names, and citations against a reliable source.
- Give the model the source material directly and ask it to answer from that material.
- Ask for step-by-step reasoning on problems with several stages, then review each step.
- Remember that a model's knowledge stops at its training cutoff date unless it is connected to search or other tools.
- Avoid pasting private or sensitive data into tools whose data handling you have not reviewed.
Working With LLMs: Prompting, Tools, And Judgment
A prompt is the instruction you give the model. Clear prompts state the goal, give relevant context, describe the format you want, and include an example when helpful. Asking for "a 100 word summary for a busy manager, in three bullet points" works better than asking for "a summary".
Models can also be connected to tools. A model that can run a web search, read a file, or call a calculator can reach information and abilities beyond its training. Systems that let a model plan several steps and use tools along the way are often called agents.
AI also raises questions that deserve open discussion: fairness when training data reflects historical bias, the ownership of material used for training, effects on jobs and education, energy use, and accountability when automated systems make decisions about people. Learning how the technology works is the first step to taking part in those conversations well.
A practical rule serves beginners well: treat an LLM as a capable assistant whose work you review. You remain responsible for the final result.
Learning to use LLMs is a skill that improves with practice. Keep a short list of prompts that worked well, note what you changed when a result disappointed you, and share good examples with classmates. Within a few weeks, most people develop a reliable feel for what to ask and how to check the answer.
Timeline
Questions
Is AI the same as an LLM?
No. AI is the broad field. An LLM is one kind of AI model, specialized in language. Other AI systems recognize images, recommend products, or control robots.
Does an LLM understand what it says?
It models patterns in language very well and often behaves as if it understands. Researchers still debate what understanding means for such systems, so the safest approach is to judge an answer by its accuracy and check it.
Why do LLMs sometimes make things up?
An LLM produces the most plausible continuation of the text, and plausible text is sometimes false. Giving the model source material and verifying key claims reduces this problem.
Does the model learn from my conversation?
Within a single conversation, the model uses what you wrote as context. Whether your messages are also used to train future models depends on the provider and your settings, so read the privacy terms of the tool you use.
Do I need to know programming to use LLMs?
No. You can use chat assistants with plain language. Programming skill helps when you want to connect a model to your own data and tools, which later modules of this course cover.
What is the best way to start?
Pick one everyday task, such as summarizing an article or drafting an email, and practice writing clear prompts. Compare the output with your own judgment and refine the prompt step by step.
Go Deeper
Each course runs twelve weeks with study notes and slides.