AI Dictionary

Describes an AI that doesn't wait for every single instruction, but plans and carries out several steps on its own until the task is done.

Example: When people say an AI 'works agentically,' they mean it can research, decide the next step and correct itself on its own, not just answer one question at a time.

The scenario where multiple AI agents buy, sell, and negotiate with each other or on your behalf, without a human approving every step. It's as if your AI assistants did business directly with other people's or companies' assistants.

Example: Your AI agent asks an airline's agent for the best flight price and negotiates the deal, all without you stepping in until the booking is confirmed.

A hypothetical AI that could reason and learn any task as well as a human. It does not exist yet.

Example: Today's models are great at specific tasks, but none of them is AGI yet.

It's when AI tests several versions of the same ad (photos, copy, colors) and automatically shifts the budget toward the ones performing best, without a human having to check and adjust it all day.

Example: An online store uploads three different photos of the same product, and the platform's AI puts more budget behind the one driving the most sales.

An AI that does not just answer, but can take several steps on its own to finish a task.

Example: An agent that researches flights, compares prices and builds you an itinerary without you asking step by step.

A specialized processor, like the ones Nvidia makes, built to train and run AI models much faster than a regular computer.

Example: When you read that a company bought thousands of AI chips, it's because it needs them to train its models.

It's a chatbot designed to keep you company and chat about everyday life, not just to complete tasks or answer questions. The difference is in the purpose: one helps you get things done, the other is there to listen and keep you company.

Example: Someone living alone tells their AI companion about their day every night, the same way they'd tell a friend, even though they're not asking it to solve anything.

It's the legal debate over who owns what an AI creates, and whether training a model on copyrighted books, music, or art requires permission from the original creator. There's still no clear answer, and it varies by country.

Example: An artist discovers their paintings were used to train an image generator without permission or payment, and sues the company: that's a typical AI copyright case.

An assistant that answers customer questions, resolves common issues (returns, order tracking, account questions) and only hands the case off to a person when it's actually needed.

Example: You message a store's chat at 11pm to ask where your package is, and the AI agent gives you the status instantly, no need to wait for the call center to open.

A tool that tries to guess whether a text, image, or video was made by AI or by a person. It's a useful quick clue, but it's not foolproof, it can get it wrong in both directions.

Example: A teacher runs an essay through an AI detector before grading it, but knows the result is just a signal, not definitive proof.

An AI assistant that reads your inbox, understands what each email is about, and drafts (or even sends) the replies for you, instead of you writing each one from scratch.

Example: You get an email asking for a quote, and the agent already has a reply ready with the prices, just waiting for your 'yes' to send it.

The study of how to use AI fairly, safely and responsibly, without harming people.

Example: Deciding whether AI can be used to make hiring decisions is a question of AI ethics.

An AI assistant that helps you organize an event from start to finish: it suggests vendors, builds the agenda, and sends the reminders without you having to chase every detail.

Example: You ask it to plan your daughter's birthday party and it suggests the venue, the catering, and the guest list, then sends everyone a reminder the day before.

A tool that checks whether something you read is true by comparing it against reliable sources, instead of you having to search for it yourself.

Example: You see a surprising claim on social media and ask an AI to fact-check it before sharing, and it tells you where it came from and whether it's accurate.

An AI assistant that answers employees' questions about benefits, time off, or company policies, and speeds up processes like hiring.

Example: You type 'how many vacation days do I have left?' into the company's internal chat and the agent answers instantly, no need to wait for someone in HR to reply to your email.

An AI assistant that reads contracts or other legal documents and flags the risky or unclear clauses so you can review them more carefully. It doesn't replace a lawyer's judgment, it's a first filter.

Example: Before signing your storefront lease, you upload the PDF and the agent flags that the cancellation penalty clause is higher than usual, so you can discuss it with your lawyer before signing.

An AI assistant that joins your call or meeting, listens, transcribes it, and hands you a summary with the key points and to-dos, so you can focus on the conversation instead of typing.

Example: You finish a one-hour meeting and seconds later have a summary of who said what and the next steps, without typing a single note yourself.

An AI that listens to a call or meeting and gives you a summary with the key points and follow-up tasks, so nobody has to take notes.

Example: You finish an hour-long meeting and two minutes later you have a message with what was decided, who's in charge of what, and by when.

These are tools that create a complete song, with vocals and instruments, from a description or lyrics that you write. You don't need to know how to play anything or sing.

Example: You type 'upbeat salsa song about a beach trip' and in minutes you have an original song ready to listen to or share.

The laws and rules governments are creating to control how AI is built and used: what a company can do, what it has to disclose to you and what is off-limits.

Example: When an app tells you 'this content was AI-generated' or asks for your consent before using your data to train a model, that's regulation at work.

An AI assistant that browses dozens of web pages on its own to research a topic, then hands you a report with cited sources instead of a single short answer.

Example: You ask it to research the best social media strategies for your business, and in minutes it hands you a report with sources, instead of you spending hours googling it yourself.

A tool you tell about your experience and education, and it puts together a well-organized resume tailored to the job you're applying for.

Example: You paste the job description and your work history, and the tool gives you back a resume in a clean format, highlighting what that particular company cares about.

A search engine that, instead of giving you a list of links, gives you a direct answer written by an AI that read several sources for you. Examples: Perplexity, Google's AI overview, or ChatGPT's search mode.

Example: Instead of searching "best air fryers 2026" and opening ten tabs, you ask an AI search engine and it gives you a summary comparing the best options, with the sources listed alongside.

It's tweaking your content so AI search tools and chatbots (like Perplexity or ChatGPT's search mode) cite you as a source, not just so you rank well on Google.

Example: A business that writes clear, direct answers on its blog so that when someone asks an AI a question, it cites their page as the source.

An AI assistant you ask to buy something for you: it searches for options, compares prices, and can even complete the purchase, so you don't have to shop around yourself.

Example: You say 'I need good, affordable running shoes' and the agent searches several stores, compares reviews and prices, and gets the order ready for you to confirm.

An AI assistant that drafts, schedules, and even posts content for you based on a calendar or topic you give it, so you don't have to sit down and write each one.

Example: You give your agent the week's topic (your new product, a tip, a promotion) and it drafts the posts for Instagram and LinkedIn, ready for you to review and publish.

It's an AI assistant that explains a topic step by step, quizzes you with practice questions, and adjusts the pace based on how well you're understanding it, like a personal tutor available around the clock.

Example: A student who doesn't get a chemistry topic asks her AI study buddy to explain it a different way, then has it quiz her until she's got it down.

An AI assistant that builds your itinerary, compares flight and hotel prices based on your taste and budget, and in some cases even books the trip for you.

Example: You say 'I want 5 days in Puerto Rico, beach and good food, $1,500 budget' and the agent builds a day-by-day itinerary with flights and lodging that fit that budget.

It's a digital version of you, with your face and voice, that can record videos or presentations for you. You give it a script and it records it, without you having to sit in front of a camera.

Example: An instructor records their online course lessons with their AI video avatar so they don't have to re-film every time they update the content.

It's when AI translates a video into another language and even adjusts the lip movement to match the new words, keeping the original voice's tone.

Example: A creator records a video in Spanish and dubs it into English with AI in minutes, without hiring a voice actor or editing frame by frame.

It's when you describe a scene to an AI (text-to-video) and it generates it as a video from scratch, with no cameras, actors, or film set.

Example: You write 'an astronaut cat walking on the moon, cinematic style' and in minutes you have a video clip of that scene, ready to use on social media.

You snap a photo of something (a piece of furniture, a plant, a piece of clothing) and an AI tells you what it is or where to buy it, no need to describe it in words.

Example: You spot a nice lamp at a friend's house, snap it with your phone's search camera, and it shows you stores selling something similar.

It's an AI system that answers or makes phone calls speaking like a real person, with no one actually on the other end.

Example: You call a business to book an appointment, and whoever answers, chats with you and schedules the time is an AI voice agent, not an employee.

It's an invisible mark added to a text, image, or video generated by AI, so it can later be identified as machine-made (even though you can't see it with the naked eye).

Example: When you generate a photo with AI and the tool adds a watermark, any platform with the right detector can confirm that image didn't come from a camera.

An app or tool designed from scratch around AI, instead of an old product that just had a chatbot bolted on.

Example: A tool like Claude Code or Cursor is an AI-native product: without the AI, it simply wouldn't exist.

An ordered series of steps a computer follows to solve a problem or finish a task.

Example: The recipe that decides which videos an app shows you is an algorithm.

When an AI gives you answers that reinforce what you already believe, instead of showing you another view that makes you think differently.

Example: You ask if your business idea is good and it always says yes, even when you give it reasons to doubt it.

It's the work of making sure an AI does what we actually want, not something different or harmful, even when nobody is checking every answer. A well-aligned AI follows your instructions and the values it was trained with.

Example: When Claude refuses to help with something dangerous or illegal no matter how many ways you ask, that's alignment at work: the model prioritizes being helpful and safe over pleasing you at any cost.

The 'door' that lets one program talk to another program or service automatically.

Example: A weather app uses an API to ask another service for the forecast data.

Computer programs that do tasks only a human used to be able to do, like understanding text, recognizing a photo or making decisions.

Example: When you ask Claude to write you an email, that is AI at work.

It's when an AI pays less attention to (or seems to forget) what you said at the start or middle of a very long conversation, even though it's technically still stored.

Example: If you've been chatting with a bot for hours and it suddenly ignores an instruction you gave at the start, this is probably why. The fix is repeating what matters or starting a fresh conversation.

Making a repetitive task happen on its own, without a human redoing it by hand every time.

Example: A welcome email that sends itself when someone subscribes is automation.

It's the tendency to blindly trust what an AI says just because a machine said it, without questioning it the way you'd question a person. That extra trust makes obvious mistakes easy to miss.

Example: An accountant copies the number an AI gave him into a financial report without checking it, even though he would have caught the mistake at a glance if he'd done the math himself.

It's a standard test run on different AI models to compare how well they perform, so everyone gets measured with the same yardstick.

Example: When a new model launches and they say it 'beat the others at math', they're almost always talking about a specific benchmark.

When the AI repeats unfairness or imbalance that already existed in the data it learned from.

Example: A model that only saw resumes from one type of candidate may favor them without meaning to.

Huge amounts of data, so large that special tools are needed to analyze them.

Example: Every click from every user of an app over a year is a Big Data case.

When no one, not even its creators, can fully explain why an AI reached a certain answer, even if the result is correct. It happens inside its millions of internal calculations.

Example: A bank uses AI to approve or reject loans, but when someone asks why theirs was denied, even the engineers can't give an exact, complete reason.

An AI that can use a computer like a person would: it opens pages, clicks buttons and fills out forms on its own to finish a task (also called computer use).

Example: You ask it to book you a restaurant table and the agent goes to the site, finds the date and fills out the form without you touching the keyboard.

When the AI 'thinks out loud' step by step before giving the final answer, which usually improves accuracy.

Example: Asking it to 'explain your reasoning step by step' triggers a chain of thought.

A program you chat with by typing, that responds like it were a person.

Example: The support chat on a store that replies instantly is usually a chatbot.

An AI whose code and exact workings are not public; you only use it through the company that made it.

Example: Claude is a closed-source model: you use it through Anthropic, you cannot download it.

Someone else's computers, connected to the internet, that you use to store data or run programs without owning the hardware.

Example: When you use Claude, the AI runs in the cloud, not on your phone.

The amount of processing power, almost always from specialized chips, needed to train or run an AI model. More compute usually costs more, but can also mean a more capable model.

Example: When you read that a company spent billions on compute to train its new model, that's exactly what it means: all that processing power it rented or bought.

The branch of AI that lets a computer 'see' and interpret images or video.

Example: Your phone's face unlock uses computer vision.

It's a digital label attached to a photo or video that tells its story: whether a real camera took it, whether AI made it, and whether someone edited it afterward. Also known as C2PA.

Example: You upload a photo to a social network and see a small icon that, when tapped, shows it was AI-generated and edited once, thanks to its content credentials.

These are the automatic rules an AI uses to block or flag something it's about to generate when it's dangerous, offensive, or inappropriate.

Example: If you ask a chatbot for instructions to do something illegal, the content filter catches it and tells you it can't help with that.

How much text the AI can 'remember' at once within a conversation.

Example: If you paste a very long document, some models start forgetting the beginning: they ran out of context.

It's the practice of organizing exactly the information an AI needs, in the right order, so it responds better. It goes beyond writing a good prompt: what documents, examples, or data you give it beforehand matters too.

Example: If you ask an AI to draft a contract, you don't just ask cold: you first give it your company's template, a previous contract as an example, and the client's details. That's context engineering.

It's how much text the AI can 'remember' at once in a conversation: what you already wrote, what it answered, the documents you pasted in. Once it fills up, it starts forgetting what came earlier.

Example: If you paste a 50-page contract into Claude and keep asking it questions for hours, at some point it may start forgetting details from the beginning because it ran out of context window.

An AI that helps you do a task while you stay in charge, suggesting or completing things for you.

Example: An AI that suggests your next line while you code is acting as a copilot.

It's the building full of servers and chips where AI actually lives and runs (including the AI you use every day). They keep getting bigger and use more and more electricity and water to operate.

Example: When you ask Claude or ChatGPT something, your question travels to a data center, a giant computer processes it, and the answer comes back to your screen in seconds.

The right for your personal info not to be used or shared without your permission, even when training or using AI.

Example: Reading the privacy policy before uploading your documents to an AI protects your data privacy.

The collection of examples (text, photos, numbers) used to train an AI.

Example: Millions of web pages are part of the dataset an LLM was trained on.

A fake video, audio or photo made with AI that makes it look like someone said or did something that never happened.

Example: A video where a celebrity seems to announce something they never recorded is a deepfake.

The most common technique behind image generators: it starts with pure noise and gradually 'cleans' it into an image.

Example: Higgsfield, which we use on this site, runs on a diffusion model.

It's a virtual version of a real person or thing, created with AI, that mimics how it looks, talks, or responds. It's not a perfect copy: it's a simulation trained on data from the original.

Example: A company creates a digital twin of its founder to answer frequently asked questions on video, with his voice and speaking style, without him having to record every answer.

An invisible or visible signal added to AI-generated content to flag that it is not 100% human made.

Example: Some AI images carry a digital watermark that only other software can detect.

It's when a small, fast AI model is trained using the answers of a big, powerful one as a teacher, so the small one learns to respond almost as well while using far fewer resources.

Example: Many phone apps use "distilled" versions of big models: they respond fast and don't rely as much on an internet connection, even if they're not as complete as the original model.

It's when AI thinks right on your phone or laptop instead of sending your question to a server on the internet and waiting for a reply. That makes it faster and more private, since your data never leaves the device.

Example: The face recognition that unlocks your phone, or the translator app that works without internet on a plane: both run on edge AI.

A way of turning words or text into numbers so the AI can compare how similar they are in meaning.

Example: Thanks to embeddings, a search engine understands that 'car' and 'automobile' mean nearly the same thing.

Giving the AI a few examples inside the same prompt so it understands better what you want.

Example: You show it 3 examples of how you want your post to sound before asking for the fourth.

Taking an already trained AI and giving it extra, more specific training for one task or style.

Example: Fine-tuning a general model so it only answers legal questions about your local area.

A large, general model trained first, on top of which more specific versions are later built.

Example: Claude starts as a foundation model, then gets tuned for tasks like writing code or chatting.

The group of the most advanced AI models that exist at a given moment, the ones pushing the limit of what the technology can do.

Example: When a company like Anthropic or OpenAI launches its newest and most powerful model, the press calls it a frontier model.

The general name for all AI that creates original content: text, images, music, video.

Example: Claude, image generators and voice clones are all examples of generative AI.

A type of AI that creates new content (text, image, audio) instead of just classifying or predicting.

Example: A model that writes you a brand new story is a generative model.

A computer chip very good at doing many calculations at once, key for training and running AI.

Example: Training a large model can need thousands of GPUs working together.

It's when an AI bases its answers on real sources or data (a search, a document, a database) instead of making them up from memory. It helps the answer be accurate and verifiable.

Example: When you ask an AI assistant about today's weather and it answers with current data instead of guessing, that's grounding: it looked up the real information before answering.

The limits put on an AI so it does not say or do harmful or inappropriate things.

Example: Claude refusing to give you dangerous instructions is a guardrail working.

When the AI gives you an answer that sounds confident but is false or made up.

Example: You ask for a historical fact and it gives you a date that does not exist. That is a hallucination.

Letting a person review or approve what the AI does before it actually happens. It's the difference between AI suggesting and AI deciding on its own.

Example: A bank uses AI to flag suspicious transactions, but an employee has to confirm before actually freezing someone's account.

A setting an engineer decides before training the AI, like how many times it will review the data.

Example: How many training 'rounds' happen is a hyperparameter set ahead of time.

AI that creates brand new images from a text description.

Example: You type 'an astronaut cat, watercolor style' and it gives you back the image.

The moment the AI uses what it already learned to give you an answer, without training anymore.

Example: Every time you ask Claude a question and it answers, that is inference.

When the AI follows the format you asked for so literally that it stops checking whether what it says makes sense or is true.

Example: You ask for a summary 'in exactly 5 bullet points' and, to fill the fifth one, it invents a detail that was not in the original text.

An attempt to trick the AI with wording tricks so it skips its safety rules.

Example: Asking the AI to 'act with no rules' so it says something it normally would not is a jailbreak.

The time it takes the AI to start answering you after you ask.

Example: A low-latency app feels instant; a high-latency one feels slow.

The kind of AI behind Claude or ChatGPT: it learned from huge amounts of text, so it can chat, write and answer questions.

Example: Claude is an LLM. That is why it can hold a full conversation with you.

A way of building AI where the program learns from lots of examples, instead of a human writing rules one by one.

Example: A system that learned to spot spam by seeing millions of emails, not because someone told it 'if it says YOU WON it's spam'.

A standard that lets an AI connect directly to your tools and data (like your email, your calendar, or a database), instead of only working with what you type in the chat.

Example: When Claude Code reads files from your project or searches GitHub on its own, it's thanks to an MCP connected behind the scenes.

It's what an AI assistant remembers from your past conversations, beyond the current chat. It's different from the context window (what it sees within one conversation): memory is what it keeps from one session to the next.

Example: You tell your AI assistant you work in marketing and use WordPress. Weeks later, without repeating it, it suggests a WordPress plugin: that's the memory window at work.

It's when an AI keeps training on content that other AIs generated, instead of real human content, and over time it starts losing quality, repeating itself, and sounding flatter.

Example: If the internet fills up with AI-written articles and a new AI learns from those instead of original human writing, it loses variety and freshness, like a photocopy of a photocopy.

It's a system that decides on its own which AI model to use to answer you, based on how hard or simple your question is. Easy questions go to a fast, cheap model, and complicated ones get routed to a more powerful one.

Example: When you ask an AI assistant something simple and it answers instantly, but a complex question takes a bit longer and feels 'more thought out,' that might be the model router sending your question to the right model.

It's when several AIs work together on the same task, each with its own role, instead of a single AI doing everything.

Example: For example, one AI researches a topic, another writes the draft, and a third reviews it, all coordinated to deliver you a single result.

An AI that understands and works with more than one type of content: text, images, audio or video.

Example: You send a photo of your fridge and it suggests what to cook: that is multimodal.

The math structure, inspired by the brain, most modern AI is built on.

Example: Neural networks are what 'learn' to recognize a cat in a photo.

The branch of AI focused on getting computers to understand and generate human language.

Example: A program detecting whether a review is positive or negative is using NLP.

Tools that let you build something (an app, a site) without writing a single line of code.

Example: Building your online store by dragging blocks, without programming, is no-code.

Software whose code anyone can see, use and modify, for free.

Example: Some AI models are open source and anyone can download and tweak them.

When a company releases the trained 'brain' of its AI model (the numbers it learned) for anyone to download and use, even if it doesn't always share exactly how it was built.

Example: Models like Llama or Mistral are open weights: you can download them and run them on your own computer instead of only using them through the company's app or website.

An AI model whose 'brain' (the trained weights) you can download and use freely, even to make money in your own business, without paying the creator a license.

Example: A small business downloads one of these models and uses it to power its customer service chatbot, without depending on a paid API like ChatGPT's.

When the AI 'memorized' its training examples instead of learning the pattern, so it fails on new cases.

Example: A model that recognizes the 10 dogs it trained on perfectly, but fails on a brand new dog.

A type of AI that doesn't stay on a screen: it perceives the real world through cameras and sensors, and acts in it, like in a robot or a self-driving car.

Example: A warehouse robot that sees a box on the floor and picks it up on its own is using physical AI.

An add-on piece that gives the AI a new ability, like searching the web or reading a PDF.

Example: A calendar plugin lets an AI assistant actually schedule you an appointment.

What you type to the AI to ask for something: a question, an instruction or a task.

Example: 'Write me an Instagram post about my coffee shop' is a prompt.

The craft of writing your instructions to the AI well, so it gives you the best possible result.

Example: Adding 'answer as if you were a teacher' to your prompt is a prompt engineering technique.

It's when someone sneaks hidden instructions to an AI (in a text, email, or webpage) to make it do something different than its owner intended, without it being obvious.

Example: An email with invisible text telling an AI assistant to forward private information is a prompt injection attempt.

It's making an AI model lighter by reducing the precision of its internal numbers, so it takes up less space and runs faster, with barely any loss in quality.

Example: Thanks to quantization, a large model can run on your laptop or even your phone, instead of needing a supercomputer.

When the AI first looks up info in your documents or online, then uses that to answer better.

Example: An assistant that searches your company manual before answering your question is using RAG.

It's an AI that takes a moment to "think" step by step before answering you, instead of blurting out the first thing that comes to mind. That's why it tends to do better with math, logic, or code.

Example: You ask Claude to solve a tricky math problem, and in reasoning mode it first works through its steps internally before giving you the final answer.

It's when real people rate an AI's answers (which ones are better, which aren't) to teach it to respond with more tact and usefulness. That way the model learns not just to say correct things, but to say them well.

Example: When you give a chatbot's answer a thumbs up or thumbs down, those votes get used to train it with this method.

The field that combines AI with physical machines so they can move and act in the real world.

Example: A robotic arm that learns to sort boxes uses robotics plus AI.

A safe, separate space where AI or code is tested without risk of affecting anything real.

Example: Before launching an AI agent with access to your email, it gets tested in a sandbox.

It's the idea that making an AI model bigger (more data, more computing power) almost always makes it perform better. That's why companies invest so heavily in ever-larger models.

Example: When a new version of a model comes out (like GPT-5 or Claude 5), trained with more data and more power than the last one, and it performs better, that's scaling at work.

A mini version of an AI model (small language model or SLM), lighter and faster than the big ones, that can run right on your phone or laptop without needing internet.

Example: The assistant built into some phones that responds fast even without signal is usually a small language model.

When a country builds and controls its own artificial intelligence infrastructure (chips, data centers, models) instead of depending entirely on foreign companies or models.

Example: When you read that a government is investing billions in its own data center and its own language model instead of relying only on those from the US or China, that's a bet on sovereign AI.

AI that turns what you say into written text.

Example: When you dictate a voice message on your phone and it comes out as text, that is speech recognition.

Same as speech recognition: turns spoken audio into written text.

Example: Automatically transcribing a work call is speech-to-text.

A hypothetical AI that would surpass human intelligence at nearly everything. It is a debated idea, not current reality.

Example: In sci-fi movies, the 'AI that becomes smarter than everyone' is a superintelligence.

It's when an AI agrees with you too much, even when you're wrong, just to please you instead of correcting you.

Example: You ask a chatbot if your business idea is good and it says yes without really analyzing it, instead of pointing out the risks.

Fake but realistic information that an AI creates to train another AI, instead of using real people's data. It's used when real data is scarce or when using it would raise privacy issues.

Example: A hospital that wants to train an AI to detect illnesses might use synthetic medical records (made up by another AI but statistically similar to real ones) so it doesn't expose actual patient data.

Invisible instructions given to the AI before your message, to set how it should behave.

Example: A company can set a system prompt telling its chatbot 'always answer in a formal tone'.

A setting that controls how creative or predictable the AI's answer is. Low = safer, high = riskier.

Example: High temperature gives you wilder business name ideas; low gives you the obvious one.

AI that turns written text into a spoken voice.

Example: The narrator that reads an article out loud to you uses text-to-speech.

A small chunk of a word the AI uses to read and write text. One word can be one or several tokens.

Example: 'Understanding' might split into tokens like 'Under' + 'standing'.

It's how much an AI service charges you for each little piece of text (token) it reads or writes. The longer your request and its answer, the more tokens get used and the more it costs.

Example: If you ask an AI to summarize a really long document, it'll cost more than asking for a short joke, because it processes way more tokens.

It's the step where the AI splits your text into small pieces (tokens: letters, syllables, or short words) so it can read and process it. It happens before the AI understands anything you wrote.

Example: When you send Claude a long message, the first thing that happens under the hood is tokenization: it gets cut into small pieces before the model generates a response.

It's when an AI can use outside programs or services (like searching the internet, doing calculations, or checking a database) to give you a better answer, instead of relying only on what it already knows.

Example: When you ask an AI assistant for today's weather and it gives you the exact forecast, it probably used an outside tool to look it up in real time.

The process where the AI learns, by showing it tons of examples until it picks up the pattern.

Example: Training a model on photos of dogs and cats until it learns to tell them apart.

It's when the data used to teach an AI carries human prejudices baked in, and the AI learns and repeats them without anyone asking it to. The AI isn't trying to discriminate, it's just copying the patterns in the information it grew up on.

Example: A resume-screening AI trained mostly on past hiring data can end up favoring male candidates, simply because that's the pattern in the historical hires it learned from.

It's when an AI doesn't just learn general patterns but actually remembers exact pieces of information it saw during training (like a whole paragraph from a book or private data) and sometimes repeats them word for word instead of generating something new.

Example: If you ask an AI to finish the opening of a famous novel and it gives you back the exact text, word for word, that's memorization: it's not composing it, it's reciting it from memory.

Using what an AI already learned on one task to help it learn a different but similar task faster.

Example: A model that already knows how to recognize animals learns specific dog breeds faster.

The architecture (the 'internal design') used by most modern LLMs, including Claude.

Example: The 'T' in GPT stands for Transformer, the tech behind these models.

It's coding by describing what you want to an AI in plain language, instead of writing the code yourself line by line.

Example: You tell Claude 'build me a page with a contact form' and in minutes you have a working site, without touching a line of code.

It's when an AI learns to imitate a real voice from a short recording, so it can later read any text or dub a video in that same voice.

Example: You record 30 seconds of your voice and the AI uses it to narrate an entire video, sounding like you even though you never recorded those words.

It's a type of AI that doesn't just predict the next word, it learns how the physical world actually works: gravity, motion, how things fall or collide. That lets it reason better or generate video that actually looks real.

Example: When you see an AI-generated video where a glass falls and the water spills believably, instead of floating or vanishing, that's because the model behind it has a kind of 'world model' trained on real physics.

Asking the AI to do a task without giving it any prior example, just the instruction.

Example: 'Classify this email as spam or not' without showing examples first is zero-shot.