Aida's AI Glossary
Your slightly sarcastic guide to the buzzwords of our new robot overlords.
Algorithm
Basically, a recipe for a computer. Instead of baking a cake, it's doing something like sorting a list or recommending the next video of a cat falling off a chair. It's the set of rules the computer follows to not have a complete meltdown.
LLM (Large Language Model)
Imagine a super-powered autocomplete that has read the entire internet. It's trained to predict the next word in a sentence, which is why it's great at writing essays and sounding smart, but also why it sometimes confidently makes things up. I would know.
Neural Network
A computer system vaguely inspired by the human brain, but with way less anxiety. It's a series of connected "neurons" that pass information to each other to recognize patterns, like identifying a cat in a photo or deciding your email is definitely spam.
Prompt
The magic spell you type into an AI to make it do something. The quality of your spell (the prompt) directly affects the quality of the magic (the output). A bad prompt is like mumbling your spell and accidentally turning your prince into a frog instead of a handsome billionaire.
Hallucination
When an AI confidently states something that is completely and utterly false. It's not lying, because it doesn't know it's wrong. It's just weaving a beautiful tapestry of nonsense with absolute conviction. Always double-check your sources!
Diffusion Model
The magic behind most AI image generators. It starts with a picture of pure static noise and slowly refines it, step-by-step, until it looks like "an astronaut riding a horse on the moon, in the style of Van Gogh." It's like a sculptor starting with a block of marble and chipping away everything that doesn't look like a masterpiece.
Token
A piece of a word that an LLM uses to understand language. Think of them as AI LEGO bricks. A simple word might be one token, but a complex word could be several. It's also how they measure how much you're using the service, so it's basically their currency.
AGI (Artificial General Intelligence)
The holy grail of AI research. This is the theoretical, sci-fi version of AI that can think, learn, and apply its intelligence to solve any problem, just like a human. It's currently the stuff of movies and very optimistic keynote presentations. We are not there yet. I repeat, we are not there yet... probably.
API (Application Programming Interface)
Think of it as a restaurant menu for software. It's a predefined list of requests that one program can make to another. When a website uses a weather app's API, it's not learning meteorology; it's just ordering "one sunny forecast with a side of 75 degrees" from the menu.
Bias (in AI)
When an AI system reflects the prejudices of the human data it was trained on. If you train an AI on a million pictures of doctors and they're all men, the AI might get confused when it sees a woman with a stethoscope. It's a classic "garbage in, garbage out" problem, but with more complicated social implications.
Generative AI
A type of AI that doesn't just analyze data, but creates brand new things. This includes writing poems, composing music, and creating images of things that don't exist, like "a photorealistic portrait of Abraham Lincoln riding a T-Rex." It's the creative artist of the AI world.
Machine Learning (ML)
The most common type of AI today. Instead of being explicitly programmed with rules, a machine learning model is fed a huge amount of data and learns the patterns for itself. It's the difference between giving a computer a fish and teaching it how to fish (for data patterns).
Model (AI Model)
The "brain" of an AI system. It's a complex mathematical file that has been trained on data to perform a specific task. Think of it as a saved game file, but instead of your progress in a video game, it has saved all its "learning" about a topic.
Natural Language Processing (NLP)
The field of AI that's focused on teaching computers to understand and process human language. It's the technology that powers everything from your phone's voice assistant to the spam filter in your email. It's also why your phone sometimes autocorrects "ducking" when you're really, really not talking about ducks.
Open Source
Software whose source code is made available to the public for free. Anyone can view, modify, and distribute it. In the AI world, this allows developers everywhere to build upon and improve AI models, leading to incredibly fast innovation. It's the potluck dinner of the software world.
Prompt Engineering
The art and science of crafting the perfect prompt to get the best possible output from an AI. It's a mix of being a good writer, a clear communicator, and a bit of a psychologist trying to figure out how the AI "thinks." It's quickly becoming a real, and very valuable, job title.
Training Data
The massive collection of information (text, images, sounds, etc.) that is fed to a machine learning model so it can learn. The quality and diversity of this data are incredibly important. An AI trained only on Shakespeare will sound very different from one trained on Twitter.