Neural Networks
The artificial "brain cells" inside a neural network that hold numbers are called these.
What are nodes
Machine learning models learn by looking at millions of examples contained in a massive collection of information called this.
What is a dataset or training data
This is the phrase that "AI" stands for.
What is Artificial Intelligence
A decision tree works by splitting data using a series of questions that only have these two possible answers.
What are Yes and No
If an AI image generator is trained on a human artist's drawings without their permission, it sparks legal debates over this right to own your creative work.
What is copyright
Information travels between nodes along connection pathways often referred to as these, similar to parts of a tree.
What are branches
If a model is trained only on pictures of orange cats, it will perform poorly on black cats due to this unfair flaw in the data.
What is bias
When an AI confidently makes up a completely fake fact or answer, it is called this word, which is usually associated with seeing things that aren't really there.
What is a hallucination
Just like a real plant, the very first question at the top of a decision tree where the splitting begins is known as this part.
What is the root
This term describes highly realistic, AI-generated fake videos of real people doing or saying things they never actually did.
What are deepfakes
In a neural network, data enters through the input layer and the final answer comes out of this layer.
What is the output layer
In this type of machine learning, humans act like teachers, giving the AI data that is already labeled with the correct answers (like pictures labeled "dog" or "cat").
What is supervised learning
This is what LLM stands for, the type of AI architecture that powers text bots like ChatGPT and Gemini.
What is Large Language Model
The final outcomes at the very bottom of the decision tree where the branching stops are called these.
What are leaves
AI ethics teams test chatbots to ensure they don't produce harmful instructions, a process known as placing these on the AI, just like the safety rails on a highway.
What are guardrails
The layers of nodes sandwiched between the input and output layers are known as these types of layers because you can't see them from the outside.
What are hidden layers
In this type of machine learning, the AI learns by trial and error in an environment, earning a "reward" score for doing things right—much like training a pet.
What is reinforcement learning
This is the term for the specific number inside an AI's neural network that gets adjusted up or down during training to change how important a piece of data is.
What is a weight
If a decision tree asks "Does it have wheels?" and then "Does it have a siren?", it is trying to group the data into specific categories, a process known by this "C" word.
What is classification
You should never type passwords or your home address into an AI chatbot because the company might save it to use as this.
What is training data
When an AI checks its answer against the correct answer and sends the error backward through the network to adjust itself, it uses this mathematical process.
What is backpropagation
When a model memorizes its training data perfectly but completely fails when given new, unseen data, it is called this "O" word.
What is overfitting
Because Large Language Models aren't actually "thinking," they are essentially just making a highly complex mathematical guess about this—the statistical likelihood of what word comes next.
What is probability
To avoid making a mistake with just one tree, AI scientists sometimes combine hundreds of different decision trees together to vote on the best answer. This technique is called a Random one of these.
What is a Forest
When an AI makes a major choice (like denying a bank loan) but its neural network is so complex that the programmers cannot explain why it made that choice, it is known as this mysterious "B.B." problem.
What is a Black Box