How AI Thinks
AI makes this based on patterns it has learned
A prediction
The data used to teach an AI model
Training data
This streaming app uses AI to recommend your next show
Netflix (or any streaming/video service or app)
When AI produces unfair results because of flawed training data, this is called...
Bias
The process of teaching an AI by showing it examples
Training
In AI, these are the input columns used to make a prediction
Features
AI uses your past likes and watch time to do this on social media
Recommend content / fill your feed
If AI Bot only learned from data about adults, it would struggle to make good predictions for this group
Kids / children
In Sprite Lab, these are the characters or objects you program
Sprites
This is what AI looks for in data to make decisions
Patterns
What AI Bot did with the snack data in Lesson 3
Found patterns to pick a snack
This type of AI can generate text and answer questions, like a chatbot
Generative AI / Large Language Model
"Garbage in, garbage out" means this about training data
Bad data leads to bad predictions
This programming concept lets a sprite do different things depending on conditions
Conditionals / if-else
A prediction AI makes about a new item based on similar past items
Classification
If you show AI Bot only 3 examples, this is likely to suffer
Accuracy
Spam filters use AI to do this to your emails
Sort / classify them as spam or not
An AI trained only on photos of light-colored cars will likely struggle to recognize these
Darker-colored cars
In your project, your sprite made a "decision" — what does it base that decision on?
Rules / conditions you programmed
This term describes an AI that learns from data rather than being programmed with exact rules
Machine Learning
The reason AI holds back some data during training to test itself
To check how well it learned (testing)
Name two industries where AI is actively used to make important decisions
Any 2 of: healthcare, banking, hiring, criminal justice, transportation, etc.
Name one way to reduce bias in an AI model
Use more diverse / representative training data
What's the connection between your Sprite Lab project and how real AI works?
Both use conditions/patterns to make decisions based on input