AI Basics
Machine learning
A branch of AI in which computers learn patterns from data so they can make predictions or decisions, instead of following a rule a person wrote for every case.
- Parents
- Educators and school leaders
What parents should know
Machine learning is how a computer finds patterns in examples and then uses those patterns on new cases. It is the reason a photo app can name a face and a chatbot can guess the next sentence. The pattern is not a reason. Someone still has to check the result.
On this page
What is machine learning?
Traditional software follows instructions a programmer wrote: if this, then that. Machine learning builds a model by showing it many examples and adjusting until its guesses improve. The model can then classify, rank, or generate something it was not handed a line-by-line script for.
Generative AI is one use of machine learning. So are spam filters, recommendation lists, and some school tools that flag writing. The shared limit is the same. The system repeats what the examples made likely. It does not know your child, and it does not owe you a true sentence unless a person checks.
Why machine learning matters
Students meet the phrase in class and in ads. A calm definition keeps the lesson from turning into magic. If they can say the model learned a pattern, they can ask whose examples, and what happens when the pattern is unfair or out of date.
Schools also buy machine-learning tools that score students. Those are a different risk from a homework hint. A tool that ranks children needs a human who can override it, and a contract that says what data it used.
How it shows up in practice
- A class sorts the spell-checker, the chatbot, and the gradebook alert as three different uses of patterns.
- A student asks why the image tool draws the same stereotype, and the teacher points at training patterns.
- A counselor refuses to let an automated score be the only reason a student is flagged.
- A parent treats a sure-sounding chat reply as a draft, not a fact.
How HeyOtto helps
HeyOtto uses machine learning inside Otto, the assistant, for ages 8–18. The parent account is the human side of that system: you can read the chat, set topic limits, and see a homework session that teaches the step instead of only emitting an answer. Schools can open the same kind of session with a teacher or advisor. The model is not the teacher of record.
- Conversations are not used to train models.
- A cited search result still has to be opened by a person.
- Schools choose which student tools run.
For schools
Request a demoFAQs
Is machine learning the same as generative AI?
Generative AI is one kind of machine learning, the kind that creates new text, images, or audio. Other machine-learning systems only sort or score things they have seen before. A chatbot that writes a reply is generative. A filter that blocks a word can be much simpler.
Does the computer understand the lesson?
It produces a likely response. Understanding, in the human sense, is the student's job and the teacher's job. That is why HeyOtto's homework help asks the student to try the step instead of ending the chat with a finished answer.
Can machine learning be fair?
It can be tested and improved, and it can still treat groups unevenly because of the data. Call that algorithmic bias and look at the actual output. Do not accept a brochure that says the model is neutral.
Should a model decide a student's placement?
Not alone. Keep a person who knows the student in the decision. Ask the school what data the tool used and whether a family can challenge the result. A homework assistant is not a placement exam.
Sources
Last reviewed September 26, 2026. This entry is reviewed twice a year.