AI for EverybodyLearn the intelligence era with better judgment.

Your progress: 0 complete

Glossary My Map AccountQuick answers Choose my route History of AI Sources About this platform

You are here

Current learning moveUnderstand

Follow material in an order that builds understanding.

Do nowUse this path’s learning plan

Your guide

AI from Zero

Your learning move

UnderstandBuild ideas in a deliberate order.

Why this page exists

Learn is for subjects where sequence matters. The foundation comes first, the core adds depth, and thematic paths apply the ideas to a direction.

What you will leave with

Eight active segments build an ordinary-language map of AI, practical judgment rules, and a learner-chosen next direction.

Before you begin

Every path states its level and time. AI from Zero assumes nothing.

Do this now

Choose the earliest route that feels new to you; do not start with the longest course by default.

Do now on this pageUse this path’s learning planThe plan states the outcome, time, prerequisite, immediate action, and prepared continuation.Another valid routeCompare every pathChoose by outcome, time, and subject rather than card position.

Account

Foundation path · about 30 minutes

AI from Zero

Build an ordinary-language map of AI, distinguish the systems people mix together, practice one privacy and verification rule, and choose where to go next.

Eight connected segments30 min

01Why AI now?
02What is AI?
03The map of AI
04Training and use
05Useful and wrong
06What an agent changes
07Privacy and stakes
08Choose the next route

Plan for this page

Build a coherent first map of AI without prior technical knowledge.

You will leave with: explain AI without treating it as one machine or one product distinguish a model, chatbot, workflow, and agent separate training from using a trained system recognize why fluent output still needs judgment apply one privacy rule and one verification rule choose a useful next learning direction
Time: 30 min
Before you begin: No account and no technical prerequisite.
Do this now: Open the next unfinished segment: Why AI now?.

Prepared next stepAn old field entered ordinary life at a new scale

Route map

Eight connected segments

01Why AI now?
02What is AI?
03The map of AI
04Training and use
05Useful and wrong
06What an agent changes
07Privacy and stakes
08Choose the next route

After this path, you should be able to

explain AI without treating it as one machine or one product
distinguish a model, chatbot, workflow, and agent
separate training from using a trained system
recognize why fluent output still needs judgment
apply one privacy rule and one verification rule
choose a useful next learning direction

Chapter

Build the map

Begin with the field, the systems, and the difference between a model and a product.

014 min

Why AI now?

An old field entered ordinary life at a new scale

Artificial intelligence did not begin with chatbots. What changed is the combination of capability, accessible interfaces, infrastructure, investment, and rapid public adoption.

Ideas now grouped under AI have developed through many periods: symbolic systems, statistical learning, neural networks, and systems that generate language, images, audio, and code.

The current moment feels sudden because several changes arrived together. Systems became more capable in familiar media, interfaces became easy to use, computing and data infrastructures grew, and organizations placed AI inside products and work. Attention is recent; the field is not.

Interactive model

A compressed timeline

Selected layer1950s–1980s

Reasoning programs, expert systems, recurring optimism and limits.

Read the complete text alternative

1950s–1980sReasoning programs, expert systems, recurring optimism and limits.
1990s–2010sStatistical learning expands across search, recommendation, speech, vision, and prediction.
Late 2010s–2020sLarge-scale models and generative interfaces make broad capabilities visible to ordinary users.
NowAI becomes a product, workflow, policy, labor, education, and public-information question.

No score and no streak. Each check asks you to make a distinction, then compare your reasoning.

Which explanation best fits why AI attention feels newly intense?

Treat current AI as a new phase of an older field, not as a machine that appeared from nowhere.Choose a response before completing this segment.

024 min

What is AI?

AI is a family of designed systems, not one intelligence

In ordinary use, AI refers to systems that perform tasks associated with perception, prediction, language, learning, planning, or decision support.

The boundary is not perfectly fixed. A technique may be called AI when it is new and later become ordinary software. Different fields and laws also use different definitions.

A useful beginner rule is to ask what the system does, how it changes with data or rules, and where judgment remains. Do not decide from a label alone.

Interactive model

Examples and nonexamples depend on the mechanism

Selected layerSpam filter

May learn patterns from labeled email and predict whether a new message is spam.

Read the complete text alternative

Spam filterMay learn patterns from labeled email and predict whether a new message is spam.
Fixed tax calculatorMay follow explicit deterministic rules without learning; not every useful automation is AI.
Image descriptionMay use trained vision and language models to identify patterns and generate a description.
Human institutionPeople, policies, incentives, and workflows remain part of the system around the model.

No score and no streak. Each check asks you to make a distinction, then compare your reasoning.

A service learns from past labeled messages to predict whether a new email is spam. Which description is most useful?

Ask what capability is learned or designed and what wider system surrounds it.Choose a response before completing this segment.

034 min

The map of AI

Different capabilities overlap without becoming the same thing

Machine learning, language, vision, speech, robotics, generative systems, and agents are related areas and system capabilities—not interchangeable names.

A single product may combine several areas. A phone assistant can use speech recognition, language processing, search, recommendation, and ordinary deterministic software.

The map is relational rather than a perfect hierarchy. Some systems learn; some generate; some act through tools; some combine all three.

Interactive model

A practical field map

Selected layerPerceive

Vision, speech, sensing, and pattern recognition.

Read the complete text alternative

PerceiveVision, speech, sensing, and pattern recognition.
PredictClassification, ranking, forecasting, and recommendation.
GenerateLanguage, image, audio, video, and code generation.
ActRobotics, software tools, workflows, and agentic systems.

No score and no streak. Each check asks you to make a distinction, then compare your reasoning.

A tool turns spoken audio into written text. Where does it fit most directly?

Classify the capability first; then inspect how several capabilities are assembled into one product.Choose a response before completing this segment.

Chapter

How current systems work and fail

Connect data, training, inference, generation, and checking without treating fluency as proof.

044 min

Training and use

Building a model and running it are different phases

Training changes model parameters from examples and feedback. Inference runs a trained model on a new input to produce a prediction or generated output.

Training can require large datasets, objectives, computing, evaluation, and repeated adjustment. The resulting model encodes patterns rather than storing one clean database of verified facts.

When you send a prompt, the service usually performs inference. It may also add retrieval, tools, memory, rules, and human review around the model.

Interactive model

Do not collapse the lifecycle

Selected layerPrepare

Collect, select, clean, label, license, or generate data and evaluation tasks.

Read the complete text alternative

PrepareCollect, select, clean, label, license, or generate data and evaluation tasks.
TrainOptimize parameters toward an objective; test and adjust the system.
RunUse the trained model on a new input—this is inference.
OperateMonitor, update, constrain, retrieve, review, and recover in a real product.

No score and no streak. Each check asks you to make a distinction, then compare your reasoning.

You ask an existing model to summarize a new document. Which phase is happening most directly?

Ask separately how the model was built, what happens when it runs, and what product systems surround it.Choose a response before completing this segment.

055 min

Useful and wrong

Fluency is a property of expression, not a certificate of truth

Generated output can be helpful, plausible, and wrong at the same time. Verification effort should follow the claim, its freshness, and its consequences.

A model can combine supplied facts, learned patterns, inference, outdated information, and invented detail in one consistent voice. A real citation can still be irrelevant or misread.

Do not respond with blind trust or total rejection. Separate possibilities and drafts from factual claims, then raise checking effort with the stakes.

Interactive model

Four questions, not one feeling

Selected layerPlausible?

Does the output sound coherent and fit the situation?

Read the complete text alternative

Plausible?Does the output sound coherent and fit the situation?
Supported?Does relevant evidence support the exact claim?
Current?Could the fact, rule, schedule, price, role, or capability have changed?
Fit for the stakes?What happens if this detail is wrong?

No score and no streak. Each check asks you to make a distinction, then compare your reasoning.

An AI says a museum closes at 21:00 tonight. What is the useful next action?

Use generated output freely for possibilities; verify claims according to recency, support, and consequence.Choose a response before completing this segment.

Chapter

Action, delegation, and privacy

See what changes when a system acts, reaches tools, or receives sensitive information.

064 min

What an agent changes

The important change is delegated continuation and action

A model produces outputs. A chatbot organizes conversation. A workflow connects steps. An agent can continue toward a goal and use tools within permissions.

These categories overlap. A chatbot may call tools; an agent may use a chat interface; a fixed workflow may contain models. Marketing labels do not settle the architecture.

For judgment, ask what goal persists, what state is kept, which tools are available, what actions are permitted, where approval sits, and how failure is recovered.

Interactive model

System layers

Selected layerModel

Transforms an input into a prediction or generated output.

Read the complete text alternative

ModelTransforms an input into a prediction or generated output.
ChatbotOrganizes turns and context around a conversation.
WorkflowConnects known steps, rules, services, and review points.
AgentContinues across steps toward a goal and may choose or use tools within controls.

No score and no streak. Each check asks you to make a distinction, then compare your reasoning.

A system receives a goal, checks a calendar, drafts options, and may book only after you approve. What makes the agent-like part important?

Judge an agent by goals, tools, state, permissions, approvals, and recovery—not by its name.Choose a response before completing this segment.

073 min

Privacy and stakes

The first privacy question is what leaves your control

Before sharing data with an AI service, identify sensitivity, authorization, provider terms, retention, recipients, and the consequence of exposure or misuse.

A useful tool does not erase confidentiality, privacy, employment, client, health, legal, or security obligations. Removing a name may not be enough if the remaining details identify a person or organization.

A beginner rule: do not paste sensitive or confidential material unless you are authorized, understand the tool and terms, minimize the data, and have an appropriate protected workflow.

Interactive model

A simple sharing boundary

Selected layerPublic and low-risk

Still check accuracy and rights, but confidentiality may be limited.

Read the complete text alternative

Public and low-riskStill check accuracy and rights, but confidentiality may be limited.
Personal or internalPause: ask whether sharing is necessary, permitted, minimized, and protected.
Sensitive or regulatedUse approved systems and qualified policy or professional guidance; ordinary public tools may be inappropriate.
Action with consequencesAdd human review, permission limits, logging, and recovery according to the stakes.

No score and no streak. Each check asks you to make a distinction, then compare your reasoning.

You want an AI tool to summarize a private salary spreadsheet from work. What should happen first?

Permission, minimization, provider behavior, and consequence come before convenience.Choose a response before completing this segment.

Chapter

Choose your direction

Bring your map together and choose the next useful depth.

082 min

Choose the next route

A learning path should end with agency, not a funnel

The useful next step depends on your question. Go deeper into evidence, manipulate conditions, build a practical rule, or leave inAi for a substantial external course.

The platform does not infer a secret profile or optimize a feed. The routes below are explicit and explain what changes in depth and activity.

Choose one now. You can change it later in My Map.

Interactive model

Five honest next directions

Selected layerLearn

Continue through the ten-unit General AI Literacy core with active checks and transfer work.

Read the complete text alternative

LearnContinue through the ten-unit General AI Literacy core with active checks and transfer work.
UnderstandFollow the useful-and-wrong ecosystem through explanation and sources.
TryChange context, source support, and stakes in the deterministic Lab.
UseBuild a personal verification checklist for a real task.
Study elsewhereOpen a reviewed external beginner course for broader multi-hour depth.

No score and no streak. Each check asks you to make a distinction, then compare your reasoning.

What is the strongest reason to choose a next route?

Choose by purpose, activity, depth, and prerequisites—not by an unexplained recommendation.Choose a response before completing this segment.

My first map of AI

The map is useful when you can use it elsewhere

You do not need to memorize every label. Keep the distinctions that change a decision: what kind of system is involved, what data it receives, what claim it makes, what it may do, and what happens if it is wrong.

Keep one caution rule and one next direction. The result appears in My Map.

One rule I want to keepMy next direction

Loading page content...

…