Week 01 · lesson

Lab: Build an AI Field Map

Mission

Run four supplied systems through the same trace. Your job is not to decide which one is “most intelligent.” Your job is to identify the task, mechanism, test, and failure boundary of each one.

The AI Field Map Lab runs locally in this page. It uses deterministic supplied cases, so every student can inspect the same evidence.

Safety, privacy, and runtime boundary

Use only the fictional cases provided. Enter no names, personal information, private prompts, credentials, or real records. The Lab makes no network request, runs no trained model, and controls no device or route. Backend View is a realistic fictional format for the same local trace, not a captured server log.

Starting cases

System	Baseline	Changed condition
Image classifier	clear bottle scores `0.91`	partly covered bottle scores `0.54`
Numeric predictor	error is `2 minutes`	unrepresented construction raises error to `13 minutes`
Route planner	eight-move route is valid	stale map sends move 6 into a blocked cell
Expert system	two complete facts fire two rules	missing humidity blocks one rule

Predict which systems use explicit rules, learned parameters, or search before you run them.

Completed guided example: numeric prediction

Field	Guided answer
Task	regression
Baseline	6 km and 2 stops represented as two numeric features
Mechanism	learned parameters
Result	18 predicted minutes; 20 observed; error `2`
Changed condition	construction exists but is not represented
Changed result	18 predicted minutes; 31 observed; error `13`
Limit	the model cannot respond to a condition absent from its features

The prediction did not change because the represented input did not change. The real trip did. That difference is the evidence.

Procedure

Select Image classifier and Baseline. Predict its mechanism, then run the local trace.
Switch to Changed condition. Record the score change and the review cutoff named in the trace.
Repeat the baseline and changed run for the numeric predictor, route planner, and expert system.
For each system, record the task and whether its mechanism uses explicit rules, learned parameters, or search.
Choose one system and preserve its exact baseline and changed result in the Evidence Record.
Explain what all four systems share without claiming they use one universal mechanism.
Complete the limitation, mark the procedure steps, and export the evidence JSON. The text fallback is equally valid.

Checkpoints

After the baseline runs, you should have four different jobs and three kinds of mechanism. After the changed runs, you should be able to name what information became uncertain, missing, stale, or unrepresented.

Success criteria

all four systems are matched to the correct task;
rules, learned parameters, and search remain distinct;
one baseline/changed comparison uses exact output or error values;
the shared pattern includes Input, Representation, Mechanism, Output, Test, and Boundary; and
the final limitation says these supplied traces are not live model or deployment evidence.

Troubleshooting

If the terminal area does not load, open Read the student trace as text or use the complete comparison table. If every result sounds identical, you are describing the interface instead of the mechanism. Reset restores the image classifier, baseline condition, Student View, and an empty evidence record.

Submission artifact

Submit the exported AI Field Map Evidence Record or its complete text equivalent. The assessment and Week Wrap-Up will inspect one conclusion without erasing the original.

AI Systems Lab · contract v1

AI Field Map Lab

Compare classification, numeric prediction, route planning, and explicit rules by changing one supplied condition in each system.

Help and boundaries

A local deterministic simulator displays supplied results and an output-only trace. It runs no trained model and contacts no service.

No personal data, private prompts, or API keys. This Lab makes no hidden remote writes and does not persist your work after the page session.

No result identifies a person, controls a robot, changes a route, or makes a real recommendation.

Lab draft is active for this page session only.

Objectives

Identify the input, representation, mechanism, output, test, and boundary of four AI systems.
Distinguish explicit rules, learned parameters, and search without treating them as the same mechanism.
Use a changed condition to explain one measured failure or uncertainty.

Procedure

Predict four jobsBefore running anything, match each supplied system to classification, numeric prediction, route planning, or rule-based reasoning.
Run the baselinesRun all four baseline cases and record the mechanism and test used by each.
Change one conditionSelect a changed case while keeping the system job fixed, then compare its result with the baseline.
Build the field mapUse exact Lab labels to explain what the four systems share and where their mechanisms differ.

Workbench

Four jobs, three mechanisms

Predict the mechanism before running the trace. The interface stays the same so the differences cannot hide behind four unrelated demos.

Local experiment · deterministic simulation

AI Field Map Simulator

Run four different AI approaches through the same trace so their jobs, mechanisms, tests, and limits stay visible.

SystemCondition

A readable explanation of the same local run.

01
InputWhat information entered the system?
02
RepresentationHow was that information encoded for processing?
03
MechanismDid rules, learned parameters, or search transform the representation?
04
OutputWhat result did the mechanism produce?
05
TestWhat comparison reveals success or failure?
06
BoundaryWhat changed condition limits the result?

Read the student trace as text

robotnix-ai-lab :: local deterministic field map
Choose a system and condition, then run the trace.
No live model, account, network request, or physical control is involved.

Image classifier · baselinebottle · score 0.91

expected bottle · match. Boundary: clear centered view only.

Baseline field map. Run changed conditions in the console to expose each boundary.
System	Task	Representation	Mechanism	Baseline test
Image classifier	classification	64 × 64 RGB pixels	learned parameters	expected bottle · match
Numeric predictor	regression	two numeric features	learned parameters	20 observed · error 2
Route planner	planning	grid cells and legal moves	search	every move reaches an open cell
Rule-based expert system	recommendation	named facts	explicit rules	two declared rules fired

Lab boundary: these are supplied deterministic traces for comparing system structure. They are not live model runs or evidence that any system is ready for real use.

Text and static fallback

Every system, condition, representation, mechanism, output, test result, and boundary appears as selectable text and in an accessible comparison table.

Image classification changes from bottle 0.91 to uncertain 0.54; numeric prediction error changes from 2 to 13 minutes; route validity changes from valid to blocked; the expert rule changes from fan and vent to insufficient input.

Evidence record

Record only the evidence requested. Do not enter names, personal information, private prompts, or credentials. This draft stays in memory only unless you explicitly export it.

Four-system comparison *For all four systems, record the task and whether the mechanism uses explicit rules, learned parameters, or search.Changed-condition evidence *Record the exact baseline and changed result for one system.Field-map conclusion *What do the four systems share, and what important mechanism difference must remain visible?Limitation *What does this deterministic supplied-case comparison fail to prove about real AI systems?

Success criteria

All four jobs are matched to their actual supplied mechanism.
One baseline and changed case use exact outputs or measurements.
The conclusion identifies the shared input-to-test pattern without claiming one universal mechanism.
The limitation states that supplied deterministic cases are not a live model evaluation or deployment study.

Complete every procedure step and required evidence field before export.

Import evidence JSON

Vocabulary lab

Flip the idea, not just the card

Explain the term before you reveal the back. Then compare your explanation with the definition, example, and warning.

Read all terms without animation

AI Field Map: A comparison of AI jobs, mechanisms, tests, and boundaries using the same system trace. Example: The Lab compares classification, regression, planning, and rule reasoning. Do not confuse it with: It is not a ranking from least to most intelligent.
baseline: The declared starting condition used for a controlled comparison. Example: The clear bottle produces a bottle score of 0.91. Do not confuse it with: A baseline is not automatically the best or normal condition everywhere.
changed condition: One altered part of a comparison used to expose system behavior or failure. Example: The bottle becomes partly covered while the classification job stays fixed. Do not confuse it with: Changing several conditions at once hides the cause.
deterministic simulation: A supplied program in which the same settings produce the same recorded result. Example: Every changed predictor run reports a 13-minute error. Do not confuse it with: It is not a live trained model or a real-world trial.
evidence record: A scoped artifact preserving observations, measurements, conclusions, and limitations. Example: The JSON records the route baseline and stale-map failure. Do not confuse it with: Exporting a record does not automatically prove mastery or deployment readiness.