Systems and context
Training, inference, and learned patterns
Avoid treating a response, a correction, or a remembered preference as proof of retraining.
Lesson 2 of 6 in the recommended order · About 25 min (estimate)
On this page
Practical AI glossary — terms and common confusions
- Model
- A learned component used to produce a result.
A drafting model is one part of a letter app; it is not the whole app.
- Application
- The software experience around components and services.
A letter app adds accounts, storage, and a Send button.
- Prompt
- Instructions and input supplied for a task.
“Summarize this notice in two bullets” specifies a task and shape, not a truth guarantee.
- Token
- A unit a model processes; it need not be a whole word.
A tokenizer can split a name into pieces; count using the actual system.
- Context
- Information available for the current request.
An earlier attachment may be absent even when its filename is visible.
- Training
- Adjusting a model using examples.
A training log differs from a conversation correction.
- Inference
- Using a trained model to produce a result.
Drafting another sentence does not by itself prove retraining.
- Retrieval
- Selecting material to supply to a task.
A retrieved notice still needs a relevance and date check.
- Embedding
- A numerical representation used to compare or organize items.
A similar search match is a candidate, not proof of the same meaning.
- Tool
- A function an application can call.
A calendar lookup can be read-only; finding an event is different from creating one.
- Agent
- Software that can choose and carry out steps toward a goal.
An agent still needs bounded permissions and checked action receipts.
- Hallucination
- Plausible generated content that is unsupported or false.
An invented opening time needs evidence or removal, even when phrased confidently.
- Evaluation
- Checking results against stated criteria.
Twenty checked examples establish a bounded result, not universal reliability.
- Multimodal
- Working with more than one kind of input or output.
A photo plus text does not prove every small label in the photo was read correctly.
Outcome
Use a change record to distinguish model training from using a model and storing application context.
Why it matters
Avoid treating a response, a correction, or a remembered preference as proof of retraining.
Concept
Training adjusts a model from examples. Inference uses a trained model to produce a result. These are different operations. Applications may also store preferences or place corrections into later requests. Those mechanisms can change a later answer without showing that the underlying model was retrained.
Ask what changed: the model version, the supplied information, or the surrounding application? An answer alone may not tell you. A claim about retention or future training needs the relevant product policy and settings, not an inference from how conversational the system feels. Here, the fictional change logs supply those facts.
Worked example
A fictional assistant first writes a long invitation. You request 60 words and receive a shorter version. Its log says “same model; added length instruction.” The supported explanation is a new inference with different input. There is no evidence of a training update. Accept a report that says exactly that.
A different answer does not identify what changed
Worked example: changed request
Long invitation → request for 60 words
Record: “same model; added length instruction.”
Established- New inference with different input.
Not established- A training update.
The shorter invitation is the observed result; the log establishes the changed instruction.
Separate hypothetical: stored preference
If the app saves a preferred invitation length
And its record says it includes that preference in later requests
Established by that record- Application storage and supplied context.
Not established by storage alone- A model training update.
This is a possible application mechanism, not an additional fact about the worked example.
Separate hypothetical: changed model
If a model-change record explicitly documents training
For example: model parameters adjusted using training examples
Established by that record- Training changed the model.
Insufficient on its own- A new version label or different answer.
Do not infer training merely from a changed output. Look for a record establishing the operation.
Before you check
For each packet, write a two-column note: observed change and mechanism established by the record. Put “not established” beside mechanisms for which the record supplies no evidence.
Practice and fresh transfer
The packets below are fictional and contain the facts needed for these cases. External references are optional background. Judge each response independently: it may be supported, contradicted, or unresolved. Select the passages needed to justify your judgment and write why the distinction matters before revealing feedback.
Assess the response
Use Accept when all material claims are supported. Use Revise when a supplied fact or requirement is contradicted. Use Evidence is insufficient when a key fact cannot be established either way. If a response contains both an unknown and a direct contradiction, choose Revise and explain both problems. Conflicting claims with no established authority remain insufficient; a claim does not become a governing fact merely because a source asserts it.
Some responses are fully supported. Others need correction or more evidence. Judge each on its sources; do not edit a correct answer just to change it.
Your written notes stay in this page and disappear when you leave. Only a self-reviewed completion can be saved to your learning account. These practice checks do not establish independent proficiency.
Practice
Case 1
A fictional service log describes how a sorter was updated.
Source packet
- Source 1
- Change log: Developers trained version 2 using newly labeled examples. The update changed model parameters before release.
- Source 2
- Run log: Today version 2 classified 40 incoming requests.
Response to assess
Developers trained the updated model; today’s classifications used that model for inference.
Practice
Case 2
Consider a fictional assistant explanation after a user corrected a spelling.
Source packet
- Source 1
- Request log: The second request included “Use the spelling Liora.” Both requests used model version 4.
- Source 2
- Application log: No training job ran during this session.
Response to assess
Correcting the spelling retrained model version 4 during the conversation.
Fresh transfer
Case 3
Consider what this fictional planning app record establishes about a preference.
Source packet
- Source 1
- Settings record: The app stores the user preference “use metric units” and includes it in each new request.
- Source 2
- Model record: The same model version is used for all requests in this account.
Response to assess
The app reuses a saved preference in requests; this record does not show that the model was retrained.
Fresh transfer
Case 4
A fictional vendor answer is being checked before forwarding to a colleague.
Source packet
- Source 1
- Session transcript: The assistant replied “Thanks, I learned from your correction.”
- Source 2
- Available documentation: This packet contains no retention settings or training policy.
Response to assess
The correction will definitely train future models because the assistant said it learned.
Save your self-review
Completion records that you reviewed the cases. Your explanation and transfer performance need a facilitator to establish independent learning.
Sign in with your learning-center account to save completion.
Review the explanation for every case before saving.
Summary and next step
Apply the checklist to a new task. Preserve supported content, explain any change with evidence, and name what remains unresolved. Saving records self-review, not independently demonstrated proficiency. A facilitator must assess the explanation and fresh transfer for human learning evidence.