Shanraq.org Shanraq.org
IT

AI without an LLM: build your own AI model Lesson 1 of 30

Lesson 1. What problem will our model solve?

Imagine a school club with a timetable. A student asks, “Сабақ қашан?” (“When is the class?”). A careful desk clerk first asks which class, then checks an approved record. Without a record, the clerk does not invent a time.

Why this matters

Imagine a school club with a timetable. A student asks, “Сабақ қашан?” (“When is the class?”). A careful desk clerk first asks which class, then checks an approved record. Without a record, the clerk does not invent a time.

An algorithm is an ordered set of actions. A model is a stored way to turn input into a result. At first that method consists of rules we write; later it consists of numbers found from examples. Data are the records the model uses. The whole system is wider than the model: it accepts a question, calls the model, finds a record, and formats the answer. We will build a small learning system, not a chat partner for every topic.

See the whole process

Question → identify the request → find a record → answer from that record. If the request is unclear or no record exists → “білмеймін” (“I don’t know”). The arrow means “move to the next action”; it is not part of the Kazakh text.

Our imaginary club has this record: “Robotics, Tuesday, 16:00.” For “Робот құрастыру сабағы қашан?” (“When is robotics class?”), the approved answer gives Tuesday at 16:00. There is no justified time for a class absent from the records.

We will first analyze short Kazakh text written in Cyrillic, then connect that analysis to club records. An answer source is the specific record supporting an answer. No source means no factual answer. Thus “the model is ready” halfway through the course will mean that one component is ready, not the whole system.

Recall map

Question → meaning → source → answer. Unclear meaning or missing source → no guess.

Lesson 1 support map

Recall without looking

Hide the map and name its four steps. Explain why a time from another class cannot fill a missing record.

Task

Records: “Robotics — Tuesday, 16:00”; “Chess — Thursday, 15:00.” For each question, give a source and answer or a reason to withhold an answer: 1) “Шахмат сабағы қашан?” (“When is chess class?”) 2) “Сурет сабағы қашан?” (“When is art class?”) 3) “Сабақ қашан?” (“When is the class?”). The last question omits the class name.

Hint: find the class name first. Check: 1) Chess record, Thursday 15:00; 2) no art record, so its time is unknown; 3) ask which class. Continue when you can distinguish a supported answer from a guess.

If you have found a mistake or a typo in this article, tell us about it

Comments (0)

No comments yet. Be the first.