CSAI Basic: AI Literacy, Data and Python Foundations
What this domain tests
The Basic CSAI domain tests AI literacy rather than deep model building: the relationship between AI, machine learning and deep learning, types of data and learning, common application areas such as computer vision and natural language processing, introductory Python, and responsible use of AI.
Core rules to keep straight
- AI is the broad goal of making machines perform intelligent tasks; ML is a subset that learns patterns from data; DL is ML using multi-layer neural networks.
- Supervised learning uses labelled data; unsupervised learning finds structure in unlabelled data; reinforcement learning learns from reward and penalty.
- Classification predicts discrete labels such as spam/not spam; regression predicts a continuous value such as price or temperature.
- Data is split into training, validation and test sets. Test data must not be used to tune the model.
- Python indexing starts at 0; a slice [start:stop] excludes the stop index; // gives integer division; ** is exponentiation.
Worked example
A school wants to predict next month's electricity bill from past consumption. The target is a continuous rupee amount, so this is a regression problem. If the school instead predicts a band such as high, medium or low, the target becomes discrete and the task is classification. For evaluation, suppose a model gets 45 of 60 test records correct: accuracy = 45/60 = 0.75, or 75%.
Traps that cost marks
- Treating AI, ML and DL as interchangeable; DL is a narrower subset, not a synonym.
- Reporting accuracy alone on an imbalanced dataset. If 90% of records are negative, always predicting negative gives 90% accuracy but is useless.
- Leaking test data into training, which inflates reported performance.
- Off-by-one slicing errors: s[0:3] returns three characters, not four.
- Mixing classification and regression when the target is continuous.