What AI actually is
AI اصل میں کیا ہے
35 min read
Three ways to see it
Artificial intelligence is a phrase that does a lot of work in 2026, often more work than it deserves. When NADRA's system matches a thumbprint to a CNIC record, that is AI. When FBR's compliance engine flags a tax return as unusually shaped, that is AI. When Easypaisa blocks a transaction that looks like fraud, that is AI. None of these systems are conscious. None of them think. They each take inputs, pass those inputs through a function that was learned from millions of past examples, and produce an output with a confidence score. The function was not written by a programmer line by line. It was found by an optimisation process running over data. That is the whole of the trick. The rest is engineering, scale, and electricity.
Way one to think about AI: rule-based programs. A traditional FBR audit script says: if income is above PKR 6 million and declared tax is below 5%, flag the return. A human wrote that rule. The system follows the rule. If the rule is wrong, the system is wrong in a way you can read and fix. This is not AI in the modern sense, but it is the kind of automation that most government and corporate systems in Pakistan still run on, and it works.
Way two: machine learning. You do not write the rule. You give the system ten lakh past tax returns labelled honest or fraudulent and tell it to find the patterns itself. The system, through a long optimisation, produces a model. Hand it a fresh return and it gives you a probability score. The model often catches frauds the human rule missed, because the patterns are too subtle for a person to spell out in if-then language. The cost is opacity: the people who built the model cannot read it like a recipe.
Quick check
Quick check: what makes modern AI different from a rule-based program?
The why-tree
Why-tree level one: why pattern recognition at all? Because most of the decisions we want machines to make are not the result of clean rules. Fraud, disease, fluent translation, voice transcription, none of these have a small set of rules a human can write. They have signatures that show up across thousands of examples. Pattern recognition is the only path that scales.
Try this with Claude
AI-edge prompt to try with Claude or ChatGPT: 'List five jobs in Pakistan that already rely on AI behind the scenes, name the AI system involved, and explain in two lines what would happen if that AI failed for a week.' Compare the answer to your own intuition. Where did it surprise you?
Sources
Sources and further reading. Russell and Norvig, Artificial Intelligence: A Modern Approach (concept level). IBM, 'What is Artificial Intelligence' (ibm.com/topics/artificial-intelligence). Anthropic, 'A primer on large language models' (docs.anthropic.com). Microsoft Learn, AI For Beginners curriculum (github.com/microsoft/AI-For-Beginners). MIT OpenCourseWare 6.034 Artificial Intelligence (ocw.mit.edu).