Traditional Software vs AI: What's the Difference?
Understand the big difference between AI and normal software: AI learns its rules from examples.
Most software is traditional software: a programmer writes exact rules, step by step, and the computer follows them. "If the total is over €50, shipping is free." It is fast, reliable and predictable, but it can only do what someone managed to describe in rules.
Some problems are almost impossible to write rules for. Try writing the rules for "is this a photo of a cat?" or "is this email spam?". There are too many cases, and every rule has exceptions.
Modern AI turns the job around. Instead of writing the rules, we show the computer lots of examples with the right answers, and it works out the rules by itself. The result is called a model. The catch: it is only as good as its examples, and it can still be wrong.
Key takeaways
- Traditional software follows rules that a person wrote. It does exactly what it was told.
- AI learns its own rules from examples with the right answers.
- Learning wins on fuzzy problems (spam, faces, speech, language) where rules are impossible to write.
- A learned model is only as good as its examples, and it can make mistakes.
Check your understanding
Go deeper (more technical)
The filter in the simulation is a tiny version of a real classic, the naive Bayes spam filter, which became popular around 2002. Real filters use probabilities instead of +1/−1 counts, learn from millions of emails, and look at much more than words (sender, links, formatting).
"AI" is a broad term: it covers any system that does something we would call intelligent, including systems built entirely from hand-written rules (once called "expert systems"). Today, when people say AI, they almost always mean systems that learn from data. The next lesson sorts out these names.