Try writing down the exact rule for whether a photo has a cat in it. Not the idea of the rule. The actual code, pixel by pixel.
You cannot do it. Nobody can, and that gap is the entire reason machine learning exists.
Normal programming: you write the rules, the computer runs them over your data, you get answers. Machine learning runs that backwards. You hand over the data AND the answers, and the computer works out the rules for you.
What "works out the rules" actually means is less magical than it sounds. A model starts life as a formula stuffed with numbers set at random (its parameters, and you can picture them as knobs). Show it a photo you have already labelled "cat". It guesses wrong, because the knobs are random. Measure how wrong, then nudge every knob a fraction in the direction that would have made the guess less wrong. Next photo. Again. Thousands of times, often millions.
When the guesses stop improving you freeze the knobs, and that frozen pile of numbers is the model. The rule now lives in those positions and nowhere else. There is still no line you can read that says what a cat is.
Which hands you a useful test: if you can write the exact rule yourself, write it. Machine learning is for rules that are real but too fuzzy to pin down.
The chatbot you used this morning is that same loop, just enormous. Billions of knobs, nudged on error, over and over.
Quick check before you scroll: What's the core difference between traditional programming and machine learning, in terms of what the human provides versus what the computer produces?
The answer
In traditional programming, the human writes the rules and the computer applies them to produce output. In machine learning, the human provides data plus the desired output, and the computer produces the rules (as model parameters) by fitting to that data.
New here? I post a bite-size AI / ML concept like this every day. Follow me for the daily drop, and it compounds fast. Why I do it: https://lnkd.in/gK8knHDH