The Ingredient Is Called Data

The lesson

The lorry at the back door

The lorry pulls up at my back door at half past five. The lamb is on ice, and the sacks of potatoes and crates of carrots come in beside it. Notice how I unpack my delivery. My pub keeps two pantries. The first holds the food: lamb, potatoes, carrots, onions, and bread. The second is an invisible pantry. It fills itself all day. And no bowl of AI has ever been cooked without it. The ingredient in that second pantry is called data.

Two pantries behind one pub

So what is data? It is not just numbers on a page. Data is anything recorded that a learner can learn from. The orders I get are data. The price of lamb is data. So is the length of my queue at 7:30 in the evening. So is the customers’ chatter, the photo a customer posts of his bowl, and the smell of my stew at 3 in the afternoon. I sort data two ways. Structured data sits in neat rows, like how I stack my plates: how many bowls of Irish stew I sell per hour, and the price list. Machines have read that happily for decades. Unstructured data is the customers’ chatter, the photos, the smells. Humans can read and make sense of these at a glance. However, one kitchen law covers both kinds of data. Good ingredients in, good bowl of stew out. Stock made from spoiled lamb, and the most skilled uncle in Singapore cannot help his dish, and he will end up serving a bad bowl of stew. Bad data works the same way. We call this: rubbish in, rubbish out.

The structured shelf and the messy one

The two pantries sit side by side everywhere. On the structured shelf: the number of bowls sold, and the cash collected today. Those rows are so tidy that the oldest computers could already total them. On the unstructured shelf: a tourist asking which size to order, and an aunty saying the stew was better last month. That shelf stayed locked to machines for decades, because the technology was not there yet. Reading and understanding this second pantry of unstructured data is the story of modern AI, because the world keeps most of its knowledge in chatter and in pictures.

Walk the day in your head

Now imagine the pub in your head, from opening to closing. Starting with the lights coming on, and ending with the last bowl of stew served. List everything my pub collects. I give you three seconds.

Pause here for three seconds.

You already know about the bowls sold and the cash collected. The three above are the ones you probably did not think about, and I track them all. Each one predicts tomorrow before tomorrow happens. And predicting is the job of a learning machine.