What is AI
Short ideas, each with a picture you can play with. Open any panel to explore it. By the end, AI will feel less like magic and more like a tool you understand.
Normal software follows rules a person writes, step by step. AI is different. You show it thousands of examples, and it works out the rules on its own.
If the email says invoice, move it to Billing
Examples it learns from
- Invoice #4021 attachedBilling
- Lunch on Thursday?Not billing
- Payment reminder: due FridayBilling
- Your order has shippedNot billing
- Receipt for your March paymentBilling
- Team meeting moved to 3 pmNot billing
A new email
Your account balance is past dueBilling, 97% sure (illustrative)
AI cannot read words or see pictures the way you do. It turns every word, sentence, and image into a long list of numbers. Things with similar meanings get similar numbers, so they end up close together on a kind of map.
AI starts out guessing badly. After each guess, it measures how wrong it was and makes tiny adjustments to billions of internal dials. Repeat that millions of times and the guesses get very good.
Every AI system follows the same journey. Data is collected, cleaned up, and turned into numbers. The AI finds connections between those numbers and uses them to answer. Messy data in means messy answers out.
Preparing the data
-
CollectBooks, articles, photos, records
-
OrganizeClean, sort, remove junk
The AI at work
-
TranslateEvery piece becomes numbers
-
ConnectSimilar ideas pull together
-
AnswerBuilt from the patterns
Inside an AI is a huge web of connections. The ones that help it get the right answer grow stronger, the way a path through grass gets worn in when people keep walking the same way.
When a chatbot answers you, it is predicting the next word, over and over. Each time, it looks at everything written so far and picks a likely next word.
Orders arrive as texts, emails, and voicemails, each worded differently. A program built on rules breaks on every new way of writing “Friday”. AI has seen so many ways of writing things that it can pull the same details out of each message and fill in a tidy form in seconds.
AI can only pay attention to a limited amount of text at one time. Think of it as a desk with room for a set number of pages. When a new page arrives, the oldest one slides off, and the AI can no longer see it.
Behind the buzzwords, most AI does one of four things. You already use all four.
-
Recognize
Identifies what is in a photo, a sound, or a document.
Everyday exampleYour phone unlocking when it sees your face
-
Predict
Estimates what is likely to happen next.
Everyday exampleThe weather forecast
-
Create
Produces new text, images, or audio.
Everyday exampleYour email suggesting how to finish a sentence
-
Act
Takes steps on its own toward a goal.
Everyday exampleA smart thermostat adjusting the temperature by itself
Knowing the limits is what makes AI useful instead of risky.
AI can
- Find patterns in huge amounts of information
- Write, summarize, and translate in seconds
- Recognize images and speech
- Work at any hour without getting tired
AI cannot
- Know anything beyond what it learned from or was given
- Check its own facts the way a careful person does
- Understand meaning the way a person does
- Experience the world (it has never held a hammer or tasted coffee)
- Tell when it does not know something
That is the whole picture: examples go in, become numbers, form connections, and come back out as answers. Everything else is detail.