# What AI is

> **What you will learn**
> What people mean by AI today, and how it differs from search. You need to know
> nothing going in. If you finish this chapter thinking "ah, so that's what it
> is," that is enough.

## What it actually is

In the [previous chapter](/guide/ai-alphago-to-now) we saw that today's AI is a
different kind of thing from AlphaGo. So what exactly is it?

If you were expecting something grand, this may be a small letdown. The
definition is this:

> **A program that read an enormous amount of human writing and got very good at
> guessing what word comes next.**

That is all. It is not conscious and it does not think for itself. But when you
get extremely good at "guessing the next word," the result **looks like writing,
summarising, translating, and answering questions.**

## How it differs from search

The fastest way to understand is to **ask search and AI the same question.**

### The situation

You are working with a staff roster in Excel. You want to type an employee
number and have that person's department appear automatically. You do not know
how.

### Asking search

You type this into the search box:

```
excel vlookup how to
```

You get back something like:

```
1. VLOOKUP explained completely — blog
2. Excel VLOOKUP from basics to practice — YouTube
3. 7 ways to fix VLOOKUP errors — forum post
...
```

Now **your work begins**. Open the links, read the explanations, and adapt them
to your own sheet. The example table has its columns in different places than
yours.

### Asking AI

You write this in a chat box. **There is no need to compress it into keywords.**

```
I'm managing a staff roster in Excel.
Column A has the employee number, B the name, C the department.
On another sheet I want to type an employee number and have the
department appear automatically. How do I do that?
```

The answer comes back like this:

```
=VLOOKUP(A2, Roster!A:C, 3, FALSE)

Put that in. What each part means:
- A2         : the cell holding the employee number to look up
- Roster!A:C : the range to search (employee number through department)
- 3          : take the 3rd column of that range (department)
- FALSE      : only accept an exact match

If the number isn't found you'll get #N/A. To show a blank instead:
=IFERROR(VLOOKUP(A2, Roster!A:C, 3, FALSE), "")
```

### Do you see the difference?

| | Search | AI |
|---|---|---|
| What it gives you | **where the material is** | **an answer fitted to your situation** |
| Your job | read, choose, adapt | check and use |
| How you ask | short, like keywords | long, like talking to a person |
| Can you follow up? | no, you search again | yes — "but what if numbers repeat?" |

In one line:

> **Search finds. AI makes.**

## One analogy

Picture a library.

- **Search** is the **catalogue**. It tells you "that subject is on floor 3,
  section B, shelf 7." Walking over and reading is your job.
- **AI** is the **librarian**. You say "here's what I want to know," and having
  already read the relevant books, they answer **organised around your
  question.**

The librarian is not always right. Memories blur; they misspeak. That is why
there is a separate chapter on [things to watch out for](/guide/ai-cautions).
The point is that **a catalogue and a librarian are for different things.**

## So what changes

Search is strong at "**things where the answer already exists somewhere.**"
Someone has to have written it down for search to find it.

AI can also answer "**things that are only true of your situation.**" Nobody on
earth has written about your staff roster, but AI builds an answer around the
situation you described.

How that plays out at work is the next chapter.

## Common questions

### "So is search obsolete?"

No. They are for different things.

- **Today's exchange rate**, **this restaurant's hours**, **breaking news** →
  search. AI may not know current information.
- **A method that fits my situation**, **summarising something long**, **a first
  draft** → AI.

### "I'm not good with computers. Can I use it?"

If anything, the opposite. **Being usable without knowing computers is this
tool's defining feature.** There are no commands or menu paths to memorise. You
write what you want in plain language.

Until now, making a computer do something meant learning its way of doing things
— menus, functions, commands. That has been reversed.

### "Does it cost money?"

There are free options and paid ones. While you are learning, free is plenty.

---

## Check yourself

**1. State the biggest difference between search and AI in one sentence.**

<details>
<summary>Answer</summary>

Search **finds where the answer is**; AI **makes an answer that fits your
situation.** That is why it can answer about things nobody has ever written
about — your sheet, your document, your circumstances.
</details>

**2. Why do you lose out by writing keyword-style questions to AI?**

<details>
<summary>Answer</summary>

AI builds its answer **around the situation you gave it.** Give it no situation
and it can only give a generic answer — roughly what search would have given
you. "Column A has the employee number, B the name, C the department…" produces
a far more useful answer than "excel vlookup."
</details>

**3. Why is asking AI for today's exchange rate risky?**

<details>
<summary>Answer</summary>

Because AI **may not know current information.** Values that change by the
minute, and facts about today, belong to search. We cover this limit in detail
in [things to watch out for](/guide/ai-cautions).
</details>

---

Next we look at what this tool **actually does in a working day**, with examples
→ [What gets better](/guide/ai-what-it-does)
