Paid Has No Maybe

I teach this to advanced students with a verb instead of a noun, because a verb carries time inside it and a noun hides it. Take the verb pay and say it four ways.

• may payThe future, wide open. The money might come, might come late, might never come.
• will payStill the future, but narrower. Someone has committed. Most of the open ground is gone.
• payingThe present. The card is in the reader. Very little is left to go wrong.
• paidThe past. Nothing is left to find out.

Read the list again and notice what changes. It is not the money. The amount is identical on every line. What changes is how much of the future is still open. Each rung down the ladder closes some of it, and the last rung closes all of it.

That last word is an artifact. An artifact is what an action leaves behind once the action is over: signed, shipped, delivered, paid. It has zero uncertainty. Not because we feel confident about it, and not because someone checked it twice, but because it is no longer in the future at all. Uncertainty is a property of the future. The past does not have any.

Even a refund does not unpay anything. Refunded is a second artifact laid beside the first. History is never edited. It only gets longer.

The oldest rule in philosophy

This is not new. Aristotle, in the Nicomachean Ethics, points out that no one deliberates about the past. No one chooses to have sacked Troy. Then he quotes the poet Agathon with approval:

For this alone is lacking even to God, to make undone things that have once been done.

Agathon, quoted by Aristotle, Nicomachean Ethics VI.2 (W. D. Ross translation)

In another book, On Interpretation, Aristotle wrestled with the other end of the ladder: is the sentence there will be a sea battle tomorrow true or false today? Logicians have argued about that sea battle for more than two thousand years. Nobody has ever argued about yesterday’s.

So hold the two ends in your mind. On the right, the Immutable Past, where every verb has an ‑ed and nothing can move. On the left, the future, where every verb needs a helper, may, might, could, will, to say how sure we are. Getting from the left to the right is nothing more than reducing uncertainty. The question every advanced student should be asking is: how?

Most likely, least likely

Here is the part that makes the ladder useful. The future is never one thing. It is a list, and the list is ranked. At the top sits the most likely future. At the bottom sits the least likely. When the client says we will pay on Friday, that is the top line. Below it: Monday, next week, half now and half later, a dispute, and at the very bottom, never.

Uncertainty is the shape of that list. When the top line and the bottom line look about equally plausible, the list is flat and you know almost nothing. When the top line towers over everything and the bottom line has nearly vanished, you are close. And in the past tense the list has exactly one line. Most likely and least likely become the same entry, because there is only what happened.

So the move from may pay to paid is not a leap across a gap. It is a ranked list getting steeper, rung by rung, until one line is left.

What makes the list steeper: patterns

How does anyone reduce uncertainty? Not by hoping. Not by insisting. Not by asking the client to promise harder. By patterns.

A pattern is history that repeats. Now look closely at what history is made of. Paid. Paid. Paid. A client who has paid every invoice on the first of the month for three years has handed you a long row of artifacts. When you read that row, their will pay stops feeling like a promise and starts feeling like a receipt that has not printed yet. The top of the list grows. The bottom thins out. You did not change the future. You read the past.

That is the whole lesson in one sentence: the only material you can build a prediction from is the past tense. Every bit of uncertainty you remove from the future, you remove with artifacts. Certainty about tomorrow is always borrowed from yesterday.

Where the ladder stops

Advanced students need the honest version, so here it is. Patterns can bring will pay very close to paid. They cannot make it paid. A pattern steepens the list. It never deletes the bottom line. The client with three perfect years can still be the one who does not pay this Friday. Watch the cloud in the figure above: push the slider all the way and it tightens dramatically, but it never collapses to a single point until the very end.

The last rung is not climbed by reading. It is climbed by acting. Only the act makes the artifact. Patterns shrink the distance between will and ‑ed; the act closes it.

The pattern machine

Claude Shannon put this on a mathematical footing. In 1948 he defined information as the reduction of uncertainty. A few years later he sat people down with English text and asked them to guess the next letter. They were remarkably good at it, because English is full of patterns: after a q comes a u, after th very often an e. A pattern is exactly the thing that makes the next item cheap to guess.

Now look at what a language model is. It was trained on text that had already been written: every word of it an artifact, past tense. When you ask it something, it produces, at every single step, a ranked list of possible next words from most likely to least likely, and picks from the top. It is Shannon’s guessing game played against a civilization’s archive. An AI is a machine that turns may into will using nothing but ‑ed.

That is the input AI changes. For all of history, reading patterns was slow, expensive and human: the shopkeeper who knew which customers were good for it, the foreman who knew which supplier always shipped late. AI makes pattern-reading nearly free and nearly instant, across far more history than any person could hold. The ranked list behind every will gets steep faster, for more things, for more people.

The consequence most people miss is this. When predictions become free, the prediction is no longer the scarce thing. The artifact is. A machine can read your history in a second, but it cannot write a single line of it for you. The world will price your will by your paid. Your history becomes the price of your future.

What I expect to happen

1. Promises will travel with their pattern.

   A will pay, will ship or will deliver will arrive with the past-tense record behind it, readable by the other side’s AI. A promise with no history attached will read the way an unsigned check reads today.

2. Reports will rank futures instead of promising one.

   Delegates will stop saying it will happen and start showing most likely and least likely side by side. A green or red status light will look as crude as a forecast that only says rain or no rain.

3. Long patterns will be treated like money.

   Someone whose history is all paid will get the terms of paid on the word will: deliver first, settle later, no deposit. The distance between the two words will carry a price, and a long pattern will make that price small.

4. Human attention will move to the bottom of the list.

   Machines will handle the most likely line. People will be called in for the least likely one: the dispute, the exception, the never. The human job is the tail of the list.

5. Forged artifacts will become the main attack.

   If patterns are made of past-tense records, the way to cheat a prediction is to fake the past. Proving that paid really means paid will become one of the most valuable services there is, which is the argument of The Cost of Being Checked.

Living on the ladder

I have lived on this ladder most of my working life. E2open, the company Paul Stern and I started, sat inside supply chains, and a supply chain is nothing but promises waiting to become past tense: will ship, shipping, shipped, delivered. Today at Charleston AI we sell deliveries, finished work, a past-tense word, because a client can feel the difference between we will help you and delivered. It is the same reason the notification everyone wants is done.

And it is why the Axiomatic approach I teach starts where it does. You write the outcome as though it has already happened, a press release for a merger that has not closed, and then you ask why isn’t it so? You are writing in the past tense on purpose, then hunting for the cheapest artifact that would steepen the list the most. The axiom is the ‑ed. The work is the pattern.

The claim ledger

Back to Friday, 4:52 p.m. The bank app, the third refresh, the buzz. Notice what you were actually waiting for. Not the money; the money was the same on every rung. You were waiting for a verb to change tense.

Payment received. Paid.

In the past tense, nothing is likely. It just is.

Background

• Aristotle, Nicomachean Ethics, Book VI, ch. 2, trans. W. D. Ross, quoting Agathon on the past.
• Aristotle, On Interpretation (De Interpretatione), ch. 9, the sea battle and future contingents.
• Claude E. Shannon, “A Mathematical Theory of Communication,” Bell System Technical Journal, 1948.
• Claude E. Shannon, “Prediction and Entropy of Printed English,” Bell System Technical Journal, 1951.
• The Reality Equation
• You Do Not Choose the Future. You Choose History.
• Sixteen Futures, One Past
• The Paradox of Intelligence
• Being an Actualizer
• Why Isn’t It So?