Artificial Intelligence — what a machine can decide, and what only you can answer for
A Koinder Global Limited production
Adam looked at every creature and gave it a name. Come and see what we taught machines to do.
In the very beginning, God said something remarkable: "Let us make man in our image." Out of everything that was made, one creature was made to resemble its Maker — able to think, to choose, to name, to build.
And then came a strange and beautiful task. Every animal was brought to Adam, "to see what he would call them."
Think about what Adam actually had to do. A creature walks up. He had never seen it before. He had to look at it, notice what kind of thing it was, and give it a name. Then the next one. And the next.
That is the oldest recorded act of what we now call classification — looking at a thing and deciding what it is.
In England there was a mathematician named Alan Turing. During the war he had helped break secret codes. Afterwards, one question would not leave him alone: "Can machines think?"
Everywhere he took that question, people argued in circles. What is thinking? What is a mind? Does a machine have a soul? Years could pass and nobody would agree on the words.
Here is the move that made Turing different. He did not try to win the argument. He changed the question.
Instead of "can it think?" — which nobody could measure — he asked something you could actually test: "If you were passing notes to someone in another room, and you could not tell whether it was a person or a machine, what exactly would be the difference?"
That was 1950. Six years later a scientist named John McCarthy needed a name for this new field for a meeting at Dartmouth College. He called it artificial intelligence.
Notice the pattern in Turing's mind, because it is worth more to you than any fact on this page: when a question cannot be answered, do not argue louder — make it smaller and testable.
TAKE IN. The creature is brought before him.
COMPARE. What is it like? What does it resemble?
DECIDE. He gives it a name.
ANSWER FOR IT. Whatever he called it — that became its name. He carried the responsibility.
TAKE IN. A photograph arrives.
COMPARE. It measures this photo against millions it has already seen.
DECIDE. It outputs a name: "cat". Or "tumour". Or "not a criminal".
ANSWER FOR IT. And here is the hard part — the machine cannot. A human being has to.
Steps 1, 2 and 3 are the same. Take in, compare, decide. A machine can do all three, faster than any of us, for a million pictures before breakfast.
Step 4 does not match, and it never will.
Adam could be asked "why did you call it that?" A machine cannot be asked. It cannot be sorry. It cannot be held responsible. So when an AI decides something about a real person's life — their loan, their medicine, their freedom — somebody human is answerable.
If you build it, that somebody is you.
TAKE IN — the machine receives something
COMPARE — against everything it has seen before
DECIDE — it produces an answer
ANSWER FOR IT — a human being. Always.
Anybody can build the first three. The engineers the world actually trusts are the ones who never forget the fourth.
You are going to build a tiny AI that decides: is this fruit ripe enough to sell? Choose your rules, then run it on real fruit.
Same shape. One change each time. Do not rush.
Find one place today where a machine decided something about a person — who gets the loan, which advert you were shown, whose message went to spam.
Then ask the only question that really matters: "Who is answerable if it got that wrong?"