What is Artificial Intelligence: A Three Minute Introduction
With everyone talking about AI, I thought you might like to know exactly what it is.
Artificial Intelligence (AI) is like compound interest: the concept is simple, but the results are counterintuitive and astonishing. Here’s a three-minute introduction to the basic concept behind AI, so three minutes from now you’ll understand what AI is.
At its core, AI only does two related things: it finds patterns and it applies those patterns.
Here’s an example:
Suppose you have some pairs of numbers: 1 and 2, 2 and 4, 3 and 6, and 4 and 8. AI is a kind of computer program that can look at those pairs and figure out the pattern —in this case that the second number in each pair is double the first number. This is the pattern-finding phase.
Then you can give the AI program a new first number, like 5. It will apply the pattern (doubling) and tell you that the corresponding second number is 10. This is the pattern-application phase.
Or you could provide a new second number, like 7. Again, the AI program will apply the pattern. This time it will tell you that the first number is halfway between 3 and 4, that is, 3.5. This is also pattern application.
This second example (deducing 3.5 from 7) works even if there are no decimals or fractions in the original data. In this way, AI can discover halfway points that no one has thought of before.
Finding the pattern seems trivial in this example, because we already know about multiplication and division. And finding the halfway point also seems trivial, because we also already know about fractions and decimals. But for a long time, people didn’t know that there was anything between 3 and 4. (Think of the Romans who didn’t have a number between III and IV.) They would have been astonished.
Here’s the real magic: AI can find halfway points between things where even today we don’t think it’s possible: for example, something halfway between “tomorrow” and “sincerity.” You might at this point reasonably object that there is no halfway point between those two, but that’s only because we are like the Romans in this regard. AI knows it’s there even though we can’t imagine it.
AI can find a halfway point between “tomorrow” and “sincerity.”
In other words, there are patterns in the world and in data that we as humans simply cannot understand. But AI can.
Here’s a more complex example of AI: Instead of pairs of numbers, we can give AI pairs of an image and a description, such as a photo of New York City and “New York City,” and a painting by Renoir and “Renoir.” Then we can ask AI to create a Renoir painting of New York City!
Similarly, from essays and summaries of those essays, AI can create new essays from new summaries. From pairs of faces that are either the same or different, AI can tell you if two new faces are the same or not. From pairs of faces that are in focus and out of focus, AI can put out-of-focus faces back in focus. From law books and cases and legal conclusions, AI can generate new legal conclusions for new cases. And so it goes.
In principle these patterns are just like multiplication, but they are so complex that no human can understand them. Still, even if you can’t understand the patterns themselves (no one can), you can understand that there is a pattern.
That’s how AI works. Simple. And counterintuitive and astonishing.
Thanks for putting this in words ordinary people like myself can follow. I have a question - if we can't figure out patterns but AI can, how did we teach it to do this? Or did we?