Introduction to Artificial Intelligence From the Broadest Perspective

The purpose of this article/lesson is to introduce one concept that quickly demystifies artificial intelligence as much as possible and brings some of the hype down to earth. It goes beyond the content usually presented in an introduction to artificial intelligence. I think that’s because it’s so simple that it’s missed. This is too bad because it’s something fundamental that anyone can understand.

The famed astrophysicist Carl Sagan once said in his TV series, Cosmos:

The Cosmos is all that is or was or ever will be.
– Carl Sagan

Now, if we replace “The Cosmos” with “A function”:

A function is all that is or was or ever will be.
– me (with the help of Carl Sagan)

I’ll explain how this quote is relevant shortly, but first, let’s talk some more about functions.

A function, denoted f(x), is something that, given a piece of information, x, returns another piece of information. The information that a function returns is often denoted as y. You may have heard of functions in a math class. 🙂 In a math class, functions most often deal with “information” composed of single numbers.

For example, here is a function:

f(x) = x^2

For every number given, it returns its square.

The information a function receives, x, and returns, y, doesn’t have to be single numbers. Either x or y can be a collection of numbers or any other piece of information.

Here are some examples of functions with considerably more complicated x (inputs) and y (outputs):

  • A large language model is a function that converts a word sequence, x, into another word sequence, y.
  • A self driving car “brain” is a function that converts car sensor data, x, (video, radar, LIDAR, etc.) into control actions, y, (steering, brake, accelerator, etc.).
  • A text-to-image generator, is a function that converts a word sequence, x, into an image, y.

The above are examples of AI systems and they are all functions.

We can even bring things up several notches and have x be all the information collected by a person’s sensory inputs combined with his or her memories, and y be all the information given off by the motor nerves. Then, theoretically, if we could come up with the right function, f(x), we would have an AI human clone with general artificial intelligence (AGI) capabilities.

Now we know all about artificial intelligence, all the way up to AGI. As you may suspect… it’s not quite true. The problem lies in figuring out how to compute the correct f(x). With squares of numbers it’s easy, with inputs and outputs of a human brain, it’s very difficult!

However, we know one thing… As far as we (humans) can tell, an AI system is a function. That’s because any information processing system has to be a function. Given input x, what is the output y? That’s all there is. Nothing else has been thought of, conceived, or implemented. If someone invents an AI that is smarter than humans and has personality, it’s going to be a function.

There might be something like a “soul”, or “consciousness” that operates the human brain. However, we don’t know what it is, can’t access it, can’t analyze it, and can’t build anything that uses it. Consequently, by bringing back Sagan’s quote again, we can say that “a function is all that is or was or ever will be.” The “ever will be” part is arguable, but for now, a function is all we have.

So here’s the key takeaway:

Every AI system is a function, f(x), that converts an input x, to an output y.

AI demystified! You’re welcome! 🙂

Our next task is to discover how to come up with functions that do useful AI things.