What Is AI?

As seen before, we have the issue that when talking about "AI", we might not be talking about the same thing. Artificial Intelligence simply means any machine or code that mimics human intelligence to solve a problem. This ultra-broad definition includes concepts that may not be recognized by those thinking of ChatGPT when they hear "AI". For this reason, it's often better to be explicit in what we mean when we say "AI".

Lets compare modern AI with traditional computing (which is a form of AI):

• Traditional Programming: Human written code that explicitly states, step-by-step, what to do. This is a form of Artificial Intelligence, as you're creating something that will act/make decisions on its own, but it can't adapt to new situations without being explicitly programmed to do so. This is known as Deterministic AI or Rule-Based System. It relies on conditions (if ____, then ___) to determine what to do next. It's limited because it can't adjust to new situations and its pattern finding abilities are pretty limited.

• Modern Artificial Intelligence: Uses Neural Networks and Machine Learning in order to take huge amounts of data, find patterns, and generate conclusions. It predicts the next step based on what it has seen before. 

A common comparison is to talk about learning to recognize a dog.

• Traditional Programming method: List all the identifiable traits of a dog (4 legs, barks,  facial shape) and use that list to determine whether an animal is a dog. The problem is, because it doesn't adapt, it can't identify a dog that doesn't fit the mold (missing a leg, bark is different, etc.). Adaptations have to be manually made by the programmer.

  • This is the type of logic that makes it so we misidentify tomatoes as vegetables, orcas as whales, and koalas as bears.

• Artificial Intelligence Method: Show billions of images of dogs, the AI trains itself to identify commonalities. This method uses pattern recognition.