Artificial Intelligence

Artificial Intelligence is a bit of a misnomer. What we have at our disposal today might seem very smart and believably intelligent, but they break down once you expect it to do something that falls outside of it's programming.

We may want to quickly make the distinction between what we refer to today as AI, something like ChatGPT or Copilot, and what they actually are: large language models, or LLMs for short. There are, of course, other "AI models" or forms of "generative AI" that take the form of image generators (like Midjourney and DALL-E) or image detection algorithms, or even in the form of audio and video deep fakes.

Some computer scientists theorize that the next big step in AI would be the completion of an Artificial General Intelligence, or AGI, for short. This definition falls more in line with pop culture ideas of AI, like Jarvis in the movie Iron Man, or HAL in 2001: A Space Odyssey) 

Most of the systems mentioned above share something in common however, they generally are built through a process called Machine Learning, or ML for short. ML is essentially natural selection, like what you may have learning in your Biology Evolution unit, but simulated and sped up to a much quicker time scale! ML is essentially a series of random events that produce a more suitable result over many generations. In a nutshell, a computer program is asked to make some random changes to its own algorithm to produce some sort of result, like a picture, or some text from a given input.

This program, often referred to as a Neural Network (or NN), receives a score based on how well it did after these random changes. If the score is higher, it knows that that change helped, so it may keep it. If the score is lower, it tries again with a new random change. These scores come from something called training data, human labelled and categorized data sets this neural network can refer to when it wants to check how well it's doing. 

How does AI learn to create? Can computers be creative?