While most of us have heard of Artifical Intelligence in science fiction for years, it can be hard to understand exactly what it is and how it works. to start, let's define a few different terms. The AI that's being discussed most often today is Generative AI, which gives users a novel product based on the prompt they input. This might be text, like from ChatGPT, or an image, like from Midjourney. The user inputs directions, and the system outputs a new text or image that has not existed anywhere else previously. You may have also heard the term LLM, or Large Langauge Model. A tool like ChatGPT runs off of a LLM, which is a neural network that has been trained on massive quantities of text, in this case almost the entire Internet. The reason why Generative AI has become such a big topic recently is the improvements in these LLMs that allow AI to provide more nuanced and humanlike responses. The video below goes into more depth on these terms:
One question that confuses many is how AI "thinks". While this is an extremely complex topic, especially for the advanced models that we see today, much of it can be based in the idea of pattern recognition. One framework for conceptualizing how AI functions is the Chinese Room thought experiment. This scenario asks you to imagine that a person who has never heard Chinese, and isn't even aware it is a language, is in a room with nothing but Chinese characters. A message is sent to him in Chinese, and, through trial and error, they eventually respond correctly. Over more and more trials, the person will get better and better at predicting which character comes next in the sequence. They may reach a point where they can respond with a string of dozens of characters placed correctly. The person outside the room reading the messages would likely assume they were written by a Chinese speaker, but the person in the room has no clue they are speaking a language at all, they're just following a pattern. The same can be said for generative AI models. They are not considering their answer in the way a human would, but instead analyzing the combination of words you've prompted them with and returning a statistically likely pattern back.
What is AI? Affordances and Constraints of AI Using AI for Planning
Using AI in the Classroom AI Resources for EducatorsApplication Activity