Before we dive into AI let’s ask ourselves, “What are computer skills?” The answer may have been simple at an earlier point in the process before computers were hardware and software and it was a job doing calculations, mostly done by women. These workers completed long, tedious calculations, often as part of a team. At that point in time speaking of someone's computer skills would mean their ability to do math calculations. Now having good computer skills may mean having excellent typing skills, being able to format a spreadsheet, being able to design a beautiful website, being able to create a website that works well. Good computer skills may mean programming, or game playing, or troubleshooting, it may be a combination or threshold of many of these skills at once. Likewise, when we ask ourselves “what is artificial intelligence” we may find our simple answers too narrow or context dependent.
Artificial Intelligence has become an umbrella term for an increasing number of capabilities and technologies. Artificial intelligence may sort the posts you see on social media or select which are videos recommended to you on YouTube, guide you through traffic, try to determine what you want to buy. Your cellphone has an AI assistant to convert voice commands into actions on your phone, your home may have a camera which uses AI to identify familiar faces from unfamiliar ones, or you may play a video game where you play against an AI opponent. This is just some of the breadth of things that might be referred to as ‘Artificial Intelligence’, but this course will focus on a subset of Artificial Intelligence that has become increasingly prominent called ‘Generative Artificial Intelligence’.
A simple explanation of Generative Artificial Intelligence (Gen AI) is an AI tool that has processed training data and can use that training data and an instruction from a user, known as a prompt, to create something that does not exactly appear in that training data. What the GenAI can produce is going to depend on it and the training data. For example, there are models trained on large amounts of text, others on images, others on sounds, code, videos and more. Some of these tools may allow for multiple media to be produced as an output. However, it is worth always remembering that even if the Gen AI model has been trained on information that is correct or accurate its output is not something that can be trusted to be true, this point will be expanded on in more detail in our section What Can GenAI Do?
While there are many types of AI it may be helpful to be aware of some categories of AI. Two terms worth being a bit more familiar with are ‘Machine Learning’ and ‘Deep Learning’. Machine Learning represents a subset of AI and Deep Learning is a subset of Machine Learning, meaning that all Deep Learning AIs are Machine Learning AIs which are all examples of Artificial Intelligence, but not all AIs fit in the Machine Learning category and not all Machine Learning AIs are Deep Learning AIs.
Machine Learning refers to AIs which can be autonomously trained rather than being explicitly programmed. These AIs may be given labeled data or may be told if their output is correct or not. They may be given the rules of a game and be left to develop strategies that let the AI win. An AI might correctly produce the correct output, but understanding the process the AI uses might be difficult and may be only true for the data it is trained on.
Deep learning uses a neural network and more processing power to guide its own development, identifying errors and adjusting its strategy. This allows for greater automation of the training process and can allow the model to process a wider range of uncategorized data. Deep learning has enabled large AI models like ChatGPT, Claude, and Bard to work the way they do.