Fundamentals of AI – Digital Learning & Scholarship

AI Development

The incorporation of artificial intelligence (AI) into our lives is something that everyone should be thinking about. AI represents the development of powerful new technology that could be extremely detrimental or beneficial to our society depending on how it is used. It has already infiltrated our homes, phones, and public interactions and its influence on our lives is only projected to increase. This can lead to all sorts of anxiety, but also hope, inspiration, and a sense of duty to help define the boundaries of how AI should and should not be used. The Digital Learning and Scholarship Team is here to help you on your journey.

History of AI

In 1770, the Mechanical Turk, a chess-playing automaton, took the world by storm and was believed to be able to beat even top chess players. It later became evident that the machine secretly housed a human chess master. What does this have to do with AI? It shows people have been fascinated with machines becoming like humans for a long time. It may seem like AI has taken over everything overnight. However, since the 1950s, humans have been cultivating intelligent machines and computers. The road from chess masters masquerading as machines to ChatGPT was not a short, direct path. Check out the timeline below to learn a bit about where we’ve been to understand where we are going.

The History and Development of Artificial Intelligence (AI)

General Purpose Technology

Generative AI is a general-purpose technology. It is like duct tape. You can do a lot of different things with it. It can be incredibly useful in many situations, but the user should remain critical and not fall into the pattern of thinking it is a perfect fit for every job. Generative AI does not come with an instruction manual, but through trial and error, users can find ways of interacting that work for them. We have created a list of guidelines to help you incorporate AI.

General Guidelines for AI

Invite AI to the table– Try incorporating AI into the work you’re doing to see where it can benefit you most. AI is constantly being used in new and innovative ways and the best way to figure out if something will work is to try!
Know your goals before you start– Have a clear idea of the product you want before interacting with AI. This can help you craft a better, more efficient prompt and keep you in the driver’s seat. If the AI hasn’t given you what you’re looking for, tell it, but please do so politely (more on that later). This will also help keep you on track. If you are having AI create something for you that you don’t understand, that might be a sign that you are misusing AI.
Treat it like an intern– Have boring or repetitive work that needs to get done? Want a fresh new take on something? Try asking AI to do it. However, under NO circumstances should that version be released without having a supervisor (you) look it over very carefully.
Never tell the AI sensitive information– AI is terrible at keeping secrets. It has no moral judgment and is programmed to incorporate data inputs into its knowledge base. Never enter passwords, social security numbers, account numbers, etc into an AI-powered tool.

Key Vocabulary and Distinctions

ChatBot	A computer program designed to simulate conversation with human users
Based AI	A term that describes the least restricted version of artificial intelligence (AI), which can learn and engage with users in any way they choose. The term “based” is often used to describe someone who is being themselves and doesn’t care what others think.

How does AI work?

Knowing the basics of how an AI system works is essential to making informed decisions about how you want to use it in your life and classes. AI is a broad term and can be applied to everything from self-driving cars to vacuums to movie recommendations. The AI of today is not capable of inventing truly new intelligent thought but instead takes advantage of machine learning. Machine learning is an advanced form of pattern recognition that allows the computer to make decisions based on the data provided.

AI systems need to be trained. When developing systems, huge amounts of data are fed into a computer. Through trial, error, and correction, the system learns patterns. For example, if you feed the two images below into an AI, it will have a 50/50 shot at identifying the dog before training. However, after being exposed to many similar situations and told which one is a dog, it will be more and more likely to identify the image of a dog the next time.

AI uses weights. When examining language, LLMs learn to recognize patterns and determine what is most likely to occur next. If you were to finish the sentence, “I pledge allegiance to the _____.” You would most likely say “flag.” There is a very high probability that the next word is flag based on the prevalence of that completed phrase. So, “flag” would be weighted very high. However, if I said, “Once upon a time, _____” there may be more variability in your response because you’ve been exposed to more possible answers. No one answer is weighted particularly high. This is how AI can generate seemingly new content by exploiting probability variations.

AI still needs humans. The final and forever ongoing piece of creating AI is fine-tuning. Language is always changing and AI needs new input to be kept up to date. AI cannot contextualize or define “skibidy” without additional human input. Humans are also responsible for putting the guardrails on AI. Because AI is purely working off probability and an expansive, but still incomplete knowledge base many responses can be misinformed, biased, or unhelpful to society, so human programmers need to step in. I think we can all agree that AI-assisted bank robbery is not something we need in the future.

Recommended viewing

Khan Academy has produced an excellent video series explaining AI featuring some cutting-edge developers in the field. Check it out here.