Artificial Intelligence

The term Artificial Intelligence (AI) was first coined in 1956 at Dartmouth College. Artificial Intelligence is the use of computers to model aspects of human reasoning and and learning. 

There are three levels of AI capabilities:

• Artificial/Narrow/Weak - this is the level of AI that exists today. This type of AI is usually developed to perform a single or narrow task, often more quickly or "better" than a human mind can. Even Generative AI is still defined as "weak".
• General/Strong AI - this level of AI is theoretical and is defined as computers being able to take an existing workflow and apply it to a new problem without human intervention or re-training/re-programing. Many believe it is possible that we will achieve this level of AI at some point. 
• Super AI - this level of AI is also theoretical and is defined as artificial intelligence surpassing human cognitive abilities. This is often the type of AI we see represented in science-fiction books and movies. 

There are four types of AI functionality:

• Reactive - this type of AI has no memory and cannot recall past interactions. These types of AI typically have one very specific task. In the 1990's, IBM designed Deep Blue, a computer programmed to play chess, which was able to beat a human Grandmaster of chess in a game. Deep Blue was able to recognize chess pieces and possible moves and predict which moves would be most successful based on its opponent's moves. However if it had played the same opponent more than once it could not have applied what it learned in previous games to their current game.  
• Limited Memory - this type of AI can recall past events or outcomes and use both past and present data to determine the "best" output. Examples of this type of AI are generative AI tools like ChatGPT, self-driving cars, and digital assistants like Siri or Alexa.
• Theory of Mind is a psychological term that means a being has the capacity to understand other beings' have mental states and that those mental states may be different from your own and influence the way that you interact. This term when applied to AI is still theoretical but may be achievable in the future. An example of theory of mind AI would be an AI that could predict that while people would normally say yes when offered ice cream they could also understand that today the person it's interacting with is sad and when that person is sad they don't say yes to ice cream. 
• Self Aware is a theoretical function of AI in which AI can form mental states of its own and those mental states in addition to the person's mental state would both affect the interaction.  

Other related terms that come up in discussions about AI can are as follow:

Machine Learning

Machine learning is a subset of AI in which computer systems are programmed to recognize patterns and make predictions based on the patterns rather than being explicitly programmed with step-by-step instructions. Machine learning allows computers to receive feedback on their performance and adjust/improve based on the feedback data. 

Neural Networks

Neural networks are used in machine learning to mimic the way that the human brain works. These networks are composed of nodes/layers (like neurons) where each node/layer receives an input, weighs the information, and passes the information along to other nodes/layers as needed. We hear the term "deep learning" when talking about any neural network that is comprised of more than 3 nodes/layers. An example of how a neural network might work is determining if you want to go for a bicycle ride. You need to consider if the weather is appropriate for a ride, if your bike is in good functioning condition, if you have all your safety gear, if you know how to ride a bike, if you have access to a trail or path that you want to ride on. Each question that needs to be answered could be considered a node/layer in the neural network that process your input, "should I go for a bike ride" and then gives you an output of yes or no.

Large Language Models

Large Language Models (LLMs) use deep learning neural networks to process huge amounts of textual data. Through analyzing billions of pages of text, LLMs can recognize a pattern of words and predict the most likely next word/words to follow. These models are the structure behind generative AI tools like ChatGPT and others. These models are now also being used on other media beyond text

Generative AI

Generative AI uses LLM's to generate new content like ChatGPT, Midjourney, Eleven Labs and others. An example of this type of AI in action would be prompting ChatGPT to write a sonnet in the style of Shakespeare about how hard it is to get up and go to class or work on Monday mornings. The AI is generating that content based on what it learned through it's LLM as the most likely combination of words that would relate to your prompt.