• AI Agent – An autonomous AI model that performs tasks or accomplishes goals on behalf of the user. These can work in a solo capacity or with multiple other AI agents (called Agentic AI) to complete more complex tasks. 

​

• AI Model – The AI model is the result of data being processed by an algorithm. There are primarily three categories of AI models, including artificial narrow intelligence (ANI or weak AI), artificial general intelligence (AGI or strong AI), and artificial super intelligence (ASI or super AI). 

​

• Algorithm – This is the logic that is used by an AI model that influences how it functions. These are typically defined with mathematical language. The algorithm is applied to a set of data and will eventually achieve a specific purpose, as defined by the code. 

​

• Artificial General Intelligence (AGI) – An AI that can use learnings from its past and skills that it’s learned to complete new tasks without being trained by human interaction. This would be the rough equivalent of a human brain. 

​

• Artificial Narrow Intelligence (ANI) – An AI trained to do a single task or a narrow set of tasks. Two types of ANI are reactive AI (such as a chess playing machine) and limited memory AI (such as ChatGPT’s early iterations). 

​

• Artificial Super Intelligence (ASI) – An AI that can think on its own, learn, make judgements, and have mental abilities that far surpass a human being. This form of AI would be self-aware.

​

• Deepfake – An AI that uses deep learning to recreate the media of an individual’s appearance, voice, or both. Deepfakes take the form of fake images, videos, audio, or a combination of these. 

​

• Deep Learning – This is a subset of machine learning. It differs from generalized machine learning in that deep learning is used to train neural networks. Deep learning (using labeled or unlabeled data) is superior to machine learning when it comes to identifying complex patterns, such as object detection, because it can learn hierarchies within the data its given. 

​

• Generative AI – These are deep learning AI models that can create complex, original content. 

​

• Hallucination – This is the name of the phenomenon of large language models that generate answers that aren’t based on training data or don’t follow any identifiable pattern. Like how humans might think they see a face on the moon (or the brain believes “a trick of the eyes”), the AI can misinterpret the data its been given and pull an answer out of “nothing.” 

​

• Large Language Model – This is a form of AI that leverages deep learning (with its own neural networks) and immense amounts of data that can accomplish a multitude of tasks at once. 

​

• Machine Learning – This is a form of AI allowing machines to learn from data without having the exact programming that tells them to do so. 

​

• Neural Network – This is a form of machine learning algorithm that is designed according to the human brain. A collection of connected nodes, arranged into layers, work together to process and analyze data for patterns, relationships, or trends.

​

• Neuron – These are the individual nodes that compose a neural network. Each neuron interacts with the other neurons around it. For example, one neuron will receive an input from another, processes it, and transmits an output to other neurons in the neural network. 

​

• Training Model – These are the methods in which an AI is trained. The methods include supervised learning (humans code the AI to achieve a specific goal), unsupervised learning (humans give the AI data and let it define its own goals), and reinforcement learning (the AI learns by trial and error). Multiple methods can be used at once to train an AI.