AI is an encompassing term for artificial intelligence (AI), machine learning (ML), and deep learning. Artificial intelligence encompasses all three within it. Machine learning encompasses machine learning and deep learning.AI in its purest form is decision trees that match conditions to actions based on what is accessed. Some examples would be video game AI which adapts to in game conditions, and automated robotics that are programmed to traverse an area and perform actions when meeting a condition, such as finding a person or item and picking up that item or giving something to that person. A third example would be an IOT controller which when triggered by its sensor readings when they reacha threshold, it sends data to a client for processing and examination as part of a data pipeline.
When people in the 2020’s think of AI, they usually think of either a chat assistant like Siri, Cortana; or a Chat AI agent like ChatGPT, Gemini, CoPilot, Grok, or Claude. Mostly, these agents are trained using deep learning on hundreds of gigabyte of text data in the first stages of training to respond to human text or voice prompts. The voice is just translated to text and large language models (LLMs) interpret the meaning of the text and words in respect to one another and in respect to words with similar meaning. This data is used for the deep learning model to train on to train it to do a particular task or set of similar tasks that result in a model that can handle conversations with people through their browser, or with the case of Claude Code, their terminal. The chat models can also handle content generative tasks like generating books, software, music, and even video games. The original and thematic usage of these models was for information summarization where the agent would traverse the internet for information pertaining to the query and contextualize it as well as summarize it and give it to the user. For a programmer, it takes 2 seconds to write a promp, and he or she is rewarded with a snippet of code that combines several articles and sources as well as libraries to create a code snippet that functions as the prompt describes. The content used to train these models now exceeds 1 petabyte per model in 2026, and it involves music, video, video games, and software, not to mention open source software is now being used to build new open source software in the matter of a few seconds.