Some AI Terminologies

Islam Religion is the religion of peace.

On May, 2026, I started AI Prompting for everyone course on Deeplearning.ai by Andrew Ng. The course was about how to become an AI Power User. I've learnt that a Power User is much better in AI usage than a Novice User. A power user can TALK to AI models in order to get the most desireable response with much less effort. Being an AI Power User as the course suggests invloves understanding:

How/Where does AI model gain his knowledge?

How much Data does an AI model rely on?

How to use AI model in your day-today tasks?

What to expect from an AI model and what to not?

What are the AI model capabilities?

I have finished this course after a month, I was lazy!!

On Aug,2026, I started Introduction to AI Fundamentals with Azure course on Udacity, by a group of instructors.

I have learnt some terminologies, like:

Term	Definition	Example	Example / Source
Anomaly Detection	Detecting data points that do not align with a recognized pattern across a set of data points.	Credit-card fraud detection: A customer's transactions are normally between $10 and $200, but a sudden $8,000 transaction in an unusual location is flagged as an anomaly.	Microsoft Azure's Anomaly Detector is an example of an AI service designed to detect unusual patterns in time-series data. Note: Microsoft currently states that the Azure Anomaly Detector service will retire on October 1, 2026. (Microsoft Learn)
API	Application Programming interfaces that are exposed to help multiple applications and systems interact with each other.	Geolocation API: A website can ask the browser for the user's location without implementing the underlying GPS/location functionality itself.	MDN Web Docs provides this exact type of example with the Geolocation API. (MDN Web Docs)
Autonomous	An autonomous system is a system that can accomplish a task without human involvement during the execution of the task.	Waymo autonomous vehicle: A passenger can request a ride and the vehicle drives from pickup to destination without a human driver in the vehicle.	Waymo, the developer of the Waymo Driver, describes it as fully autonomous and capable of operating without a human driver in the vehicle. (Waymo)
Azure Cognitive Services	A suite of AI Services targeting specialized AI workloads to help developers implement AI capabilities into their solution with minimal AI and ML expertise.	Document/image analysis: A developer can use Microsoft's prebuilt vision services to extract text from an image rather than developing an OCR model from scratch.	Microsoft originally called these capabilities Cognitive Services; Microsoft now generally uses Azure AI services / Foundry Tools terminology. Microsoft describes its AI services as prebuilt capabilities that developers can access through APIs. (Microsoft Learn)
Azure Machine Learning	A cloud-based service that delivers a platform supporting the creation and deployment of Machine Learning Models.	Predicting house prices: A developer can train a regression model using historical housing data, register the trained model, and deploy it as an endpoint that applications can call for predictions.	Microsoft provides a complete Azure Machine Learning example covering training, registering, deploying, and calling a model for predictions. (Microsoft Learn)
Computer Vision	A domain of artificial intelligence focused on processing and understanding visual inputs through complex machine learning algorithms.	Traffic-camera analysis: A computer-vision system can identify cars, pedestrians, and other objects in a road image and determine their locations.	IBM describes object detection as a computer-vision task that identifies and localizes objects in images, including applications such as self-driving cars. (IBM)
Dataset	A dataset is a collection of data that is used to train the machine during the AI and ML processes.	House-price dataset: A table might contain thousands of houses, with columns such as area, number of bedrooms, location, and selling price. The data can be used to train a model to predict prices.	Google for Developers defines a dataset as a collection of examples and explains how tabular datasets can be used as input to ML models. (Google for Developers)
Experiment	In the context of AI and ML, experiment refers to using the AI and ML processes to see if a particular question can be answered with AI and ML, or a hypothesis can be supported or validated.	Comparing models: A data scientist trains several models using different algorithms or parameters and compares their accuracy to determine which approach best predicts customer churn.	Microsoft describes an ML experiment as a collection of runs that can be compared using parameters, metrics, and outputs. (Microsoft Learn)
Go	An abstract strategy-based board game.	AlphaGo: Google DeepMind developed AlphaGo, an AI system that learned to play Go and defeated world champion Lee Sedol 4–1 in 2016.	Google DeepMind documents AlphaGo's 2016 match against Lee Sedol and its significance in AI research. (Google DeepMind)
Intent	An intent represents the action the user wants to execute. It usually refers to a preferred action when a user delivers a sentence to be interpreted by the machine.	A user tells a travel assistant, "Book me a flight to Cairo next Friday." The system identifies the intent as BookFlight and extracts information such as destination and date.	Microsoft's Language Understanding documentation uses examples such as BookFlight, where different user utterances are mapped to the same intended action. (Microsoft Learn)
Jupyter Notebook	A virtual notebook environment where developers can write notes and combine live code and visualizations into a single document.	Data analysis notebook: A data scientist writes Python code to load sales data, calculates monthly revenue, displays a chart, and explains the results in Markdown cells—all in one document.	Project Jupyter defines a notebook as a shareable document combining code, plain-language descriptions, data, visualizations, and interactive elements. (Jupyter Documentation)
Machine Learning Model	A machine learning model is a logic that has been trained to recognize certain types of patterns.	House-price prediction model: After training on historical house data, the model learns relationships between features such as size and location and the selling price, then uses those patterns to predict the price of a new house.	Microsoft explains that the goal of an ML model is to identify patterns in training data and use them to make predictions on new data. (Microsoft Learn)
Machine Learning Operations (MLOps)	The practice involves the automation of processes required to build and deliver end-to-end Machine Learning Models and experiences.	Production ML system: After training a fraud-detection model, an organization automatically tests it, deploys it, monitors its performance, detects degradation, and retrains/redeploys it when necessary.	Google Cloud describes MLOps as practices covering the ML lifecycle from development through deployment and monitoring. (Google Cloud)
Natural Language Processing	A domain of artificial intelligence focused on understanding human language and helping computers and humans interact through human language.	Customer-service chatbot: A customer writes, "My package hasn't arrived yet." An NLP system processes the text, identifies its meaning and intent, and helps the system provide an appropriate response.	IBM describes NLP as enabling computers to process human language in text or speech and gives applications such as chatbots, search, voice assistants, and question answering. (IBM)
Turing Test	A test where a human tested interacts with two entities, one machine, and one human, without knowing which one is which. The goal of the test is to assess if the tester can identify which one is the machine. If the tester cannot determine the machine or is not sure, the machine passes the test.	Text conversation: A human judge communicates through text with another human and a computer. If the judge cannot reliably distinguish which participant is the machine, the machine is considered to have succeeded in the test.	The concept comes from Alan Turing's 1950 paper, "Computing Machinery and Intelligence," in which he proposed the "imitation game." (arXiv)