Domains of AI

Domains Of AI

By Fatema Dhorajiwala

​The 3 domains of AI

AI operates in these three core domains depending on the type of data it has to process:

1. Data sciences: Handling and deriving insights from data.

2. Computer Vision (CV): Enabling machines to interpret and act on visual information.

3. Natural Language Processing (NLP): Teaching computers to understand and interact using human language.

​Imagine you use an app to translate text into another language. The app processes your words using NLP, and if you add a photo of text for translation, it uses CV.

The entire system relies on large datasets, which are part of data sciences.

Collection of large amount of data, organised in a way that makes it easier to process and analyse them is called a dataset.

How AI “learns” when it’s a machine

AI systems use data as input to various algorithms and the output (result of all the analysing and processing that the algorithms do) is what they “learn.” The algorithms used for this “learning” are called machine learning algorithms, or simply machine learning. And this learning is called training.

AI can learn the patterns in the data to automate the tasks.

Think of YouTube recommending next videos in the right sidebar on a computer. Have you wondered how it does that? It uses your watch history (data) to suggest content tailored to your interests. Every time you watch a YouTube video, you are training, or telling, the YT algorithm what types of videos you like. Sometimes it looks at data from users from your demographic (location, age, gender, education, ethnicity, etc.) as well.

​How data fuels AI

Large amounts of data is collected, cleaned and analysed to train machine learning models.
More data means better accuracy, as AI learns from diverse scenarios, and hence the need for large datasets.

​Real-world data is often messy and can contain errors, missing values or irrelevant information, which may mislead AI models

Data needs to be cleaned before it is usable

​For instance, a dataset of customer ages might incorrectly include negative values or impossibly high numbers. These are obviously incorrect and/or impossible. So they are removed during the cleaning process.

​Someone might write date in the “DD/MM/YYYY” format while someone else may write in “MM-DD-YYYY” format.

Data is put into standardize formats, to maintain uniformity, before storage.

Consistent data is easier and faster to process

For example, if 20% of a dataset lacks gender data and you are trying to predict gender-specific preferences using the AI model, you will not get an accurate output. Or worse, a biased output.

When faced with inaccurate data, a data scientist either removes the data completely or uses statistical techniques (such as estimation and extrapolation) to fill the missing data points.

Incomplete data can give incorrect output

For example, for a weather prediction model, storing data about the city’s population may be irrelevant.

Data analysis tools help you avoid storing repetitive and unnecessary information, which reduces project cost for storage.

Data is analysed before it is part of a dataset

​for example Amazon uses Predictive Analysis to get information on what the customers are actually looking for to grow their business.

This is also the phase where data is again checked for inconsistencies, missing entries or corrupted data.
This helps identify issues that might have been missed during data cleaning.

That’s all you need to know for now about data needed for AI models

Lets test your knowledge now

Multiple Choice

How does an AI device learn and think?

• It hacks our computer

• It accepts data and trains itself, then creates a model

• It hacks our brain

• None of the above

Multiple Choice

Voice to text is considered as ________

• Data

• NLP

• CV

Multiple Choice

The ability of a computer to understand images and videos is __________

• Data Analysis

• Nature Language Processing

• Computer Vision

Multiple Choice

Artificial Intelligence is the process that allows computers to learn and make decisions like humans

• 
  True

• False

Multiple Choice

The ability of a computer to understand human language is called _____________

• Data Analysis

• Nature Language Processing

• Computer Vision

Multiple Choice

Which of these is NOT an AI technology?

• 
  Facial recognition

• Image recognition

• Robotics

• Animation

Multiple Choice

___________________________ is to read, decipher, understand, and make sense of the human languages in a manner that is valuable.

• NLP

• CV

• Data Analysis

Multiple Choice

What is object tracking in computer vision?

• Counting the number of objects in a frame

• 
  Measuring the size of objects in a video

• Locating objects in a video sequence over time.

• Identifying objects in a single image

Multiple Choice

________ is a domain of AI related to data systems and processes, in which the system collects numerous data, maintains data sets and derives meaning/sense out of them.

• NLP

• CV

• Data Analysis

Multiple Choice

Objects tracking is considered as ________-

• NLP

• CV

• Data

Multiple Choice

Alexa is an example of _____________________ domain

• NLP

• CV

• Data Analysis

Multiple Choice

Which of the following is NOT a goal or feature of AI?

• Natural Language

• Reasoning/decision making

• Forward planning

• Displaying emotion

Multiple Select

Select the various applications of Data science in today's time

• 
  Auto-Correct

• Netflix

• Snapchat

• Weather forecasting

Multiple Choice

Face Lock in Smartphones is an example of ________________

• NLP

• CV

• Data