AI Concepts and Project Framework

Examples of expert systems and their uses:

1. INTERNIST-1: It is a medical expert system used for the diagnosis of complex internal medicine diseases by analysing patient symptoms.

2. HEARSAY-II: An expert system for speech recognition that combines multiple knowledge sources to interpret spoken language.

3. DART: Used in engineering and military logistics to plan and schedule complex operations.

4. PUFF: A medical expert system that helps in diagnosing lung function problems using pulmonary test data.

5. CADEX: A chemical expert system used to assist scientists in identifying chemical compounds.

The given applications are classified below:

True

Reason — Broad AI is also known as Strong AI or Artificial General Intelligence. It refers to artificial intelligence systems that have human-like cognitive abilities and can understand, learn, and accomplish any intellectual work that a human can perform.

False

Reason — In AI project development, the collected data is divided into two forms: data for training (training dataset) and data for testing (testing dataset). Complete data is not used only for training, a separate testing dataset is used to test the performance of the trained model.

True

Reason — An automatic answering system works by using predefined rules and knowledge to provide responses to user queries. Such systems are designed to offer specialised problem-solving capabilities, which is a characteristic of a basic expert system.

False

Reason — Computer vision enables computers and machines to perceive, analyse, and extract meaningful information from visual data, such as images and videos. Speech signals are processed using Natural Language Processing, not computer vision.

True

Reason — Image segmentation is the process by which computer vision algorithms divide images into meaningful sections or segments, allowing for more detailed analysis and understanding of the visual content.

True

Reason — Video subtitles involve processing visual data from videos as well as understanding and generating natural language text. Therefore, video subtitles are an example of both Computer Vision and Natural Language Processing.

True

Reason — Natural Language Processing is used to process natural language that may be in the form of text or speech. Voice search on YouTube is an example of Natural Language Processing.

True

Reason — Autonomous vehicles use artificial neural network based algorithms. Neural networks are critical components of autonomous driving systems and help in real-time object detection and tracking, lane detection, and decision-making.

True

Reason — In any AI project, data is an important aspect. In machine learning and deep learning systems, data is used to train the algorithms. The success of the AI project highly depends on the quality, authenticity, and relevance of the data.

False

Reason — Alexa is designed to perform specific tasks such as answering questions, setting reminders, and responding to voice commands. Since it works in a specialised domain and does not have human-like general intelligence, it is an example of Narrow (Weak) AI, not Broad AI.

Weak AI

Reason — Narrow AI refers to artificial intelligence systems that are designed to perform a single task or a set of closely related tasks. Such task-specific AI systems are also known as Weak AI.

Expert system

Reason — Eliza is an early artificial intelligence program designed to simulate human conversation using rule-based pattern matching. It works by applying predefined rules and responses to user input, which is a characteristic of an expert system.

Narrow AI

Reason — A weather prediction system is designed to perform a specific task, that is, predicting weather conditions using available data. Since it works in a specialised domain and cannot perform multiple human-like tasks, it comes under the category of Narrow AI.

Both a and b

Reason — Computer vision enables computers and machines to perceive, analyse, and extract meaningful information from visual data, such as images and videos.

Image classification system

Reason — An image classification system is designed to identify and categorise images using image processing and computer vision techniques. It does not use rule-based reasoning or a knowledge base to provide specialised problem-solving like an expert system. Therefore, it is not an expert system.

Natural Language Processing

Reason — Gmail’s subject suggestion involves understanding and processing written text and predicting suitable words or sentences. This task deals with analysing and generating natural language, which is an application of Natural Language Processing.

Human brain

Reason — Artificial neural networks are inspired by the structure and functioning of biological neural networks in the human brain.

Computer vision

Reason — Computer vision enables computers and machines to perceive, analyse, and extract meaningful information from visual data such as images. Since X-ray images are visual data, disease diagnosis through X-ray images is an application of computer vision.

Problem definition and scoping

Reason — Problem definition and scoping is the first step in the AI project framework. It is the process by which we identify the problem that needs to be solved and define its scope, objectives, and limitations before developing an AI project.

All of these

Reason — The performance of a developed AI model can be evaluated using different performance metrics. Accuracy, precision, and recall are commonly used measures to assess how well a model performs.

1. In biological neurons, dendrites are responsible for accepting inputs.
2. In AI project development, collected data is split into training and testing data.
3. Broad AI refers to artificial intelligence systems that have human-like cognitive capacities.
4. Rule-based reasoning is used in expert systems to make decisions.
5. Image recognition is possible through computer vision.

Broad AI, also known as Strong AI or Artificial General Intelligence, refers to artificial intelligence systems that have human-like cognitive abilities and can understand, learn, and accomplish any intellectual work that a human can perform. Broad AI means to imitate human intelligence across multiple domains, allowing the system to adapt and apply its knowledge and skills to a variety of scenarios and contexts.

Examples of Broad AI can be seen in virtual world like in movies where robots do all human tasks, like heading an organisation, managing workforce, etc.

Two examples of expert system are:

1. MYCIN – An early expert system that diagnoses bacterial infections and recommends antibiotic therapy.
2. DENDRAL – A chemical expert system that analyses mass spectrometry data and identifies organic molecules.

Data acquisition is the stage of an AI project in which the required data is identified and collected for building the AI solution. In this stage, suitable data sources are selected and, if data is not available, it is gathered using methods like surveys, interviews/meetings, observations or experiments. The collected data must be relevant, authentic and of good quality, because the accuracy and success of the AI system depends on it.

The collected data may have noisy or incomplete information. It may also contain information that is not relevant to the problem. Data cleaning helps in handling all these issues so that the data becomes suitable and useful for developing an AI project.

Comparison between narrow and broad AI:

Some important applications of computer vision are:

1. Image recognition: Computer vision algorithms can analyse and identify objects, sceneries, and patterns from an image or video. Object identification, image categorisation, facial recognition, and visual search are examples of image recognition.
2. Object tracking: Computer vision algorithms can track and follow certain objects or subjects in real-time camera feeds or videos. This is important in applications such as surveillance, self-driving cars, and augmented reality.
3. Image segmentation: Image segmentation is the process by which computer vision algorithms divide images into meaningful sections or segments, allowing for more detailed analysis and understanding of the visual content.
4. Scene comprehension: Computer vision techniques allow machines to understand and interpret the context and semantics of a visual scene. This includes tasks such as scene recognition, scene comprehension and context-based image comprehension.
5. Three-dimensional (3D) modelling: Computer vision algorithms can recreate three-dimensional models of objects or scenes from two-dimensional photographs or video recordings. This is useful in 3D mapping, virtual reality, and robotics.
6. Visual inspection and quality control: In industries, computer vision systems are used for automated inspection, defect identification, and quality control. They can detect flaws, irregularities, or errors in products or production processes.
7. Medical imaging: Computer vision is used for activities such as tumour identification, diagnosis, and analysis of medical scans like X-rays, CT scans, and MRIs.
8. Robotics: Computer vision is essential in robotic systems as it allows robots to observe and interact with their surroundings. Vision-based robots can manoeuvre, grip objects, recognise objects, and work with humans.

Natural Language Processing (NLP) helps a computer understand, interpret and respond to human language (both spoken and written). It converts human language into a form that the computer can process, so that meaningful actions or responses can be produced.

Example: In voice search on Google/YouTube, when we speak a query like "Show ICSE Class 9 physics videos", the system understands the spoken words and performs the required search.

Artificial neurons are similar to biological neurons in the way they receive, process, and transmit information. In biological neurons, inputs are accepted through dendrites and passed to the nucleus. If the summation of inputs crosses a particular threshold, an output is generated and passed to the next neuron through the axon and axon terminals.
Similarly, in artificial neurons, input signals are received, their sum is calculated, and an activation function (usually for thresholding) is applied. If the condition is satisfied, the output is passed to the next neuron. Just like biological neurons form a network in the human brain, artificial neurons are interconnected to form artificial neural networks, which help machines learn, analyse, and make decisions.