VISION INSTITUTE OF TECHNOLOGY, Subject: Artificial intelligence

ALIGARH

Unit: 5

APPLICATIONS:
AI applications – Language Models – Information Retrieval- Information Extraction – Natural
Language Processing – Machine Translation – Speech Recognition – Robot – Hardware – Perception
– Planning – Moving

AI Applications: Overview

Artificial Intelligence (AI) has become a transformative technology, impacting various

industries and domains. Below are key areas where AI is widely applied:

1. Natural Language Processing (NLP)

• Applications: Chatbots, Virtual Assistants (e.g., Siri, Alexa), Sentiment Analysis, Machine
Translation, Text Summarization, Speech Recognition
• Examples: OpenAI's GPT, Google's BERT, IBM Watson

2. Language Models

• Definition: AI systems trained on vast amounts of text data to understand, generate, and
process human language.
• Applications:
o Content Generation (e.g., articles, poetry)
o Conversational AI (e.g., customer support chatbots)
o Code Assistance (e.g., GitHub Copilot)
• Examples: GPT (Generative Pre-trained Transformer), LLaMA (Large Language Model Meta
AI), Claude (Anthropic)

3. Information Retrieval (IR)

• Definition: The process of obtaining relevant information from large datasets based on user
queries.
• Applications:
o Search Engines (e.g., Google, Bing)
o Recommendation Systems (e.g., Netflix, Amazon)
o Enterprise Search Tools (e.g., ElasticSearch)
• AI Techniques Used: Natural Language Understanding (NLU), Semantic Search, Vector
Embeddings

Page1 Faculty: Kajal saxena

VISION INSTITUTE OF TECHNOLOGY, Subject: Artificial intelligence
ALIGARH

Unit: 5
4. Information Extraction (IE)

• Definition: Automatically extracting structured information from unstructured text.

• Applications:
o Named Entity Recognition (NER)
o Relationship Extraction
o Event Extraction
• Examples: AI systems scanning medical records to identify patient information or extracting
financial data from reports.

5. Computer Vision

• Applications: Object Detection, Image Recognition, Facial Recognition, Autonomous

Vehicles, Medical Imaging Analysis
• Examples: Tesla Autopilot, Google Lens, OpenCV

6. Robotics and Automation

• Applications: Industrial Robots, Warehouse Automation, Humanoid Robots, Surgical Robots

• Examples: Boston Dynamics’ Spot Robot, Amazon's Warehouse Robots

7. Healthcare

• Applications: Disease Diagnosis, Drug Discovery, Personalized Medicine, Virtual Health

Assistants
• Examples: IBM Watson Health, AI in Radiology

8. Finance

• Applications: Fraud Detection, Algorithmic Trading, Credit Scoring, Risk Assessment

• Examples: PayPal's Fraud Prevention System, Bloomberg Terminal

9. Autonomous Vehicles

• Applications: Self-driving Cars, Drone Navigation, Traffic Management

• Examples: Waymo (Google's Self-Driving Car), Tesla Autopilot

Page2 Faculty: Kajal saxena

VISION INSTITUTE OF TECHNOLOGY, Subject: Artificial intelligence
ALIGARH

Unit: 5
10. Gaming and Entertainment

• Applications: AI Game Characters, Personalized Recommendations, Content Creation

• Examples: OpenAI Five (Dota 2), Netflix Recommendations

11. Education

• Applications: Intelligent Tutoring Systems, Personalized Learning, AI-Powered Grading

• Examples: Duolingo, Khan Academy

12. Cybersecurity

• Applications: Threat Detection, Malware Analysis, Anomaly Detection

• Examples: Darktrace, CrowdStrike

Language Models: Overview

Language Models (LMs) are a type of Artificial Intelligence (AI) system designed to
understand, generate, and process human language. They are built using large datasets and
advanced machine-learning algorithms, primarily based on Natural Language Processing
(NLP) techniques.

1. What Are Language Models?

• Definition: A language model predicts the probability of a sequence of words appearing in a

sentence.
• Goal: To understand context, semantics, and relationships between words to perform tasks
such as text generation, summarization, and translation.
• Example: Given the phrase "The cat sat on the ___," a language model might predict "mat"
as the most likely next word.

2. Types of Language Models

a. Statistical Language Models

• Description: Early models based on probability and statistical relationships.

• Examples: N-gram models, Hidden Markov Models (HMM)

Page3 Faculty: Kajal saxena

VISION INSTITUTE OF TECHNOLOGY, Subject: Artificial intelligence
ALIGARH

Unit: 5
b. Neural Language Models

• Description: Use deep learning architectures for better context understanding.

• Examples:
o Recurrent Neural Networks (RNNs)
o Long Short-Term Memory (LSTM)
o Transformer Models (e.g., GPT, BERT)

3. Key Transformer-Based Models

a. GPT (Generative Pre-trained Transformer)

• Developer: OpenAI
• Purpose: Text generation, summarization, chatbot interactions
• Strength: Generates human-like text, understands context deeply

b. BERT (Bidirectional Encoder Representations from Transformers)

• Developer: Google AI
• Purpose: Question answering, text classification, sentiment analysis
• Strength: Processes text in both directions (left-to-right and right-to-left)

c. LLaMA (Large Language Model Meta AI)

• Developer: Meta (Facebook)

• Purpose: Open research, general language tasks
• Strength: Designed to be efficient and lightweight for research

d. T5 (Text-to-Text Transfer Transformer)

• Developer: Google
• Purpose: Text summarization, question answering, classification
• Strength: Treats all NLP tasks as text-to-text problems

4. Applications of Language Models

a. Text Generation

• Creative writing, story generation, and article creation

b. Chatbots and Virtual Assistants

• Customer support (e.g., ChatGPT, Alexa, Siri)

Page4 Faculty: Kajal saxena

VISION INSTITUTE OF TECHNOLOGY, Subject: Artificial intelligence
ALIGARH

Unit: 5
c. Information Retrieval

• Search engines, document retrieval systems

d. Machine Translation

• Real-time translation (e.g., Google Translate)

e. Sentiment Analysis

• Product reviews, social media analysis

f. Code Generation

• Assisting developers (e.g., GitHub Copilot)

5. Training Language Models

a. Pre-training Phase:

• Model learns patterns from massive text datasets.

• Example datasets: Wikipedia, Common Crawl

b. Fine-tuning Phase:

• Model is adjusted for specific tasks or domains (e.g., medical, legal language).

c. Reinforcement Learning from Human Feedback (RLHF):

• Helps align model outputs with human intent.

6. Challenges with Language Models

• Bias: Models can inherit biases from training data.

• Ethical Concerns: Misinformation, harmful content generation.
• Resource-Intensive: High computational power and energy requirements.
• Interpretability: Hard to fully understand how decisions are made.

7. Future of Language Models

• Multimodal Models: Combining text, image, and audio (e.g., GPT-4, Gemini).
• Efficiency Improvements: Reducing energy and computational requirements.

Page5 Faculty: Kajal saxena

VISION INSTITUTE OF TECHNOLOGY, Subject: Artificial intelligence
ALIGARH

Unit: 5
• Domain-Specific Models: Tailored models for healthcare, finance, etc.

• Robot: This is the physical embodiment, the actual machine or system, often consisting of
hardware components like actuators, sensors, processors, and power sources.

• Hardware: Refers to the physical components of the robot, such as:

• Actuators (motors, servos) for movement.

• Sensors (cameras, LIDAR, accelerometers) for perception.
• Processors (computers, microcontrollers) for controlling operations.

• Perception: This is the robot’s ability to sense and interpret its environment. It includes:

• Sensors that gather data (e.g., visual, auditory, haptic).

• Algorithms to process raw sensor data (e.g., computer vision, object recognition).
• Understanding of the environment (mapping, localization, detecting obstacles).

• Planning: The robot needs to make decisions on what actions to take based on its
perceptions. This includes:

• Path planning: Deciding how to move from one location to another while avoiding
obstacles.
• Task planning: Determining the sequence of actions to achieve a goal (e.g., pick-
and-place tasks, navigation).
• Decision-making algorithms (e.g., reinforcement learning, planning under
uncertainty).

• Moving: This is the execution phase, where the robot carries out planned actions. It
involves:

• Movement control: Translating plans into motor commands to move the robot.
• Feedback systems: Adjusting motion in real-time based on sensory feedback (e.g.,
adjusting speed, avoiding obstacles).
• Motion planning: Managing smooth, efficient movement trajectories.

Page6 Faculty: Kajal saxena