Prathima's Miles & Milestones

Introduction to concepts, their applications & tools

1. Generative AI makes new content using the data it was trained on.
2. Discriminative AI analyses and predicts, while Generative AI creates.
3. It uses models like GANs, VAEs, and Transformers to mimic creativity.
4. GANs (Generative Adversarial Networks) create new data by having a generator make content and a discriminator check if it looks real.
5. VAEs (Variational Autoencoders) compress data into patterns and then reconstruct it to generate new, similar content.
6. Transformers use attention mechanisms to understand context in sequences (like text) and generate coherent outputs.
7. Foundation models can be adapted for specialised tools and use cases.
8. Generative AI works across many industries and domains. It can produce text, images, audio, and video that feel realistic and relevant.
9. It can write or complete code and even generate new data to strengthen datasets.
10. It can build virtual worlds, avatars, and digital personalities with high realism.

Applications & Tools of Generative AI

In a question-and-answer format for interviews

Q: In which fields is Generative AI being applied?
 A: Generative AI is used in IT, DevOps, entertainment, finance, medicine, and HR.

Q: How is Generative AI changing the workplace?
 A: It improves efficiency, productivity, and helps us work smarter.

Q: What can Generative AI do with text?
 A: It can generate text, translate languages, summarise, and answer questions.

Q: Which tools are popular for text generation?
 A: ChatGPT (best for conversations) and Google Gemini (best for research).

Q: How does Generative AI handle images?
 A: It can create or transform images with style transfer, inpainting, or outpainting.

Q: Which image generation tools are most common?
 A: DALL-E, Stable Diffusion, StyleGAN, Bing Image Creator, and Adobe Firefly.

Q: What can Generative AI do with code?
 A: It can write new code, optimise existing code, and translate between languages.

Q: Which tools are used for code generation?
 A: GPT, GitHub Copilot, PolyCoder, IBM Watson, plus ChatGPT & Gemini for basic coding.

Q: What’s the difference between Generative AI and AI Agents?
 A: Generative AI is reactive (responds to prompts), while AI Agents are proactive (pursue goals with minimal human input).

Q: What is Agentic AI?
 A: Agentic AI uses LLM reasoning to break down complex tasks into smaller steps — a process called chain of thought reasoning.

Glossary:

A. Core Concepts

1. Machine learning (ML): Teaching computers to learn from data and make predictions.
2. Deep learning: A branch of ML that uses neural networks to learn patterns from lots of data.
3. Neural networks: Computer systems inspired by how the brain works; the core of deep learning.
4. Training data: The large set of examples used to teach an AI model.
5. Data augmentation: A Trick to boost training data variety by slightly changing existing data.

B. AI Model Types

Generative AI: AI that creates new content like text, images, audio, or video.

1. Generative AI models: Models that understand input and generate new content in response.
2. Discriminative AI: Focuses on spotting differences and classifying data into categories.
3. Discriminative AI models: Predict or classify by recognising patterns in data.
4. Foundation models: Big, general AI models that can be adapted for specific use cases.

C. Language & NLP

NLP (Natural Language Processing): AI that understands and works with human language.

1. LLMs (Large Language Models): AI trained on huge text datasets to do tasks like writing, summarising, and translation.
2. GPT (Generative Pre-trained Transformer): OpenAI’s language models that understand and create text.
3. Transformers: A type of AI architecture that excels at language and sequence-based tasks.

D. Generative Techniques

GAN (Generative Adversarial Network): Two AIs (generator + discriminator) compete: one makes content, the other checks if it’s real or fake.

1. VAE (Variational Autoencoder): AI that compresses data into a smaller form and recreates it, useful for generating variations.
2. Diffusion model: Creates images by adding noise, then learning to remove it step by step.

E. Interaction

1. Prompt: The instruction/question you give an AI to get a response.