Introduction to Generative AI
Course Title:
Introduction to Generative AI
Course Duration:
3 months (12 weeks)
Frequency:
2 classes per week
Course Content
Week 1: Introduction to Generative AI
- Lecture 1 : What is Generative AI?
- Overview of AI and Machine Learning.
- Applications and Use Cases.
- Tools Used in Generative AI
- Lecture 2 : Descriminative VS Generative Models
- Generative Models: What are they and why are they important?
- Exercise: Build a simple text generation model.
Week 2: Understanding Neural Networks
- Lecture 1: Deep Learning and Neural Networks: An Overview
- Deep Learning Basics and Neural Networks Architecture.
- Feedforward and Backpropagation.
- Tools: TensorFlow, Keras.
- Lecture 2: Generative Models in Image Generation: Understanding GANs and VAEs
- Use Case: How Generative Models are used in image generation (GANs, VAEs).
- Exercise: Build a simple Neural Network using Keras.
Week 3: Autoencoders
- Lecture 1: Autoencoders: How They Work and Why They Matter
- Understanding Autoencoders.
- Different types: Variational Autoencoders (VAEs).
- Tools: TensorFlow, Keras.
- Lecture 2: Building and Training Autoencoders for Data Compression
- Use Case: Data Compression and Reconstruction with Autoencoders.
- Exercise: Build an Autoencoder for image compression.
Week 4: Generative Adversarial Networks (GANs)
- Lecture 1: Introduction to GANs: Theory and Architecture
- Introduction to GANs: Architecture, Generator, and Discriminator.
- Real-World Applications of GANs.
- Tools: PyTorch, TensorFlow.
- Lecture 2: Hands-on: Building a Simple GAN for Image Generation
- Use Case: Image Generation using GANs.
- Exercise: Build a basic GAN for generating images.
Week 5: Advanced Generative AI Techniques
- Lecture 1: Overcoming Challenges in GANs: Techniques for Improvement
- Issues with GANs: Mode Collapse, Training Instabilities.
- Solutions: Wasserstein GAN, Conditional GAN.
- Tools: Tensorflow.
- Lecture 2: Building Conditional GANs: A Practical Guide
- Use Case: Generating specific image categories with Conditional GAN.
- Exercise: Build a Conditional GAN for a specified dataset (e.g., Fashion MNIST).
Week 6: Sequence Models and Generative Text Models
- Lecture 1: RNNs, LSTMs, and GRUs: Understanding Sequence Models
- Introduction to Recurrent Neural Networks (RNNs), LSTMs, and GRUs.
- Sequence-to-Sequence Models and Language Models.
- Tools: TensorFlow, Keras.
- Lecture 2: Text Generation with RNNs: Hands-on Implementation
- Use Case: Text Generation and Completion.
- Exercise: Train an RNN for simple text generation tasks.
Week 7: Transformers and Attention Mechanism
- Lecture 1: Transformers: Revolutionizing AI with Attention Mechanisms
- Understanding Transformers: Self-Attention and Encoder-Decoder Architecture.
- From Transformers to GPT (Generative Pretrained Transformer).
- Tools: Hugging Face Transformers.
- Lecture 2: Implementing Transformers for Language Translation
- Use Case: Language Translation, Text Summarization.
- Exercise: Implement a transformer model for text translation.
Week 8: Large Language Models (GPT, BERT)
- Lecture 1: Exploring Large Language Models: GPT and BERT
- Overview of GPT, BERT, and other large language models.
- How to fine-tune pretrained models.
- Tools: Hugging Face, OpenAI API.
- Lecture 2: Fine-tuning GPT for Custom Text Generation Tasks
- Use Case: Chatbots and Conversational AI.
- Exercise: Fine-tune GPT for custom text generation tasks.
Week 9: AI Art and Creative Applications
- Lecture 1: Generative AI in Art: From DeepDream to Neural Style Transfer
- Generative AI in Art: DeepDream, Neural Style Transfer.
- Tools: Runway ML, Artbreeder, DALL·E.
- Lecture 2: Hands-on: Creating AI-Generated Art with Style Transfer
- Use Case: Creating AI-generated artwork.
- Exercise: Use Neural Style Transfer to blend styles between images.
Week 10: Generating Music and Audio
- Lecture 1: Audio and Music Generation with AI: Tools and Techniques
- Audio Generation with Generative AI: WaveNet, Jukedeck.
- Tools: Magenta, MuseNet.
- Lecture 2: Hands-on: Building a Music Generator with Magenta
- Use Case: AI-generated music and sound effects.
- Exercise: Build a simple music generation model using Magenta.
Week 11: Ethics and Limitations of Generative AI
- Lecture 1: Ethics in Generative AI: Bias, Deepfakes, and Data Privacy
- Ethical concerns in Generative AI: Deepfakes, Bias, Data Privacy.
- Tools: OpenAI API, AI Ethics frameworks.
- Lecture 2: Case Study: Assessing Ethical Implications of Generative Models
- Use Case: Assessing bias in generative models.
- Exercise: Analyze the ethical implications of a given generative AI model.
Week 12: Capstone Project and Deployment
- Lecture 1: Deploying Generative AI Models: Tools and Best Practices
- How to deploy generative AI models.
- Tools: Docker, AWS, TensorFlow Serving.
- Lecture 2: Capstone Project: Building and Deploying Your Own Generative Model
- Capstone Project: Students create and deploy their own generative model (e.g., an image generator or text generator).
- Review and Final Q&A.