1. Review of Previous Topics
2. Understanding of Data Types
- 3D Point Cloud
- Learning about spatial data and how AI uses three-dimensional shapes.
3. Neural Network Training Basics
- Dimensions of Input Data
- Exploring how different types of data have various dimensions.
- Types of Dataset
- Understanding and dividing data into training, validation, and test sets.
- Structure of Dataset
- Learning how data is organized into features and labels for training.
- Overfitting and Underfitting
- Understanding the learning quality and generalization of neural networks.
- Learning Rate
- Learning about a parameter that determines the speed at which a model learns.
4. Types of Neural Networks
- Generative Adversarial Networks (GAN)
- AI models that generate new data resembling the training data.
- Transformer
- Advanced models for processing sequences, especially in language.
- GPT (Generative Pre-trained Transformer)
- Architecture behind advanced language models like ChatGPT.
5. NLP Basics
- Parts-of-speech Tagging
- Identifying grammatical parts of words in sentences.
- Vectorization
- Converting text into numerical vectors for processing.
- Embeddings
- Representing words in a way that captures their meanings and relationships.
6. Large Language Models
- Retrieval Augmented Generation
- Learning about enhancing AI responses using external data sources.
- Zero-shot Prompting
- Understanding how AI can perform tasks without prior specific training.
- Fine Tuning
- Learning about adjusting pre-trained models for specialized tasks.
- LLM Parameters
- Exploring settings like temperature, top_p, and top_k that control output randomness and creativity.
7. Use of AI in Robotics
- Manipulation
- Learning how robots handle and move objects.
- Conversational Intelligence
- Understanding that robots can engage in human-like dialogue.
- Task Planning
- Learning how robots plan and execute steps to achieve goals.
8. Limitations of LLM
- Limited Reasoning
- Recognizing challenges with complex problem-solving and logical reasoning.
- Inability to Learn After Training
- Understanding that AI models may not update knowledge in real-time after initial training.
- Lack of Reverse Prompting
- Learning about difficulties in adjusting context based on user feedback.
9. Types of Problems AI Needs to Solve
- Time Series Forecasting
- Predicting future data points based on past trends.
- Pose Estimation
- Determining the position and orientation of objects or body parts.
- Gesture Recognition
- Interpreting human gestures and movements.
- Object Detection/Recognition/Segmentation
- Identifying and isolating objects within images.
- Anomaly/Outlier Detection
- Finding unusual patterns that do not conform to expected behavior.
- Weather Forecasting
- Using AI to predict weather conditions.
- Reasoning
- Learning how AI solves problems through logical deduction and inference.
10. AI Development Process
- In-depth Review
- A comprehensive examination of each stage in developing AI systems.
11. Safe and Fair AI
- Socially Beneficial
- Ensuring AI contributes positively to society.
- Accountability
- Establishing responsibility for AI decisions and actions.
- AI Misuse
- Discussing ethical concerns, including AI in weapons and surveillance.