Notes on Introduction to Generative AI from Coursera

I am now taking the Introduction to Generative AI Learning Path Specialization courses from Coursera, which are composed of three parts. The first part is called Introduction to Generative AI. Here are some notes I have taken from this course:

1. There are two types of deep learning models: discriminative models and generative models. The discriminative model is used for classification or prediction and is often trained with labeled data. The neural network aims to identify the relationship between the features of the data and the corresponding labels. On the other hand, the generative model creates new data similar to the data on which it was trained, aiming to understand the underlying distribution of the data.
2. Generative AI can be supervised, semi-supervised, or unsupervised learning. It can be a generative language model or a generative image model.
3. One big challenge of generative AI is that it may generate “hallucinations” due to improper training data, lack of context, or constraints in the prompt.
4. The foundation model plays an important role in generative AI, and it can be adapted for specific tasks. Generative AI language foundation models include BERT, PaLM, etc. Generative AI vision foundation models include Stable Diffusion v1–5, CLIP, etc.