Summary

We started this chapter by introducing the fundamental topics that underpin GenAI, including concepts such as embeddings and latent space. We then described what GenAI is and how it contrasts against “traditional AI,” whereby traditional AI typically tries to predict a specific answer, such as a revenue forecast based on historical data or identifying whether a image contains a cat, but GenAI goes beyond those kinds of tasks and creates new content.

We dived into the role of probability in GenAI versus traditional AI, and we discussed how traditional AI often uses conditional probability to predict the values of a target variable based on the values of features in the dataset. On the other hand, GenAI approaches typically try to learn the joint probability distribution of both the features and the target variable.

Next, we explored the evolution of GenAI and various model development milestones that led to the kinds of models we use today. We began this...