With Einstein Studio, Model Builder and all the recently announced Low Code and Pro Code Model Training features on AWS Sagemaker and Google Vertex AI, I am confused what the use cases are compared to AI features that exist for many years like Einstein Discovery, Prediction and Recommendation Builder.
Will this new version of AI tools slowly replace the old generation?
Einstein Studio has an Add Predictive Model action which allows you to either connect to an external predictive model, or to create a model from scratch. When you choose to create a model from scratch, you are led through the configuration of a predictive model with essentially the same features as have previously existed in Einstein Prediction Builder. The major difference is that all of the data sources for training and test data, as well as data sources when you're ready to make inferences, are the Data Cloud DMOs.
The Einstein Prediction Builder feature which has been in the platform for several years will probably continue to exist for customers who have no interest in using Data Cloud, and have good data management practices within Salesforce Core. But it is inherently more limited due to its needing to work with data within the legacy Salesforce object storage. My own prediction is that we will see the Einstein Model Builder take on a greater diversity of data science use cases and product evolution in years to come, and that the Einstein Prediction Builder will always be limited to its current feature set of regression and classification only.
