AI Associate Exam Questions and
Answers| Latest Update
Overfitting ✔️✔️Occurs when using too many variable fields in a predictive model. Overfitting
captures the noise in your data with an overly complex, unreliable way so that the model
memorizes unnecessary details. When new data comes in, the model fails. To avoid overfitting,
exclude variables that are too detailed
Underfitting ✔️✔️Often the result of an excessively simple model. The statistical algorithm
cannot capture the underlying patterns in the data
Data Cloud ✔️✔️- Embedded in SF platform
- External data lake/warehouse can now drive actions and workflows inside your CRM
- Allows you to integrate structured and unstructured data (pdfs, emails) into SF
- Can use data cloud to get generative AI - evolving as you get more data.
Einstein Discovery ✔️✔️Includes statistical modeling, supervised machine learning in a no-code
environment
Einstein Discovery: rows limits ✔️✔️Can analyze millions of rows and many columns. Ideal to
select most relevant columns.
Requires at least 400 rows *with outcomes* for analysis.
Einstein Discovery: data preparation ✔️✔️- Ideally data includes relevant factors to the business
outcome; omits extraneous factors; contains high-quality data that is representative.
- Mechanisms available in CRM Analytics data platform include:
- - Extract data from many different data sources
- - Load the data into CRM analytics datasets that you design
- - Transform the data to maximize quality and readiness for analysis
, Einstein Discovery: options/settings ✔️✔️Can update model versions, settings.
- Settings - can view rows, select variables, validation, algorithm, etc.
- First variable is outcome. Then, explanatory variables. Importance is shown (seems to be a
stand-in for correlation but there's also an option to see correlation)
- "analyze for bias" option. "Transform" option to transform data - for model only - e.g. fix typos
or update categorization
- "include only" - allows you to choose picklist values to include or omit or put into "other"
group
Model Metrics ✔️✔️To help you determine how well your model performs, Einstein Discovery
provides model metrics that visualize common measures of model performance. (Data
scientists recognize these as fit statistics, which quantify how well your model's predictions fit
the real-world data.)
Einstein Prediction Service: ✔️✔️- Public, REST API Service
- Allows you to programmatically interact with einstein - discovery-powered models,
predictions
- Can get predictions, get suggested actions, manage prediction definitions/models, manage
bulk scoring jobs, manage model refresh jobs
- Can access Einstein Prediction Service from a REST client using a connected app, REST API
request...
Ethical Maturity Model - Stages ✔️✔️Ad hoc
Organized & Repeatable
Managed & Sustainable
Optimized & Innovative
EMM Stage: Ad Hoc ✔️✔️- Individuals might raise their hand and ask "should we do this?"
- Informal advocacy builds awareness
Answers| Latest Update
Overfitting ✔️✔️Occurs when using too many variable fields in a predictive model. Overfitting
captures the noise in your data with an overly complex, unreliable way so that the model
memorizes unnecessary details. When new data comes in, the model fails. To avoid overfitting,
exclude variables that are too detailed
Underfitting ✔️✔️Often the result of an excessively simple model. The statistical algorithm
cannot capture the underlying patterns in the data
Data Cloud ✔️✔️- Embedded in SF platform
- External data lake/warehouse can now drive actions and workflows inside your CRM
- Allows you to integrate structured and unstructured data (pdfs, emails) into SF
- Can use data cloud to get generative AI - evolving as you get more data.
Einstein Discovery ✔️✔️Includes statistical modeling, supervised machine learning in a no-code
environment
Einstein Discovery: rows limits ✔️✔️Can analyze millions of rows and many columns. Ideal to
select most relevant columns.
Requires at least 400 rows *with outcomes* for analysis.
Einstein Discovery: data preparation ✔️✔️- Ideally data includes relevant factors to the business
outcome; omits extraneous factors; contains high-quality data that is representative.
- Mechanisms available in CRM Analytics data platform include:
- - Extract data from many different data sources
- - Load the data into CRM analytics datasets that you design
- - Transform the data to maximize quality and readiness for analysis
, Einstein Discovery: options/settings ✔️✔️Can update model versions, settings.
- Settings - can view rows, select variables, validation, algorithm, etc.
- First variable is outcome. Then, explanatory variables. Importance is shown (seems to be a
stand-in for correlation but there's also an option to see correlation)
- "analyze for bias" option. "Transform" option to transform data - for model only - e.g. fix typos
or update categorization
- "include only" - allows you to choose picklist values to include or omit or put into "other"
group
Model Metrics ✔️✔️To help you determine how well your model performs, Einstein Discovery
provides model metrics that visualize common measures of model performance. (Data
scientists recognize these as fit statistics, which quantify how well your model's predictions fit
the real-world data.)
Einstein Prediction Service: ✔️✔️- Public, REST API Service
- Allows you to programmatically interact with einstein - discovery-powered models,
predictions
- Can get predictions, get suggested actions, manage prediction definitions/models, manage
bulk scoring jobs, manage model refresh jobs
- Can access Einstein Prediction Service from a REST client using a connected app, REST API
request...
Ethical Maturity Model - Stages ✔️✔️Ad hoc
Organized & Repeatable
Managed & Sustainable
Optimized & Innovative
EMM Stage: Ad Hoc ✔️✔️- Individuals might raise their hand and ask "should we do this?"
- Informal advocacy builds awareness
