About Einstein Generative AI
42 Questions
1 Views

Choose a study mode

Play Quiz
Study Flashcards
Spaced Repetition
Chat to Lesson

Podcast

Play an AI-generated podcast conversation about this lesson

Questions and Answers

What is the primary function of Einstein in creating marketing content?

  • To analyze past campaign performance data.
  • To generate content such as subject lines and email copy. (correct)
  • To manage email distribution lists.
  • To design visual elements for marketing materials.

What functionality does Commerce.Semantic provide to enhance user experience during searches?

  • It offers personalized product recommendations based on previous searches.
  • It reduces the occurrence of 'no results found' by accounting for variations in search terms. (correct)
  • It provides real-time customer support during the search process.
  • It generates product reviews based on customer feedback.

How does Commerce.Smart facilitate the creation of promotions?

  • By leveraging natural language instructions and generative AI to draft promotions quickly. (correct)
  • By using a manual process to categorize products.
  • By requiring advanced programming skills to launch promotions.
  • By connecting with external marketing platforms for promotion management.

Which component does Einstein draft as part of creating an email campaign?

<p>Campaign brief and subject line. (D)</p> Signup and view all the answers

What is a key function of Commerce Concierge for Commerce Stores?

<p>To guide customers through their entire shopping experience using generative AI. (C)</p> Signup and view all the answers

In what context can Einstein generate targeted content?

<p>It can generate content for both standalone emails and SMS messages. (D)</p> Signup and view all the answers

In what manner does Data Cloud.Einstein enhance audience segmentation?

<p>By utilizing trusted customer data within Data Cloud for targeted audience segments. (D)</p> Signup and view all the answers

What type of report can Einstein utilize to provide recommended answers?

<p>Uploaded disclosure reports. (C)</p> Signup and view all the answers

What role does Education.Intelligent Question serve in program assessments?

<p>To aid in drafting assessment questions for program intakes. (A)</p> Signup and view all the answers

How does Einstein assist in segmenting the audience for a campaign?

<p>By creating a target segment based on campaign parameters. (B)</p> Signup and view all the answers

What is the primary goal of prompt engineering?

<p>To maximize the performance and reliability of models (A)</p> Signup and view all the answers

What best describes prompt injection?

<p>A technique to manipulate a model’s output using specific prompts (C)</p> Signup and view all the answers

Which of the following statements about prompt instructions is true?

<p>They consist of a verb-noun structure with a defined task. (A)</p> Signup and view all the answers

What role does a prompt template play in the use of LLMs?

<p>It includes placeholders that are filled with business data for final instructions. (A)</p> Signup and view all the answers

How does retrieval-augmented generation (RAG) enhance prompting?

<p>By enriching prompts with context from an information retrieval system. (C)</p> Signup and view all the answers

What is the primary function of prompt management tools?

<p>To effectively build, manage, package, and share prompts. (B)</p> Signup and view all the answers

Semantic retrieval allows LLMs to utilize what type of data?

<p>Similar and relevant historical business data from a CRM system. (C)</p> Signup and view all the answers

What do system cards aim to address in AI systems?

<p>The integration of multiple AI models within a holistic framework. (C)</p> Signup and view all the answers

What characterizes model deprecation in AI development?

<p>It means that a model is still functional but not recommended for use. (B)</p> Signup and view all the answers

What happens during the rerouting process?

<p>Requests to a retired model are redirected to a replacement model. (A)</p> Signup and view all the answers

When should users start migrating their applications according to the guidelines?

<p>As soon as the deprecation is announced. (B)</p> Signup and view all the answers

What is one of the first steps when configuring a model in Einstein Studio after deprecation?

<p>Choose a new foundation model instead of the deprecated one. (A)</p> Signup and view all the answers

How are model deprecation and rerouting notices communicated to users?

<p>In the monthly release notes of the Einstein Platform. (C)</p> Signup and view all the answers

What is the main purpose of vectorization in the context of a search index?

<p>To create semantic similarities from numeric representations. (C)</p> Signup and view all the answers

What should users do with testing results from the initial implementation phase?

<p>Use them for comparison against the results from the new model. (C)</p> Signup and view all the answers

What can be concluded about the nature of model deprecation and rerouting?

<p>They are considered temporary and users should adjust accordingly. (B)</p> Signup and view all the answers

What role do retrievers play in the search index process?

<p>They serve as the bridge between search indexes and prompt templates. (C)</p> Signup and view all the answers

What is the recommended action if a model is deprecated?

<p>Transition to a recommended replacement model as per the announcement. (C)</p> Signup and view all the answers

Which step directly follows the retrieval of relevant context from the indexed data?

<p>The original prompt is populated with the retrieved information. (D)</p> Signup and view all the answers

What advantage does Retrieval Augmented Generation (RAG) provide to prompt template users?

<p>It enhances generated responses with the latest proprietary data. (D)</p> Signup and view all the answers

Which of the following statements accurately describes the workflow when a prompt template with a retriever is executed?

<p>The query created from the prompt template is vectorized before retrieval. (D)</p> Signup and view all the answers

What is the default rate limit for Large Language Model (LLM) generation requests at the Salesforce Organization ID level?

<p>300 requests per minute. (C)</p> Signup and view all the answers

What can custom retrievers in Einstein Studio achieve beyond the default options?

<p>They provide enhanced search criteria and result relevance. (B)</p> Signup and view all the answers

What is a significant limitation of many Large Language Models (LLMs) mentioned in the content?

<p>They are limited to data processed within a specific timeframe. (B)</p> Signup and view all the answers

What is the primary function of the Field Service.Pre-Work Brief?

<p>To inform mobile workers about their upcoming work order. (C)</p> Signup and view all the answers

How does Contracts AI enhance the handling of legal documents?

<p>By extracting data from PDFs and drafting legal clauses. (A)</p> Signup and view all the answers

Which feature of the Marketing Cloud is aimed at content creation?

<p>Einstein Assistant for drafting email content. (A)</p> Signup and view all the answers

What benefits does Einstein for Loyalty Management provide?

<p>It summarizes loyalty programs and promotions. (C)</p> Signup and view all the answers

What capabilities does the Field Service.Dispatcher offer to users?

<p>It provides an overview of appointments needing attention. (A)</p> Signup and view all the answers

In what way does Einstein enhance the educational experience?

<p>By providing personalized skill recommendations. (D)</p> Signup and view all the answers

How does the Field Service.Post-Work Summary contribute to efficiency?

<p>By providing a detailed report of their completed tasks. (A)</p> Signup and view all the answers

What is a function of the Marketing Cloud.Subject Line & Body Copy tool?

<p>It generates subject lines and body copy for messages. (D)</p> Signup and view all the answers

Flashcards

Einstein Product Description Enhancement

Einstein analyzes and suggests revised product descriptions based on your instructions and any linked reference fields.

Commerce Search with AI

Commerce Search leverages AI to automatically account for various search variations, including synonyms, misspellings, abbreviations, and typos.

Smart Promotions with AI

This feature simplifies promotion creation by using AI-powered suggestions and natural language instructions.

AI-powered Commerce Concierge

Commerce Concierge for Commerce Stores uses AI to guide shoppers through their online experience, from discovering products to resolving issues.

Signup and view all the flashcards

Einstein Segments for Targeting

Einstein Segments helps create targeted groups by analyzing trusted customer data in Data Cloud, ensuring effective marketing campaigns.

Signup and view all the flashcards

Einstein Skills Generator (Education Cloud)

Uses AI to suggest skills that fit your learning courses and programs based on data from the public domain and your Education Cloud org.

Signup and view all the flashcards

Pre-Work Brief (Field Service)

Provides mobile workers with a brief using generative AI that tells them everything they need to know about their upcoming work order.

Signup and view all the flashcards

Post-Work Summary (Field Service)

Uses AI to create a summary of a completed job, saving time for mobile workers.

Signup and view all the flashcards

Field Service Dispatcher Actions (Field Service)

Provides a daily overview of service appointments needing immediate attention, such as appointments with rule violations, overlaps, SLA risks, or emergencies.

Signup and view all the flashcards

Contracts AI (Industries)

Uses generative AI to automate contract drafting and clause extraction from PDFs.

Signup and view all the flashcards

Einstein for Loyalty Management (Loyalty)

Uses AI to summarize loyalty programs and promotions.

Signup and view all the flashcards

Einstein Assistant (Marketing Cloud)

Uses AI to generate content for forms, landing pages, and email drafts.

Signup and view all the flashcards

Subject Line & Body Copy Generator (Marketing Cloud)

Uses AI to quickly generate subject lines and body copy for messages.

Signup and view all the flashcards

What is Einstein Copy Insights?

Einstein Copy Insights helps you write better marketing messages. It can generate subject lines, body copy, target audiences, and even entire email campaigns.

Signup and view all the flashcards

How does Einstein Copy Insights create email copy?

You provide a target audience and key message, and Einstein Copy Insights will recommend an email subject line, preheader, and body copy.

Signup and view all the flashcards

What can Einstein Copy Insights do for a campaign?

Einstein Copy Insights can create a campaign brief with target audience and key messages, draft campaign components, and generate email subject lines, preheaders, and body copy.

Signup and view all the flashcards

Can Einstein Copy Insights help with SMS messaging?

Einstein Copy Insights can generate text for SMS messages, helping you create effective mobile marketing campaigns.

Signup and view all the flashcards

What is Einstein Generative AI?

Einstein Generative AI can analyze disclosure reports and provide insights using artificial intelligence.

Signup and view all the flashcards

Deprecated Model

A model that is no longer recommended for use. This usually happens when a new and improved model is released.

Signup and view all the flashcards

Model Rerouting

The process of redirecting requests from a retired model to a newer, available alternative model.

Signup and view all the flashcards

Einstein Platform Release Notes

Published monthly, these documents outline details about deprecated and rerouted AI models within the Einstein Platform.

Signup and view all the flashcards

Reroute Date

The date marked for a model's retirement, when it will no longer be available.

Signup and view all the flashcards

Model Migration

The process of updating your application to use a new AI model, ensuring functionality remains consistent.

Signup and view all the flashcards

Rerouting Testing

Testing your application with the new model to ensure its performance and output align with your original requirements.

Signup and view all the flashcards

Foundation Model

Foundation models are AI models that serve as the base for other models, providing core functionality.

Signup and view all the flashcards

Einstein Studio Model Playground

Einstein Studio Model Playground is a tool for building and testing AI configurations.

Signup and view all the flashcards

Prompt Engineering

The process of designing and crafting effective prompts to maximize the performance and reliability of AI models, often involving a systematic and scientific approach.

Signup and view all the flashcards

Prompt Injection

A technique used to manipulate the output of an AI model by providing specific prompts that can influence its behavior, often bypassing designed constraints.

Signup and view all the flashcards

Prompt Instructions

Natural language instructions that are part of a prompt template, providing task directions for the AI model. These instructions have a verb-noun structure and define the task to be performed.

Signup and view all the flashcards

Prompt Management

Tools for the effective creation, management, packaging, and sharing of prompts, streamlining the process of AI prompt development.

Signup and view all the flashcards

Prompt Template

A string with placeholders that are replaced with business data values, resulting in a final text instruction sent to the AI model.

Signup and view all the flashcards

Retrieval-augmented Generation (RAG)

A method of grounding AI models by integrating information retrieval systems, like knowledge bases, to enhance prompts with relevant context for inference or training.

Signup and view all the flashcards

Semantic Retrieval

A situation where an AI model utilizes related historical business data from a customer's CRM system, enabling semantic understanding and context.

Signup and view all the flashcards

System Cards

An expansion of model cards to address the complexity of an entire AI system, which may involve multiple models, ensuring transparency and responsible use.

Signup and view all the flashcards

What is Vectorization in RAG?

Vectorization converts large chunks of text into numeric representations that capture semantic similarities, enabling efficient search for relevant information.

Signup and view all the flashcards

What is a Retriever in RAG?

A retriever is a key component in RAG that acts as a bridge between your search index and prompt templates. It's used to retrieve relevant information from a knowledge store.

Signup and view all the flashcards

What is a Search Index in RAG?

A search index is a structured database containing vectorized text, allowing efficient searches based on semantic similarity.

Signup and view all the flashcards

What is RAG (Retrieval Augmented Generation)?

RAG (Retrieval Augmented Generation) enhances LLMs by providing them with relevant context from a knowledge store. This context is retrieved through a retriever and incorporated into prompts.

Signup and view all the flashcards

What are Custom Retrievers?

Custom retrievers allow you to fine-tune search criteria and retrieve the most relevant information for your prompts, enhancing the quality of responses.

Signup and view all the flashcards

How does RAG work?

The RAG process involves dynamically querying a retriever, vectorizing the query, retrieving relevant context from the search index, populating the prompt with retrieved information, and submitting the augmented prompt to the LLM for response generation.

Signup and view all the flashcards

What are the benefits of RAG?

RAG allows you to bring proprietary data into an LLM's knowledge base without retraining or fine-tuning the model, making generated responses more relevant to your context and use case.

Signup and view all the flashcards

Rate limits for Einstein generative AI

Einstein generative AI applies rate limits to manage the usage of LLMs, with a default limit of 300 requests per minute at the Salesforce Organization ID level.

Signup and view all the flashcards

Study Notes

Einstein Generative AI & Trust

  • Data safety is paramount when innovating with new technology
  • Einstein Generative AI prioritizes data security and accuracy
  • Salesforce values trust as its primary principle
  • Agreements with LLM providers (like OpenAI) protect user data
  • Private data isn't retained by LLM providers

Trusted Generative AI

  • Salesforce's Einstein Generative AI solutions adhere to five principles
    • Accuracy: Model responses backed by explanations and sources
    • Safety: Detects and mitigates bias, toxicity, and harmful content
    • Transparency: Models respect data provenance and grounding
    • Empowerment: Aids users by augmenting their capabilities
    • Sustainability: Prioritizes accuracy and reduces carbon footprint

Reviewing Generative AI Responses

  • Generative AI is a tool to boost creativity, productivity, and decision-making
  • It's crucial to ensure the accuracy and helpfulness of the content, and aligning with company values.
  • AI can sometimes generate inaccurate responses (hallucinations), avoid this by checking for accuracy and ensuring grounded content (e.g. using up-to-date knowledge articles in customer service).
  • AI responses may contain bias from training data (humans) and should be checked for appropriateness (avoid harmful language)
  • Users can edit or discard responses that don't meet standards or business needs.

Einstein Generative AI Glossary of Terms

  • Artificial intelligence (AI): Computer systems using data for inference, tasks, and problem-solving resembling human reasoning
  • Bias: Systematic and repeatable errors in AI systems creating unfair outcomes due to inaccurate assumptions during machine learning.
  • Corpus: A large collection of text datasets used to train LLMs
  • Domain adaptation: Adding organization-specific knowledge to a prompt or foundation model
  • Fine-tuning: Adjusting a pre-trained language model for a specific task using smaller, task-related data
  • Generative AI gateway: Provides normalized APIs for interacting with foundation models and services from different vendors.

Generative Pre-Trained Transformer (GPT)

  • A family of language models trained on vast text data to produce human-like text.

Grounding

  • Incorporating domain-specific information or customer data into prompts.
  • Improves response accuracy by providing context.

Hallucination

  • Generating factually incorrect text that may seem semantically correct within a given context.

Temperature

  • Parameter controlling the predictability and variety of model outputs.
  • High temperature = more diverse responses
  • Low temperature = more consistent responses

Toxicity

  • Offensive, harmful, or abusive language
  • Models are trained on data including this and may generate this.

Trusted AI

  • Salesforce's guidelines for responsible AI development and implementation.

Einstein Requests

  • A usage metric for Generative AI, used for both production and sandbox environments. Using generative Al capabilities consumes Einstein Requests.
  • Data Cloud or other credits might be consumed depending on usage

Large Language Model (LLM) Support

  • Salesforce offers managed LLM models
  • Using and customizing/integrating your own models is supported (BYOLLM)
  • Using the Models API provides custom solutions for developers.

Model Deprecation

  • Process of model updates, where an existing model is phased out in favor of a newer version
  • Model providers announce deprecation with a shutdown date.
  • Rerouting of requests occurs if model is retired
  • Deprecation notices are in monthly release notes.
  • Updates should be made as soon as possible.

Geo-Aware LLM Request Routing

  • LLM requests are routed to servers closest to the Einstein Generative Al platform instance location
  • Supports OpenAI and Anthropic models in relevant regions.

Proximity and Routing

  • Closest LLM server determined by the Einstein generative AI platform instance location.

Retrieval Augmented Generation (RAG)

  • A framework for grounding large language models (LLMs) in data.
  • Improves relevance and value of LLM responses by using structured or unstructured data.
    • Retrieve relevant information
    • Augments the prompt
    • Generates a response

Offline Preparation (RAG)

  • Steps in preparing data for RAG use in Data Cloud. Load, chunk, vectorize, and index data.

Online Usage (RAG)

  • Steps that use data retrieved from the knowledge store. Query, locate, retrieves/retrieve.

Rate Limits

  • Customers have default rate limits for LLM generations per minute.

Record Pages in Lightning App Builder

  • For optimal viewing of generative AI-related data in Salesforce pages, fields shouldn't be positioned in highly constrained spaces, which may obstruct viewing.

Generative AI-Enabled Fields

  • Icons highlight fields that utilize generative AI for prompt template support

Commerce Features that use Generative AI

  • Concierge for Commerce Stores
  • Segment Creation
  • Campaign Brief & Campaign
  • Subject Lines, Preheader & Message Copy
  • Emails for Salesforce partners
  • Copilot actions for Fundraising Gift Proposals
  • Generative Al for ESG Reports

Data Cloud Features that use Generative AI

  • Einstein Segments for commerce
  • Intelligent Question Generation for Education
  • Pre-work briefs, summaries, and dispatcher actions in Field Service

Other Features that use Generative AI for various Use Cases

  • Report Formula Generation, Return Insights
  • Product Fields
  • Semantic Search for Commerce
  • Smart Promotions for Commerce
  • Contracts Al for Commerce
  • Loyalty Management
  • Marketing Cloud Engagement, Subject Lines & Body Copy, and Body Copy
  • Net Zero
  • Nonprofit & Education Cloud
  • Agentforce for Developers
  • Service Replies for Services; Summarize Records; and Work Summaries

Studying That Suits You

Use AI to generate personalized quizzes and flashcards to suit your learning preferences.

Quiz Team

Related Documents

Einstein Generative AI PDF

Description

Explore the principles of Salesforce's Einstein Generative AI, focusing on data safety, accuracy, and user trust. This quiz highlights key factors that sustain the integrity and effectiveness of generative AI technologies while ensuring user data remains confidential. > https://help.salesforce.com/s/articleView?id=sf.generative_ai_about.htm&type=5

More Like This

Guide de l'IA Générative à l'UNIGE
24 questions

Guide de l'IA Générative à l'UNIGE

InsightfulZeugma6668 avatar
InsightfulZeugma6668
Generative AI: Ethical Considerations and Risks
18 questions

Generative AI: Ethical Considerations and Risks

DignifiedWilliamsite5901 avatar
DignifiedWilliamsite5901
Use Quizgecko on...
Browser
Open
Browser
Browser