Slalom Gen AI Flashcards
Key terms to know for Gen AI and partner's product names for all the different components of the solutions
What does the abbreviation Gen AI stand for?
Catch all term and abbreviation for Generative Artificial Intelligence
What is Generative AI?
Category of AI solutions that generate text, images and data
What are the 7 features of a Large Language Model (LLM)?
- Text generation; 2. Translation; 3. Content summary; 4. Rewriting content; 5. Classification and categorization; 6. Sentiment analysis; 7. Conversational AI ( chatbots)
What is a Vector Database?
A vector database is a type of database that indexes and stores vector embeddings (data representations) for fast retrieval and similarity search
What does the acronym LLM stand for?
Large Language Model
What are Chat Logs?
A log of the transactions and data passed between customer, Salesforce and the APIs to the LLM
What is Salesforce Service Cloud?
Salesforce application for Customer service functions
What are Knowledge Articles?
Salesforce feature of Service Cloud that stores versions of customer service documents for use by Service teams and customers while they are online using the web and the customer’s self service site
What is OmniChannel?
Salesforce feature of Service Cloud that manages inbound requestss for servivce and routes to the available teams
What are Business Benefits of Conversational AI?
Based on National Bureau of Economic Research (Working Paper 31161 from April 20231) reports 13.8%; Virgin Voyages achieved 22% call deflection with a legacy Einstein Bot and an additional 20% call deflection with Slalom’s enhanced bot with Gen AI
What is the Acronym NLP?
Natural Language Processing
What are the 5 design principles of the Slalom Gen AI Framework for Salesforce
- Trust; 2. LLM Agnostic; 3. Guardrails; 4. Content Targeting; 5. Feedback and Refinement
What is “Slalom Gen AI Framework for Salesforce”?
Offical legally apporved name of our solution.
What are “Gen-AI Impacted Contact Center Dimensions”?
key slide in our sales pitch that visually shares how Gen AI in customer service impacts other area, such as Assisted Service, Self-Service & Digital Channel, Customer Insights, Agent Experience, Training & Onboarding, Organization Structure, Knowledge Management, Workforce Operations, Special Operations, Agent Desktop, Reporting & Analytics, Call Center Technology, and Enterprise Systems Integration
Describe the “Roadmap to Trust”; or “Onramp to Trust”
Key message in sales presentation helping customers start with Crawl, Walk, Run with GenAI starting with generic content applicable to anyone (FAQs); then moving to customer specific content in read only format (past or open orders) and eventually read/write and edit customer data or transaciotn data (update order, change shipment)
Who and what is Vivi?
Name of Virgin Voyages solution on Virgin Voyage’s Contact Us page; also name of the avatar woman in the Soul Machines demo that Slalom created for Dreamforce 2023 and use to show vision of text to speech and digital human connected to Gen AI enabled customer Service. Virgin’s Customer care team is called “Sailor Services”
Who is Soul Machines?
Soul Machines is the vendor behind the Vivi avatar, also known as Digital Person
What is the Acronym: KCS?
Knowledge Ceneterd Service. Global framework to help customer service teams us knowledge to improve customer service
What are the 5 Steps in a new Knowledge Generation Process?
- Identify Question and Answer Pairs; 2. Systematic Fact-Checking Exercise; 3. Publish Proposed Articles; Embed Knowledge Articles in Business Processes; 5. Implement Continuous Monitoring
What is the Salesforce Trust layer?
Salesforce has added a layer between your Salesforce data and the Generative AI solution delivering:
Secure data retrieval
Dynamic Grounding
Data Masking
Toxicity detection
Auditing
Zero retention
What is Secure data retrieval?
Feature of Salesforce Trust Layer, The Secure Data Retrieval Layer is designed to safeguard access to data. This layer ensures that only authorized users can access specific data, thereby preventing unauthorized access and maintaining the integrity of the information
What is Dynamic Grounding?
Feature of Salesforce Trust Layer, Dynamic Grounding is responsible for grounding the AI’s responses in real-time data, enabling it to respond accurately to current events and trends.
What is Data Masking?
Feature of Salesforce Trust Layer, uses data masking techniques to create an inauthentic but structurally similar copy of the actual data.
What is Toxicity detection?
Feature of Salesforce Trust Layer, it is designed to detect and neutralize any potentially harmful content in real time.
What is Auditing?
Feature of Salesforce Trust Layer, serves as a backbone for accountability. It’s designed to meticulously track and record all activities and actions within the system. The Audit Trail Layer operates by producing thorough documentation of activity records. This covers every query, every AI response, and every action taken by the system.
What is Zero retention
Feature of Salesforce Trust Layer, ensures that no user-specific data is retained by the system.
What is Data Cloud?
Product from Salesforce (old name Customer Data Platform or CDP; then renamed Genie) that enables entity recognition and matching of identity across data systems
What is Salesforce Prompt Builder?
Feature of Salesforce Copilot Studio that enables Admins to create reusable prompts connecetd to data in Salesforce
What is Salesforce Action Builder?
Feature of Salesforce Copilot Studio that enables Admins to create reusable actions as a result of using Prompt Builder in CRM
What is Salesforce Model Builder?
Feature of Salesforce Copilot Studio that enables Admins to create predicitve AI models for use in Salesforce
List Three Features of Salesforce Copilot Studio
1 Prompt Builder
2 Action Builder
3 Model Builder
Who is Open AI?
Vendor of GPT 3.5 and 4.0 and other Gen AI Models
What is a Foundational Model - aka FM?
Gen AI model, also referd to as LLM or Large language model. Can also refer to multi-modal models for image recognition and generation
What is Microsoft Azure OpenAI Service?
MSFT Azure capability that hosts Open AI’s language models, such as GPT 3.5
What is AWS Kendra?
Vector Database from Amazon Web Services
Who is Pinecone?
Pinecone is the vendor of PineconeIO, that offers a market leading Vector Database
Who is Zilliz and what is Milvus?
Vendor and product name of Vector database embedded in Salesforce Data Cloud vector database
What is Google Vertex AI?
Google’s brand name for Gen AI solutions - a machine learning (ML) platform that lets you train and deploy ML models and AI applications, and customize large language models (LLMs) for use in your AI-powered applications.
List Leading Vector databases
Pinecone IO, Google Vertex; AWS Kendra; Zilliz Milvus
What is Nemo?
Nemo is an LLM from nVidia
What are Tokens?
The unit of measure that is used to calcualte consumption of characters into and out of a LLM. Tokens can be thought of as pieces of words. Before the API processes the prompts, the input is broken down into tokens. These tokens are not cut up exactly where the words start or end - tokens can include trailing spaces and even sub-words. Here are some helpful rules of thumb for understanding tokens in terms of lengths: 1 token ~= 4 chars in English, 1 token ~= ¾ words, 100 tokens ~= 75 words
What are Slalom’s top Gen AI salesforce stories?
Q4 2023: Virgin Voyages for consumer generative AI Converstaional chatbot; Quidel Ortho for summarization of medical device cases during investigation
What is “Conversational AI”?
Gartner name for the software category for chat bots enabled by Gen AI, NPL and LLMs.
What are the 4 Phases of Salesforce product release cycle?
- Press Release (usually 3-6 months before GA)
- Pilot (usually 3-4 months)
- Beta (usually 3-4 months)
- GA - Generally Available (and usually takes 1-3 releases to be useful by customers)
Who is Hugging Face?
Hugging Face, Inc. is a French-American company based in New York City that develops computer tools for building applications using machine learning.
Hugging Face is a machine learning (ML) and data science platform and community that helps users build, deploy and train machine learning models. It provides the infrastructure to demo, run and deploy artificial intelligence (AI) in live applications
What is XGen?
XGen-7B is Salesforce’s LLM
What is Google Gemini?
Google’s LLM, was renamed from Bard in Feb 2024.
Gemini is a family of multimodal large language models developed by Google DeepMind, serving as the successor to LaMDA and PaLM 2
What is the Acronym RAG?
Retrieval augmented generation (RAG) is a natural language processing (NLP) technique that combines the strengths of both retrieval- and generative-based artificial intelligence (AI) models.
A technique for enhancing the accuracy and reliability of generative AI models with facts fetched from external sources.
What is Retrieval Augmented Generation?
aka RAG - Retrieval-augmented generation (RAG) is a technique for enhancing the accuracy and reliability of generative AI models with facts fetched from external sources.
What is Constitutional AI?
Term coined by Anthropic - Constitutional AI refers to a set of techniques developed by researchers at Anthropic to align AI systems like myself with human values and make us helpful, harmless, and honest.
See https://www.constitutional.ai/ for more details
Who is Anthropic?
LLM developer…
Anthropic PBC is an American artificial intelligence startup company, founded by former members of OpenAI. Anthropic develops general AI systems and large language models. It is a public-benefit corporation, and has been connected to the effective altruism movement
What is Claude, Claude instant?
Claude is a next-generation AI assistant for your tasks, no matter the scale. Our API is currently being offered to a limited set of customers and researchers.
See https://www.anthropic.com/news/introducing-claude
List popular LLMs (as of April 2024)
Open AI Chat GPT 3.5
Open AI Chat GPT 4.0
Anthropic Claude
Google Bard (renamed Gemini)
Google PaLM, PaLM 2 (renamed Gemini)
Meta’s LLaMA
Google LaMDA (Language Model for Dialogue Applications)
MosaicML’s MPT
Databricks Dolly
Others are listed here https://www.algolia.com/blog/ai/examples-of-best-large-language-models/
What is Google Vertex AI?
Google’s One AI platform with every machine learning and AI tools
What is Google Model Garden?
Google’s single place to discover, customize, and deploy a wide variety of Gen AI models. Model Garden provides a curated collection of 130+ models including enterprise-ready foundation model APIs, open models, and task-specific models from Google and 3rd parties.
What is Google PaLM?
PaLM is Pathways Language Model, a 540 billion parameter transformer-based large language model developed by Google AI
What is Google Generative AI Studio?
A solution that enables customers to customize and deploy generative models. Vertex AI provides APIs for leading foundation models, and tools to rapidly prototype, easily tune models with your own data, and seamlessly deploy to applications.
What is Google Vertex AI Search & Conversation?
Google’s tools to build generative AI search and conversations for customers and employees with Vertex AI Search and Conversation (formerly Gen App Builder).
What is Google Bard?
Bard is now Gemini
Gemini is the best way to directly access Google’s best family of AI models
What is Google Chirp?
Google’s speech-to-text models for many languages
What is Google Codey?
Codey generates code based on natural language input. Good for writing functions, classes, unit tests, and more.
What is Snowflake Cortex?
Snowflake Cortex fully-managed service that can analyze data and build AI applications. Features include Forecasting, Contribution Explorer, Document AI, Snowflake Copilot and Universal Search
Name Three GenAI features from Snowflake
- Document AI
- Universal Search
- Snowflake Copilot
What is Snowflake Streamlit?
Streamlit in Snowflake helps developers securely build, deploy, and share Streamlit apps on Snowflake’s data cloud
What is Snowflake Document AI?
New workflow to easily extract content from PDFs and other documents using a built-in LLM developed by Snowflake
What is Snowflake Universal Search?
Snowflake Universal Search enables
search and discover tables, views,
databases, schemas and available
Marketplace listings
What is Snowflake
Copilot?
Snowflake
Copilot is An LLM-powered assistant that brings the power of Generative AI to everyday coding tasks. Generates SQL from natural language, refines queries through conversation.
What is Snowsight UI?
Snowflake Snowsight provides a unified experience for working with your Snowflake data by using SQL or Python: Write and run SQL queries. Write and run Snowpark Python. Monitor query performance and history.
What is Snowflake Snowday?
Snowday is a Snowflake sponsored trade show.
What is Snowflake Snowpark?
Snowflake Snowpark is a set of libraries and runtimes that securely deploy and process Python and other programming languages in Snowflake to develop data pipelines, machine learning models, apps, and more
What is Google Dialogflow CX ?
A Dialogflow CX agent is a virtual agent that handles concurrent conversations with your end-users
What is Google Bison?
Google Bison is the best value PaLM model that handles a wide range of language tasks, such as classification, summarization. It is optimized for accuracy and latency at a reasonable cost.
What is Google Gecko?
Google Gecko is the smallest and lowest cost model for simple tasks.
What are the 4 Gen AI APIs in Google’s Vertex AI?
- PaLM api for Text
- PaLM api for Chat
- Text Embedding API
- Codey API
What are 3 ways to improve Gen AI responses and tune Google Vertex?
- Prompt design tuning - no updates to the model weights
- Prompt Tuning - base model weights are fixed, uses a vector embedding as the prompt instead of the user’s input.
- Fine Tuning - updates the model weights.
What is Prompt Design?
The art and science of figuring out what text to feed your model to get it to take on the behavior you want.
What is the difference between Zero Shot and Few Shot prompting?
The primary difference between zero-shot and few-shot prompting is the amount of task-specific data given to guide the model to help it understand and perform the task
https://shelf.io/blog/zero-shot-and-few-shot-prompting/
What does the Temperature setting in an LLM prompt impact?
Temperature impacts the randomness of the response.
A lower temperature = less random (potentially more correct, less variations);
A higher temperature = more random, possibly repetitive and unpredictable.
What Is Semantic Search?
Semantic search is a data searching technique which uses the intent and contextual meaning behind a search query to deliver more relevant results.
What is a vector embedding?
Vector embeddings are a way to convert words and sentences and other data into numbers that capture their meaning and relationships
What is Langchain?
LangChain is an open source framework for developing applications powered by large language models (LLMs).
What is Google Duet?
Gemini for Google Cloud (formerly Duet AI) Duet AI lives within Google Workspace to help you get tasks done. (Similar to a Copilot helper for Google code environments and office / workspace)
What is Mosaic AI?
Built on the Databricks Data Intelligence Platform, Mosaic AI enables organizations to securely and cost-effectively integrate their enterprise data into the AI lifecycle.
What is parameter-efficient fine-tuning (PEFT)?
Parameter-Efficient Fine-Tuning (PEFT) methods enable efficient adaptation of large pretrained models to various downstream applications by only fine-tuning a small number of (extra) model parameters instead of all the model’s parameters. This significantly decreases the computational and storage costs.
What is DeepSpeed?
DeepSpeed is a deep learning optimization library that makes distributed training easy, efficient, and effective. DeepSpeed wasD developed by Microsoft.
What is pretraining?
Pretraining is the practice of building a new LLM model from scratch to ensure the foundational knowledge of the model is tailored to your specific domain.
What is Databricks Lakehouse ?
Databricks Lakehouse AI provides one unified toolkit for building and serving AI models.
What is Databricks’ Dolly?
Dolly-v2-12b is a 12 billion parameter causal language model created by Databricks
What is Databricks MosaicML Foundation Series?
MPT is a family of LLMs from Databricks
https://www.databricks.com/blog/mpt-7b
Google IO
Google’s annual developer event
Google AI Overviews
New name of Google search results with GenAI answers above sources and links
DeepMind
DeepMind Technologies Limited, doing business as Google DeepMind, is a British-American artificial intelligence research laboratory which serves as a subsidiary of Google. https://deepmind.google/technologies/
DeepMind developed the Gemini family of models.
Veo
Google’s DeepMind Video generation solution.
ImaGen
Google’s DeepMind image generation solution.
SynthID
Google’s DeepMind image annotation and watermarking solution.
Gemma
Google DeepMind’s family of lightweight, state-of-the art open models built from the same research and technology used to create the Gemini models
AlphaFold
Google DeepMind’s AI system for treatments for diseases freely available to anyone in the scientific community
multimodal
A multimodal model is capable of processing information from different modalities, including images, videos, and text. For example, Google’s multimodal model, Gemini, can receive a photo of a plate of cookies and generate a written recipe as a response and vice versa.
What are the different versions of Google Gemini
Gemini 1.0 Ultra: Google’s largest model, designed for complex tasks
Gemini 1.5 Pro: Google’s best model for general performance across a wide range of tasks, can handle up to 1 million tokens at once, and can process large amounts of data at once, including 1 hour of video, 9.5 hours of audio, codebases with over 30,000 lines of code or over 700,000 words
Gemini 1.5 Flash: A lightweight model, optimized for speed and efficiency
Gemini Nano: Google’s most efficient model for on-device tasks
AI Agent
An intelligent agent is an agent acting in an intelligent manner. It perceives its environment, takes actions autonomously in order to achieve goals, and may improve its performance with learning or acquiring knowledge
