Training: Introduction to Large Language Models
Documentation: Generative AI on Vertex AI
Overview
What is a large language model (LLM)?
A large language model (LLM) is a statistical language model, trained on a massive amount of data, that can be used to generate and translate text and other content, and perform other natural language processing (NLP) tasks.
LLMs are typically based on deep learning architectures, such as the Transformer developed by Google in 2017, and can be trained on billions of text and other content.
What are some examples of popular large language models?
PaLM 2 is our next generation large language model that builds on Google’s legacy of breakthrough research in machine learning and responsible AI. It excels at advanced reasoning tasks, including code and math, classification and question answering, translation and multilingual proficiency, and natural language generation better than our previous state-of-the-art LLMs, including PaLM. We continue to implement the latest versions of PaLM 2 in generative AI tools like Duet AI and Vertex AI.
What are the use cases for large language models?
Text-driven LLMs are used for a variety of natural language processing tasks, including text generation, machine translation, text summarization, question answering, and creating chatbots that can hold conversations with humans.
LLMs can also be trained on other types of data, including code, images, audio, video, and more. Google’s Codey, Imagen and Chirp are examples of such models that will spawn new applications and help create solutions to the world’s most challenging problems.
What are the benefits of large language models?
What large language model services does Google Cloud offer?
Generative AI on Vertex AI: Gives you access to Google's large generative AI models so you can test, tune, and deploy them for use in your AI-powered applications.
Vertex AI Search and Conversation: Enterprise search and chatbot applications with pre-built workflows for common tasks like onboarding, data ingestion, and customization.
Contact Center AI (CCAI) : Intelligent Contact Center solution which includes Dialogflow, our conversational AI platform with both intent-based and LLM capabilities.
How It Works
LLMs work by using a massive amount of text data to train a neural network. This neural network is then used to generate text, translate text, or perform other tasks. The more data that is used to train the neural network, the better and more accurate it will be at performing its task.
Google Cloud developed products based on its LLM technologies, catering to a wide variety of use cases you can explore in the Common Uses section below.
Tips to becoming a world-class Prompt Engineer
Common Uses
Build a chatbot
Build a LLM powered chatbot
Vertex AI Conversation facilitates the
creation of natural-sounding, human-like chatbots.
Generative AI Agent is a feature within
Vertex AI Conversation that
is built on top of functionality
in Dialogflow CX.
With this feature, you can provide a website URL
and/or any number of documents, and then
Generative AI Agent parses your content and
creates a virtual agent that is powered by data
stores and LLMs.
Research and information discovery
Find and summarize complex information in moments
Extract and summarize valuable information
from complex documents, such as 10-K forms,
research papers, third-party news services,
and financial reports—with a click of a
button. Watch
how Enterprise Search
uses natural language to understand semantic
queries, offer summarized responses, and
provide follow-up questions in the demo on the
right.
Research and information discovery solution architecture
The solution uses
Vertex AI Search and Conversation
as its core component. With Vertex AI Search
and Conversation, even early career developers
can rapidly build and deploy chatbots and
search applications in minutes.
Document summarization
Process and summarize large documents using Vertex AI LLMs
With Generative AI Document Summarization,
deploy a one-click solution that helps detect
text in raw files and automate document
summaries. The solution establishes a pipeline
that uses
Cloud Vision Optical Character Recognition (OCR)
to extract text from uploaded PDF documents in
Cloud Storage,
creates a summary from the extracted text with
Vertex AI,
and stores the searchable summary in a
BigQuery database.
Build an AI-powered contact center
Build an AI-powered contact center with CCAI
Powered by AI technologies such as natural
language processing, generative AI, and text
and speech recognition,
Contact Center AI (CCAI)
offers a Contact Center as a
Service (CCaaS) solution that helps
build a contact center from the ground up. It
also has individual tools that target
specific aspects of a call center,
for example
Dialogflow CX
for building a LLM powered chatbot,
Agent Assist
for real-time assistance to human agents, and
CCAI Insights
for identifying call drivers and sentiment.
Train custom LLMs
Use TPUs (Tensor Processing Units) to train LLMs at scale
Cloud TPUs
are Google’s warehouse scale
supercomputers for machine
learning. They are optimized for performance
and scalability while minimizing the total
cost of ownership and are ideally
suited for training LLMs and generative AI
models. With the
fastest training times
on five MLPerf 2.0 benchmarks, Cloud TPU v4
pods are the latest generation of
accelerators, forming
the world's largest publicly available ML hub
with up to 9 exaflops of peak aggregate
performance.
