Fundamentals of GenAI Webinar - 28-0948

Download as pdf or txt
Download as pdf or txt
You are on page 1of 22

The Fundamentals of Generative AI

Tom Wyrwich
Senior Corporate Counsel
EMEA AI & Strategic Partnerships, Microsoft
Agenda

 Machine learning & generative AI: the basics

 Large language models: how they work

 Microsoft’s Generative AI Services

 Key legal considerations for generative AI

 How legal departments are using generative AI

 Q&A

Microsoft AI Webinars | The Fundamentals of Generative AI


AI & ML  Artificial Intelligence or AI solves tasks that
require human intelligence
 Machine Learning or ML is a subset of
artificial intelligence that solves specific tasks
by learning from data and making
predictions

Microsoft AI Webinars | The Fundamentals of Generative AI


Microsoft has released AI-powered features
for decades. You use AI 100s or 1000s of
times per day:
 Bing search
 Artificial background, echo cancellation,
"lower your hand" in Teams
AI experiences
 Translation services
over the years
 “Did you forget an attachment” in Outlook
 Autocomplete: Predictive text
 Speech to text
 Your LinkedIn feed

Microsoft AI Webinars | The Fundamentals of Generative AI


AI and ML through the years
1950s
Artificial Intelligence
Artificial Intelligence the field of computer science that seeks to create intelligent
machines that can replicate or exceed human intelligence.

1997
Machine Learning Machine Learning
subset of AI that enables machines to learn from existing data
and improve upon that data to make decisions or predictions.

2017
Deep Learning Deep Learning
a machine learning technique in which layers of neural networks
are used to process data and make decisions.

2021
Generative AI
create new written, visual, and auditory content given prompts
Generative AI or existing data.

Microsoft AI Webinars | The Fundamentals of Generative AI


What is Generative AI
(As told by Copilot for M365)

Generative AI is a type of artificial intelligence that can create new things, like
pictures, text, or speech, that are similar to examples it has seen before.

It does this by learning from a set of examples, figuring out the patterns and rules
that make them similar, and then using those patterns and rules to make new
examples that are similar to the ones it learned from.

It's different from other types of AI because it can create new things, instead of
just recognizing or classifying things it has seen before.

Microsoft AI Webinars | The Fundamentals of Generative AI


AI in a Minute: Large Language Models
Watch here

Microsoft AI Webinars | The Fundamentals of Generative AI


Common Generative AI terms
Response (or “completion” or
Prompt (or “input”): the
“output”): the generated
question or instruction the user
content with which the model
provides the model
responds to the prompt

Fine-tuning: a process of using


Grounding: data provided with
data to further train a model
a prompt to provide it context
so that it provides more specific
for responding to that prompt
responses in a given domain

Metaprompt: an instruction to Red-teaming: An adversarial


the model intended to apply to system of rigorously testing a
all prompts/responses model prior to deployment

Microsoft AI Webinars | The Fundamentals of Generative AI


A word on OpenAI’s Large Language Models
 Microsoft makes the LLMs that OpenAI trained available in its
products as a licensee.

 Microsoft and OpenAI are both free to commercialize those models


as they wish.

 Microsoft does not share any Customer Data with OpenAI.

Microsoft AI Webinars | The Fundamentals of Generative AI


How enterprises use Generative AI with Microsoft

Develop or enhance their own Reinvent everyday experiences with Unlock the full power of Web search
applications using GenAI models their cloud applications
Azure OpenAI Service Copilot applications like: Copilot (formerly Bing Chat Enterprise)
Azure AI Studio (preview) Copilot for M365, Copilot for Sales, Copilot for
Service, GitHub Copilot, Copilot for Security
(preview), and
Copilot features for Dynamics 365, Microsoft Fabric,
and the Power Platform.

Microsoft AI Webinars | The Fundamentals of Generative AI


Deployed in your Azure subscription,
secured by you, and tied to your datasets
and applications

Azure OpenAI Large, pretrained AI models to unlock


new scenarios
Service
AI models custom-tunable with your
data and hyperparameters

Built-in responsible AI to detect and


GPT-3.5 Turbo GPT-4 mitigate harmful use

DALL·E Whisper
Enterprise-grade security with role-based
access control (RBAC) and private networks

Microsoft AI Webinars | The Fundamentals of Generative AI


Copilot for M365 Dynamics 365 Copilot

Empower everyone with a Specialized copilots for


copilot that works every role and function
alongside you

A copilot for Copilot in Power Platform Copilot for Security


every Microsoft Imagine it, describe it, and Defend at machine speed
Cloud Power Platform builds it

experience
Windows Copilot GitHub Copilot

The first centralized AI Increase developer


assistance on a platform productivity to accelerate
innovation

Microsoft AI Webinars | The Fundamentals of Generative AI


Generative AI: key legal considerations

Privacy and Data Confidentiality Intellectual Property Responsible AI

Using Generative AI without compromising Leveraging Generative AI to create new and Deploying Generative AI in a safe, fair,
your commitments to, and obligations more efficient content without violating the secure, inclusive, transparent, and
concerning, personal and sensitive data. rights of authors. accountable manner.

Microsoft AI Webinars | The Fundamentals of Generative AI


 Our generally available Generative AI Services are built
Overview: on Microsoft Azure and therefore meet our industry-
Privacy and standard privacy and security commitments.

Data  Copilot for M365 and Azure OpenAI Service are Core
Confidentiality Online Services and EU Data Boundary Services.

 Copilot for M365 leverages your existing data access


controls to leverage only the data each user already has
access to (with additional controls available).

Microsoft AI Webinars | The Fundamentals of Generative AI


 Microsoft always stands behind the IP in its services,
Overview: including how our models were trained.

Intellectual
 But with our Customer Copyright Commitment,
Property Microsoft will also defend against 3rd-party claims based
on the Output Content that our Copilots and Azure
OpenAI Service generate (with reasonable conditions).

 Our Copilot and Azure OpenAI Webinars will go deep


into this new commitment.

Microsoft AI Webinars | The Fundamentals of Generative AI


 Microsoft is committed to ensuring AI is built and used
Overview: responsibly and ethically, behind our six core principles:

Responsible AI  Fairness
 Reliability and Safety
 Privacy and Security
 Inclusiveness
 Transparency
 Accountability

 Much more in next month’s webinar!

Microsoft AI Webinars | The Fundamentals of Generative AI


Our AI  Sharing our learnings about developing and deploying
AI responsibly
Customer  Creating an AI Assurance Program
Commitments  Supporting our customers as they implement their own
AI systems responsibly

Microsoft AI Webinars | The Fundamentals of Generative AI


The EU AI Act  A new regulation for “AI systems” that has not yet gone
into force
 Classifies AI systems according to risk:
 “Unacceptable” → Prohibited
 “High risk” → Regulated
 “Limited risk” → Transparency
 “Minimal risk” → Unregulated

 After entry into force, regulations will apply between 6


months (prohibited AI systems) and 24-36 months (high-
risk AI systems)

Microsoft AI Webinars | The Fundamentals of Generative AI


How Legal Departments Are Using Generative AI

Knowledge Sharing Contracting Regulatory Compliance


Using Generative AI to access Generating contracts from Monitor and understand
knowledge from a wide variety of templates and adjusting terms regulations and public policy with
resources that already exist in the through prompts to Generative AI Generative AI to better
legal department models understand key global or regional
trends

Microsoft AI Webinars | The Fundamentals of Generative AI


Resources
 Microsoft’s “AI in a Minute” series
 Microsoft: Empowering responsible AI practices
 Microsoft's AI Customer Commitments
 The Customer Copyright Commitment
 Azure OpenAI Service: data, privacy, and security
 Copilot for Microsoft 365: data, privacy, and security

Microsoft AI Webinars | The Fundamentals of Generative AI


Q&A

Microsoft AI Webinars | The Fundamentals of Generative AI


Microsoft Legal AI Webinars: the series continues

Microsoft’s Responsible AI Journey: March 20-21

Unpacking Copilot for Microsoft 365: April 24-25

Exploring the Power of the Intelligent Cloud – Azure OpenAI: May 22-23

Register here! - ai-legalcounselwebinar.com

Microsoft AI Webinars | The Fundamentals of Generative AI

You might also like