Practices For Responsible AI At Microsoft

Question 1

At Microsoft we have established our own governing practices for responsible and I will stop this includes former governing body is developing our policies and providing engineering teams with actionable guidance and tools. Finally we aim to empower people and organisations to use AI to improve lives around the world

Question 2

Understanding the Microsoft governance model - aether + office of responsible AI

Question 3

Microsoft responsible governance structure:
Our governance structure today uses a hub-and-spoke model to provide the accountability and authority to drive initiatives while also enabling responsible AI policies To Be implemented at scale

Question 4

Centralised governance:
The senior leadership team is ultimately accountable for the company’s Direction on responsible as setting the companies are principles values and human rights committee’s. Building off of our culture and integrity and trust this group is the final decision maker on the most sensitive novel and significant AI development and deployment matters.

Question 5

The office of responsible tires for key functions:
Internal policy in new line enablement
Case management
Public policy

Question 6

Internal policy: Setting the company-wide rules for an acting responsible as well as defining roles and responsibilities for teens involved in this effort

Question 7

Enablement:
Readiness to adopt responsibility practices both within our company and among our customers and Partners

Question 8

Case management:
Review of sensitive use cases to help ensure that our AI principles are upheld in our development and deployment work

Question 9

Public policy:
Help to shape new laws norms and standards that will be needed to ensure that the promise of AI technology is realised for the benefit of society at large

Question 10

The Aether committee serves and advisory role to the senior leadership and the office of responsible guy on questions challenges and opportunities with the development and Fielding of AI technologies.
If they’re also provides guidance to teams across the company to ensure that our products and services align with our principles.
The committee brings together the top Talent and technology is ethics law and policy from across Microsoft formulate recommendations on policies processes and best practices

Question 11

The Aether committee has six working groups that focus on specific topics grounded in our our principles.
The working group’s play a key role in developing tools best practices and tailored implementation guidance through the related to their respective areas of expertise

Question 12

Decentralized governance

Question 13

Policies and procedures turn on new line to help every Microsoft employee live up to our commitment to developing and deploying responsible as we have created principles a sensitive use framework and company-wide rules to help employees develop a better understanding of the company’s commitment with respect to AI development and deployment

Question 14

We expect every Microsoft employee to:
Develop a general understanding of our our principles
Report and escalate sensitive uses
Contact their responsible and champlain they need guidance on responsible

Question 15

Responsible as standard:
In order to implement responsible AI practices the policy requirements procedures and Tools need to be tightly embedded with the AI development lifestyle and organisation already uses.

Question 16

Responsible are frameworks in action:
As discussed in the previous unit Microsoft has been developing and refining its own internal process to govern responsibly

Question 17

Microsoft sensitive use case framework:
For responsible governance documentation and development or employment scenario was considered a sensitive us if it falls into one or more of the following categories:

Answer 17

Denial of consequential services
Risk of harm
Infringement on human rights

Question 18

Niall of consequential services:
The scenario involves the use of AI in a way that made directly result in the denial of consequential services or support to an individual for example financial housing insurance education employment health care services etc

Question 19

Risk of harm:

Question 20

Infringement on human rights

Question 21

Microsoft sensitive use case review process:
Identification
Assessment
Mitigation

Question 22

Identification

Question 23

Assessment

Question 24

Mitigation

Question 25

Establishing design principles and guidelines in engineering:
If you are developing implementing or managing a internally you may want to consider how to honour your organization’s guiding principles at every step of the AI life cycle.
To empower technical employees to do so we have found that it helps to translate your principles into an actionable guidance

Question 26

Principles and guidelines

Question 27

Security and privacy:
Security and privacy of key pillars of trustful stop there are a number of emerging tools to help protect security and privacy in AI systems will stop techniques like homomorphic encryption multi-party computation and differential privacy make it possible to train a r models using private data without sharing it.

Answer 27

Once the AI system is built you can use technologies like counterfeit to conduct a security risk assessments

Question 28

Resource type:
Guidelines
Technology tools
Third-party tools

Question 29

Guidelines

Question 30

Technology tools

Question 31

Third-party tools

Question 32

Fairness:
Ai system to treat everyone fairly and avoid affecting similarly-situated groups of people in different ways
Fairness is a fundamentally social socio-technical challenge Sophia classification tools are not be all and end all solutions for stop however there are two key steps for reducing unfairness:
assessment and mitigation

Question 33

Implementing fairness in your organisation:
Resourcetype:
guidelines
Technology tools
Third-party tools

Question 34

Guidelines:

Question 35

Technology tools

Question 36

Third-party tools

Question 37

Inclusiveness:
Inclusive design practices help ensure that AI models perform well for all users

Answer 37

Resource type:
Guidelines:
Reference Microsoft inclusive design practices inclusive design toolkit and algorithmic green lining paper.

Question 38

Reliability and safety

Question 39

Transparency

Question 40

Accountability

Question 41

Implementing accountability in your organisation:
Guide line
Management tools
Technology tools

Question 42

Guideline

Question 43

Management tools

Question 44

Technology tools

Question 45

Engaging externally-ai for good

Question 46

Contributing solutions to societal challenges

Question 47

AI for accessibility

Question 48

AI for Earth

Question 49

AI for humanitarian action

Question 50

AI for cultural heritage

Question 51

AI for health