Skip to content

Latest commit

 

History

History
executable file
·
539 lines (398 loc) · 9.85 KB

contents.md

File metadata and controls

executable file
·
539 lines (398 loc) · 9.85 KB
Raw

The Institute for Ethical AI & ML

Meditations on First Deployment




Alejandro Saucedo | [email protected]

@AxSaucedo
in/axsaucedo

[NEXT]

The Institute for Ethical AI & ML

Meditations on First Deployment


![portrait](images/alejandro.jpg)
Alejandro Saucedo 

    <br>
    Chief Scientist
    <br>
    <a style="color: cyan" href="http://e-x.io">The Institute for Ethical AI & ML</a
    <br>
    <br>
    <br>
    <hr>
    <br>
    Head of Solutions Eng. & Sci.
    <br>
    <a style="color: cyan" href="http://eigentech.com">Eigen Technologies</a>
    <br>
    <br>
    Chief Technology Officer
    <br>
    <a style="color: cyan" href="#">Hack Partners</a>
    <br>
    <br>
    Software Engineer
    <br>
    <a style="color: cyan" href="#">Bloomberg LP.</a>

</td>

[NEXT]

Today

The Institute

AI Procurement Framework

Next stesp

Questions

[NEXT]

#LetsDoThis

[NEXT SECTION]

1. Overview

[NEXT]

About the institute

We are a UK-based research centre formed by cross functional teams of applied STEM researchers, philosophers, industry experts and software engineers

[NEXT]

Applied research

We develop industry frameworks and practical research that empowers technologists to design, develop and deploy of machine learning systems responsibly.

[NEXT]

The new complexity

full_height

[NEXT]

It gets harder

full_height

[NEXT]

The core Principles

full_width

http://ethical.institute/

[NEXT]

Let's put these into practice

[NEXT SECTION]

2. The Framework

[NEXT]

AI Procurement Framework


  • A set of tempaltes to support industry practitioners and suppliers in AI tenders.

  • Fully open source, built using our "Machine Learning Maturity Model".

[NEXT]

An Assessment Criteria

The Machine Learning Maturity Model

#1 Practical benchmarks
#2 Explainability by justification
#3 Infrastructure for reproducible operations
#4 Data and model assessment processes
#5 Privacy enforcing infrastructure
#6 Operational process design
#7 Change management capabilities
#8 Security risk processes

[NEXT]

Creating a checklist

From principles to a practical questionnaire

#1 Practical benchmarks
#2 Explainability by justification
#3 Infrastructure for reproducible operations
#4 Data and model assessment processes
#5 Privacy enforcing infrastructure
#6 Operational process design
#7 Change management capabilities
#8 Security risk processes

  • Each has a set of questions for supplier compliance

  • Each question points out potential Red Flags
<style> .fragment.visible.fade-out.current-fragment { display: none !important; height:0px; line-height: 0px; font-size: 0px; } </style>

[NEXT]

Practical benchmarks


Ensuring there are process to evaluate the right metrics:

  • Accuracy
  • Cost functions
  • Time
  • Time-to-accuracy

line

[NEXT]

Explainability by justification


Explainability through domain knowledge, together with feature importance analysis

line

[NEXT]

Data and model assessment processes

line

[NEXT]

Infrastructure for reproducible operations

line

[NEXT]

Privacy enforcing infrastructure

Build processes to use and protect user data & privacy, and make sure they are communicated

line

[NEXT]

Operational process design


Assess impact of incorrect predictions

and design with human-in-the-loop review

where reasonable

line

[NEXT]

Change management capabilities

Identifying and documenting impact of technology towards workers being displaced

line

[NEXT]

Security risk processes

Develop processes and infrastructure to ensure data and model security are taken into consideration

[NEXT]

Next steps

Applying this thinking into your actual projects

[NEXT SECTION]

3. Wraping up

[NEXT]

Today

The Institute

AI Procurement Framework

Next stesp

Questions

[NEXT]

The ML Principles

ethical.institute/principles.html


## Procurement Framework ethical.institute/rfx.html

[NEXT]

The Institute for Ethical AI & ML

Meditations on First Deployment




Alejandro Saucedo | [email protected]

@AxSaucedo
in/axsaucedo