Skip to content

Latest commit

 

History

History
executable file
·
143 lines (99 loc) · 5.67 KB

from-innovation-to-deployment.md

File metadata and controls

executable file
·
143 lines (99 loc) · 5.67 KB
Raw
title subtitle abstract author blog date venue transition
From Innovation to Deployment
Auto AI and Machine Learning Systems Design
In this talk we introduce a five year project funded by the UK's Turing Institute to shift the focus from developing AI systems to deploying AI systems that are safe and reliable. The AI systems we are developing and deploying are based on interconnected machine learning components. There is a need for AI-assisted design and monitoring of these systems to ensure they perform robustly, safely and accurately in their deployed environment. We address the entire pipeline of AI system development, from data acquisition to decision making. Data Oriented Architectures are an ecosystem that includes system monitoring for performance, interpretability and fairness. The will enable us to move from individual component optimisation to full system monitoring and optimisation.
family given gscholar institute twitter url
Lawrence
Neil D.
r3SJcvoAAAAJ
University of Cambridge
lawrennd
2019-10-24
Data Science Africa, Ashesi University
None

\include{talk-macros.tex}

\subsection{Introduction}

\notes{Artificial Intelligence (AI) solutions are based on machine learning algorithms (ML), but each ML solution is only capable of solving a restricted task, e.g. a supervised learning problem. Consequently, any AI that we deploy today takes the form of an ML System with interacting components. As these ML systems become larger and more complex, challenges in interpretation, explanation, accuracy and fairness arise. This project addresses these issues. The challenges include [@Lawrence:threeds19]: the decomposition of the system, the data availability, and the performance of the system in deployment. Collectively we refer to these challenges as the "Three Ds of ML Systems Design".}

\include{_ai/includes/turing-ai-fellowship.md}

\subsection{Announcement} \slides{

\newslide{Announcement} \slides{

  • Five year program in collaboration with

. . .

\aligncenter{Element AI}

. . .

\aligncenter{Open ML}

. . .

\aligncenter{Professor Sylvie Delacroix}

. . .

\aligncenter{and}

. . .

\aligncenter{Data Science Africa!} } \notes{As of 24th October 2019, the Turing Institute announced that this work has been funded through a Turing Institute Senior AI Fellowship. This is the first Senior AI fellowship and it provides funding for five years.

The project partners are Element AI, Open ML, Professor Sylvie Delacroix and Data Science Africa.}

\include{_ai/includes/ride-allocation-prediction.md}

\include{_ai/includes/the-promise-of-ai.md}

\notes{This proposal is about addressing that gap, but to first understand the gap, let's look at comparisons between the approach we take to systems design, and the way that natural systems evolve.}

\include{_ai/includes/artificial-vs-natural-systems-short.md}

\notes{Currently, our main approach to systems design involves designing a system in a component-wise manner. Attempts to replicate the capabilities of evolved systems through specifying the objective, rather than evolving behaviour.}

\include{_ai/includes/ml-system-decomposability.md} \include{_ml/includes/ml-paradigm-shift.md}

\notes{This gives vulnerabilities that we are exposing to the natural environment. Many security problems that we face today are the result of bugs that mean that code and data are not separate in thee systems we deploy, imagine what will happen when we deploy systems that purposefully short-circuit this protection into uncontrolled environments.}

\include{_ai/includes/intelligent-system-paolo.md} \include{_ai/includes/peppercorn.md}

\subsection{The Three Ds of Machine Learning Systems Design}

\slides{

  • Three primary challenges of Machine Learning Systems Design.
  1. Decomposition
  2. Data
  3. Deployment } \notes{We can characterize the challenges for integrating machine learning within our systems as the three Ds. Decomposition, Data and Deployment.}

\addblog{The 3Ds of Machine Learning Systems Design}{2018/11/05/the-3ds-of-machine-learning-systems-design}

\notes{The first two components decomposition and data are interlinked, but we will first outline the decomposition challenge. Below we will mainly focus on supervised learning because this is arguably the technology that is best understood within machine learning.}

\newslide{The Three Ds of Machine Learning Systems Design}

\slides{

  • Three primary challenges of Machine Learning Systems Design.
  1. Decomposition
  2. Data
  3. Deployment } \notes{In this talk, we will focus on the third challenge, the deployment challenge.}

\include{_ml/includes/ml-deployment-challenge.md}

\include{_ai/includes/ml-system-decomposability.md} \include{_ai/includes/ride-allocation-prediction.md}

\include{_data-science/includes/data-oriented-architectures.md}

\include{_uq/includes/emulation.md} \include{_uq/includes/deep-emulation.md} \include{_uq/includes/bayesian-system-optimization.md} \include{_uq/includes/auto-ai.md}

\include{_data-science/includes/data-oriented-conclusions.md}

\thanks

\references