The Machine Learning Imperative: Empowering Businesses to Innovate Faster

The Machine Learning Imperative: Empowering Businesses to Innovate Faster

Sally Eaves 15/09/2021
The Machine Learning Imperative: Empowering Businesses to Innovate Faster

Machine Learning (ML) took centre stage at AWS reInvent from being a key focus in the opening keynote by CEO Andy Jassy, to the entire focus of the inaugural ML keynote by Swami Sivasubramanian, VP of Amazon Machine Learning – very fitting given AWS being such an early adopter of ML at the core of their business.

In combination, this resulted in an inspiring plethora of transformational announcements covering Instances; Data, Search and Analytics; Amazon SageMaker; Development and Application Performance; Devices; Amazon Connect; Industrial/Manufacturing/Commercial; Conversation and Text. These aim to benefit all user types and with tools tailorable to different environments - from novices to field experts, from SMBs to large scale enterprises, and across a diversity of industry sectors.

Heralding Machine Learning as ‘one of the most disruptive technologies we will ever encounter in our generation’ and emphasising its application for increasingly core business functions with AWS ML moving further up the stack, there were three core tenets or pillars underpinning Swami Sivasubramanian’s keynote:

Provide_Firm_Foundations .jpeg

Provide Firm Foundations 

Optimise the foundations - frameworks and infrastructure - upon which models are built and used, speeding up training and deployment, and reducing cost.

Create the Shortest Path to Success - Having the tools for builders to be able to satisfy and explore their ideas quickly, without barriers, is a significant accelerator to business success.

Expand Machine Learning to more Builders – Bringing ML to builders such as data analysts and data developers with tools closer to those they natively use, integrating ML into data stores and BI tools - alongside bringing higher abstraction level ML services to more people.

I believe this trajectory is absolutely vital to increase accessibility to learn and apply ML wherever you are, and wherever you want to go. The capacity to innovate necessitates experimentation (including failure), cultural support and the right tools and services, tailored to different roles, as Swami Sivasubramanian describes, this is the freedom to invent:

‘Freedom to invent requires that builders of all skill levels can reap the benefits of revolutionary technology. The technology itself allows for experimentation, failures, and limitless possibilities’

Additionally, the criticality of being able to solve real business problems end-to-end and to embrace continual learning came to the fore throughout. Providing context to all the announcements and focus areas, the image below shows the AWS AI/ML ecosystem. At the bottom of the stack is ML capability for expert practitioners, with Amazon SageMaker in the middle layer of the stack allowing developers and scientists to build train and deploy ML models at scale and at the top layer, a suite of AI services. There are 250+ new features this year alone - from generic ML infrastructure components to business-centric, high abstraction, ML enabled PaaS and SaaS services.

ML_enabled_PaaS_and_SaaS_services.png

Swami Sivasubramanian presents the AWS AI/ML Ecosystem

And reflecting across all the content, there is so much news to consider! From new Anomaly Detection Services (Lookout, DevOps Guru) to the embedding of Responsible Artificial Intelligence (SageMaker Clarify). And right across AI Services (HealthLake, Monitron, Lookout, Panorama) to ML Services with SageMaker (Pipelines, Clarify, DataWrangler, DistributedTraining, FeatureStore) and much more besides. Taking this all into consideration, here are my personal reflections on some of the leading developments.

Detecting Bias by Design - SageMaker Clarify

Detecting_Bias_by_Design_-_SageMaker_Clarify.jpeg

A difficult call to make! But my No 1 Machine Learning announcement from reInvent is the introduction of Clarify to Data Science tool Amazon SageMaker enabling bias detection in data and training models, with the bias computation aspect being open sourced. A key step change to increase the transparency of ML Models and in advancing ethical AI development – and within a ML platform already used by millions of users around the world.

This transparency is key to building trust (Edelman 2020) and can benefit multiple stakeholders, for example End Users having a better understanding on why specific results are generated and Developers being able to more easily debug, tune and optimise ML models. With Accuracy vs Explainability trade-offs a key driver for model selections, especially in highly regulated environments, SageMaker now also supports Machine Learning Interpretability and with the bias and explain-ability tools fully available visually in all stages of the model life cycle. This all aligns with the ethos of Shapley Values to afford visibility and transparency into the decision making of a model and can help shorten the path to success. 

Faster_Distributed_Training.png

Faster Distributed Training

Supporting the tenet of providing firm foundations upon which models are built and used is my next highlight, the announcement of Faster Distributed Training on Amazon Sagemaker, as part of a critical AWS focus on Machine Learning infrastructure scalability and performance. Utilising the forthcoming EC2 instances built for ML training by design and powered by Habana Gaudi processors, this development aims to complete distributed training up to 40% faster.

Other related aspects here are 1) Managed Data Parallelism which simplifies training on large datasets, negating trade-offs between training time and training cost and so enabling Machine Learning teams to benefit from accelerated results-iteration-innovation cycles. And 2) Managed Model Parallelism to automatically and efficiently partition models across several smaller and more cost-effective GPUs. This enables the capacity to work with very large models without memory bottlenecks, whilst eliminating the need for accuracy compromises or for complex manual work. Perfect to add to PyTorch and TensorFlow training scripts!

An Up-Shift in Access to ML 

An_Up-Shift_in_Access_to_ML .png

Another key development is the launch of Amazon Redshift ML which aligns with the tenet to expand Machine Learning to more builders by enabling easier experimentation and deployment of models. This tool will allow developers to create, train and apply ML algorithms in Amazon Redshift data warehouses without the need for manual selection, building, or training of models, but instead use more familiar SQL commands via SageMaker Autopilot. In addition, the Autopilot tool allows for easy integration with more data sources natively, including Tableau, Qlik and Snowflake.

Enabling Intelligent Industry

And my final selection is expanding the application of Machine Learning for Industry use with the introduction of 5 new industry and manufacturing focussed tools to help organisations identify productivity bottlenecks, safety violations and potential equipment failures. This is ever more important given the impact of COVID19 with shocks across supply chains, growth strategies and pricing structures. Bringing Cloud to Edge and 24x7 machine intelligence – integrating AI, ML, sensor analysis, and computer vision – into production processes is vital. This can drive a move beyond a reactive to more proactive stance across operational efficiency, quality control, security and workplace safety. A step change for optimising manufacturing and industrial operations, and creating a robust foundation for digital industrial transformation. 

An_Up-Shift_in_Access_to_ML .jpeg

Final Thoughts

Machine Learning is a leading catalyst to enhance, transform and disrupt across multiple industry verticals, with use cases from prediction and forecasting, to analytics and forensics. To optimise that potential, it is imperative to empower businesses and the individuals within to innovate with ML by having the right tools available at the right time and tailorable to different roles and environments. I believe the examples highlighted and overarching Machine Learning focus of reInvent is a critical step forward in this trajectory, underpinned by tangible tool advancements and investment in education which come together to Provide Firm FoundationsCreate the Shortest Path to Success and Expand Machine Learning to more Builders.

Share this article

Leave your comments

Post comment as a guest

0
terms and condition.
  • No comments found

Share this article

Sally Eaves

Tech Expert

Dr. Sally Eaves is a highly experienced Chief Technology Officer, Professor in Advanced Technologies and a Global Strategic Advisor on Digital Transformation specialising in the application of emergent technologies, notably AI, FinTech, Blockchain & 5G disciplines, for business transformation and social impact at scale. An international Keynote Speaker and Author, Sally was an inaugural recipient of the Frontier Technology and Social Impact award, presented at the United Nations in 2018 and has been described as the ‘torchbearer for ethical tech’ founding Aspirational Futures to enhance inclusion, diversity and belonging in the technology space and beyond.

   

Latest Articles

View all
  • Science
  • Technology
  • Companies
  • Environment
  • Global Economy
  • Finance
  • Politics
  • Society