There’s been a lot of attention on Machine Learning (ML) lately, and for very good reason.
Machine Learning has been one of the most fascinating advancements of our time, and it continues to improve and innovate every single day. Using ML to solve problems or strip away routine tasks has benefited many industries, especially in the last several years.
By simply starting small, many businesses have now evolved to leverage and develop robust ML models capable of analysing more complex data sets while delivering more accurate results faster than ever.
AWS has been leading the way when it comes to advancing ML products, making ML more usable and accessible to everyone. In fact, ML is becoming more mainstream, with over 100,000 customers now using ML on AWS.
Earlier this month, AWS hosted its much-anticipated virtual summit with some of the brightest minds in ML, exploring how ML impacts businesses and showcasing real-world examples.
Here are some of our key takeaways and favourite moments of the 2021 AWS Machine Learning Summit.
Opening keynote with Swami Sivasubramanian
As the VP of AI and ML at AWS, Sivasubramanian kicked off the summit with an insightful chat about how increasingly embraced and transformative ML has been, used to improve and personalise customer experience, create efficiencies, and spurring new discoveries.
Sivasubramanian noted that from the way we do business, to the way we seek entertainment, to how we get things done – ML is transforming everything. Entire business processes around the world are being made easier with ML, and the barriers of entry to ML have been significantly lowered to enable builders to build and apply ML to their most critical challenges.
Examples included how BMW Group uses AWS SageMaker to globally forecast demand of model makes and individual equipment, and iFood Brazil uses AWS SageMaker to optimise delivery routes to decrease the distance travelled by riders by 12%.
Another, more interesting, example was how Swedish frozen meals company Dafgårds used ML to ensure a consistent quality of their meals. Dafgårds achieved this by using “few-shot learning”, where their visual quality inspection only needs a few image examples to identify inconsistencies. We’re talking about avoiding “not enough cheese” on your pizza.
Ashok Srivastava, Senior Vice President & Chief Data Officer at Intuit was also an intriguing keynote speaker, offering an in-depth real-world scenario of how they successfully applied ML.
Srivastava discussed the Artificial Intelligence (AI) hierarchy of needs, and how at Intuit the goal is for their AI scientists to focus on the (productive) top of the pyramid, while the unnecessary workloads at the (unproductive) bottom of the pyramid – where up to 70% of a person’s time could be wasted – become eliminated thanks to greater infrastructure capabilities.
Intuit treated their data “as a product” and built critical data infrastructure to help their team get access to clean data as rapidly as possible. By modernising with AWS, Intuit achieved a 30% decrease in downtime, a 60% increase in mobile app deployments, and tripled their speed of delivery.
Fireside chat with Andrew Ng
In this fireside chat with host Swami Sivasubramanian, the brilliant Andrew Ng, who is the founder of DeepLearning.AI, co-founder Coursera, and leader of many other AI-focused organisations discussed getting started with AI.
Ng noted that over the last few decades, as paper records across various industries globally was converted into digital records, the foundation for many industries to be “AI-ready” was unknowingly laid. Still, Ng believes that even though it has created tremendous value, the most exciting activities are still to come.
The most common comment by CEOs and CTOs encountered by Ng as to why they can’t get started on ML projects is that “my data is a mess and still needs to be cleaned up”. Ng assures that almost every company has some level of messy data, which means that most companies today have enough to simply get started, deliver a small project or proof of concept, achieve a quick win, and then formulate a thoughtful strategy to do bigger and better ML projects.
Ng believes that until a business has done their first few ML projects, strategies and goals are often more theoretical.
Ng also advises on the importance of getting executive-level team members non-technical AI knowledge, as it would make it easier for them to collaborate with technical team members down the chain, delivering more consistent success.
How AstraZeneca is transforming pharmaceutical R&D through modern ML
Dr. Ian Dix, Senior Director AI/ML and Analytics R&D at AstraZeneca discussed how their journey towards building a flexible, compliant platform that helped accelerate drug research.
The average cost in the industry for producing a drug is an estimated $2 billion and 10 years of R&D, which is not sustainable in the long run. Data and AI are going to be critical for transforming R&D.
Only one in 10 drugs in early clinical trial stages make in through as a licensed product, and much of that failure is attributed to the inability to demonstrate efficacy.
Dix explained that the vision they have at AstraZeneca is three-fold, focused on accelerating, enabling, and transforming R&D activities through AI and data:
- A lot of AstraZeneca’s data activities involve routine tasks such as data processing, document sorting, and image processing – and so automating these activities is a must to ensure they can be accelerated.
- By provisioning the data scientists with high quality, accessible, interoperable and reusable data, they are enabled to improve their way they work.
- By accelerating routine tasks and supporting data scientists with the tools and resources they need, the organisation would be in a position to transform their R&D in previously impossible ways; building advanced AI solutions and bringing teams together to improve target selection, drug design, clinical trial design, and drug responses.
“When we started a lot of this work, teams often outsourced activities such as some of their data science or data management work, but they didn’t necessarily outsource this to all the same organisations, so we ended up with a lot of data silos,” explained Dix.
AstraZeneca now looks to repatriate these activities to internalise their data mesh, build an end-to-end understanding of their data science activities, and maintain data compliance.
How 3M cultivates a company-wide ML culture
David Frazee, Vice President at 3M began with an overview of their unique periodic-table-inspired tech capabilities diagram, which Frazee explains is at the heart of 3M today, and used to guide and define the creation of new capabilities.
3M utilises AWS Machine Learning technologies to modernise mundane activities, such as record management and billing cycles, to benefit and improve their entire operations.
For a few years now, 3M hosts their widely popular hackathon events. Over time, the 3M team realised that this became part of their culture and reframed their flywheel model. From cloud tech enabling them to scale, to scrum allowing them to achieve with speed, to having the right people passionate about what they do, to instilling a “hacking” culture of leveraging modern tech to create something meaningful for customers.
With so many other bright minds leading insightful stories and real-world examples of Machine Learning with AWS, make sure you catch up to it all by signing up for on-demand content from the AWS Machine Learning Summit.
Want to learn more about Machine Learning and ways it can revolutionise your business? Get in touch with us to find out more.