eWEEK content and product recommendations are editorially independent. We may make money when you click on links to our partners. Learn More.
I spoke with Bratin Saha, VP and General Manager, Machine Learning Services, AWS, about how the cloud has enabled machine learning, and SageMaker’s ability to ease ML deployment. (See podcast and video below.)
Among his key points:
Exponential Growth in ML Models
“Now to give you some statistics, let’s say in 2018, just about when SageMaker started, our customers used to deploy maybe a dozen models. Today, our customers deploy millions of models. And at that point of time, the state of the art models that they would train would maybe have 20 million parameters or so – now they have more than 100 billion parameters.”
ML Use Case
“Domino’s Pizza had a project called Project 310, where they wanted to have a pizza ready for pickup within three minutes of an order, and have it delivered within 10 minutes of an order. Well, they were able to do it by using machine learning to predict when a pizza would be ordered.”
Help with Data Prep
“In SageMaker, Data Wrangler provides you a no-code visual way of doing all of the data preparation. So you can, with a single click, connect to multiple data sources – it could be S3, it could be Redshift, it could be Snowflake. So you can use any of these data repositories and then pull in the data. And again, with a few clicks, you can choose the kind of data transformations you want to do.”
Explainability in ML
“We have a feature called SageMaker Clarify. What Clarify does is it gives you tools to measure bias in your data sets. There are a number of different measures that you can use. It’s a very visual, no-code way of doing it; you can basically choose which measures of bias you want to use.
And it also explains the predictions that a model is making. So let’s say a model is predicting customer churn. You can use it to say, Why do you think that the customer is going to churn? So it converts the [data] into actionable insights.”
“I think the next phase of machine learning will deliver what scientists dreamed of when they came up with machine learning, which is truly industrializing, and democratizing machine learning.
Future of Machine Learning: Democratization
“I think the next phase of machine learning will deliver what scientists dreamed of when they came up with machine learning, which is truly industrializing, and democratizing machine learning.
And I think we are at the cusp of a major shift, where industries across the board are going to make machine learning integral to how they operate and how they solve their customers problems, just like an Amazon has been innovating with machine learning and transforming the customer experience.
I do think machine learning is going to get accessible to a lot more people. Currently, it’s accessible to ML developers, data scientists, and so on. But I think we’ll come up with tools where we make it much easier for a wider set of personas to do it.”
Download the podcast:
Watch the video:
- James Maguire on Twitter: https://twitter.com/JamesMaguire
- eWEEK on Twitter: https://twitter.com/eWEEKNews
- eWEEK on Facebook: https://www.facebook.com/eWeekNews/
- eWEEK on LinkedIn: https://www.linkedin.com/company/ewee…