Data infrastructure optimization software
Data integration and quality software
Data availability and security software
Cloud solutions

Machine Learning as a Service: Is It Really Here?

It looks like the days of paying a one-time fee to purchase and install software and applications is coming to an end. The new IT business model is to sell software and even IT infrastructure As A Service. So far, we’ve got SaaS (Software as a Service), IaaS (Infrastructure as a Service), and PaaS (Platform as a Service), and the all-in-one solution, SPI (Software PLUS Platform PLUS Infrastructure services). Or, you can use the catch-all term, XaaS.

Introducing MLaaS, or Machine Learning as a Service. While a number of upstarts are entering the MLaaS realm (or attempting to), including BigML, Wise.io, Precog, and Ersatz, the usual suspects also have hats tossed in the ring, including Google, Microsoft, and now Amazon.

Amazon’s Andy Jassy, Senior Vice President of Amazon Web Services, announced their new offering at the Amazon Summit in San Francisco on April 9.

How Amazon’s MLaaS Works

data transfer
MLaaS is a way for businesses that can’t hire their own data scientists to remain competitive in the age of data analytics.

Amazon’s MLaaS is being offered as part of Amazon’s overall EC2 cloud, which has expanded by an eyebrow-raising 516 new features in just over a year. Their Machine Learning model is a homegrown service that they developed to use in-house, and are now offering to other businesses, of course, at a fee.

You are familiar with this already. It’s the predictive analytics used for Amazon’s Recommendations, in other words, products Amazon suggests for you when you place items in your Cart. Like when you shop for a rake and it also recommends a hoe and shovel and perhaps a lawn tractor.

In light of the current crunch to find qualified data scientists, it’s a good bet plenty of businesses will leverage such a service. The platform is designed to be used by people with little or no real data analytics experience. Currently, Amazon’s Recommendations are managing some 5 billion predictions per week, and if you’ve ever saved your kid’s birthday surprise because Amazon reminded you that the remote control car you ordered indeed also needed a battery, you’re probably already a fan.

The MLaaS by Amazon will feature a machine Learning API and developers’ guide in the form of wizards. The current pricing structure sits at 42-cents per hour for data analytics, model training, and model evaluation, 10-cents per 1,000 predictions in batch mode, and 10-cents per 1,000 predictions in real time. These fees are on top of the hourly reserved capacity charges of $0.001 per hour per 10 MB, which will likely be the primary moneymaker for Amazon’s model.

Others Offering (or Soon to Offer) MLaaS

data transfer
Nobody is surprised to see Microsoft and Google in line with MLaaS offerings, and it’s a natural extension of Amazon’s Web Service products.

Aside from the aforementioned newcomers attempting to break into MLaaS (BigML, Wise.io, etc), the two major players currently are Google’s Prediction API, and PSI Project, which is a Microsoft brainchild. It will be interesting to see whether marketing genius Amazon or owns-everything Google will emerge on top of the MLaaS race. For now, the winners are businesses with the need for predictive analysis to remain competitive but the inability to do their own data transformation operations or acquire their own data scientists.

0 comments

Leave a Comment

Related Posts