Ever wanted a machine-learning platform capable of making predictions based on your in-house data? Even if the thought never actually crossed your mind, you can’t deny that the concept has a certain appeal: Who wouldn’t want a system capable of offering up solid advice, based on everything your business has done?
Amazon, which has evidently never met some aspect of storage and analytics it didn’t want to try and sell for cheap to businesses all over the world, has just announced a Machine Learning service, complete with visualization tools.
The cloud-based platform will walk developers through the creation of machine learning (ML) models; it also includes “simple” APIs that allow apps to call the eventual predictions. Amazon’s overarching claim, of course, is that it can carry out all these functions without the development team needing to learn complex algorithms, implement custom code, or deal with any of the infrastructure concerns usually associated with data storage and heavy computation.
As with Amazon’s other cloud services, the Machine Learning’s pricing is based on use: $0.42 an hour for data analysis and modeling, $0.10 per 1,000 batch predictions, and $0.0001 per real-time prediction.
In theory, businesses can use machine-learning products for everything from fraud detection and customer-churn prediction to modeling for customer support and marketing campaigns. In reality, though, it remains to be seen how many businesses feel the need (and think they have the ability) to use tools this powerful, even if the cloud does eliminate much of their underlying complexity.