Cloud Amazon

Amazon AI: The smart person's guide

Amazon offers robust AI and machine learning capabilities with AWS. Find out how Amazon AI compares to similar offerings from Google and Microsoft.

Video: breaking down Amazon's 3 newest AI services for AWS

Amazon doesn't lead the pack on cloud hosting for no reason: Amazon Web Services (AWS) was the first platform of its kind on the market, has a wide range of services, and according to Garner it's number one in terms of having a strategic vision it can actually deliver on.

One of the features that may push Amazon further ahead of its competition is Amazon AI. Announced in November 2016 at re:Invent, Amazon AI is one of AWS's newer features: Artificial intelligence and machine learning software that can be completely customized from the ground up. That could give it a real edge.

TechRepublic's smart person's guide about Amazon AI is a quick introduction to this artificial intelligence platform, as well as a "living" resource that will be updated periodically as new breakthroughs in AI and machine learning change how it operates.

SEE: All of TechRepublic's smart person's guides

Image: Amazon

Executive summary

  • What is Amazon AI? Amazon AI is Amazon's machine learning and AI platform that is part of AWS. It has three ready-to-use services--Rekognition, Polly, and Lex--as well as AI platforms, AI frameworks, and AI infrastructure that can be used collectively or separately to create a custom AI or machine learning product.
  • Why does Amazon AI matter? Amazon's two biggest cloud competitors, Google and Microsoft, offer AI frameworks on their cloud platforms, and for some purposes they offer more. Google and Microsoft don't offer, however, a robust free tier that can serve as an entry point into AI and machine learning without an overburdening amount of limitations.
  • Who does Amazon AI affect? Amazon AI can affect anyone who needs AI and machine learning software. It especially affects customers who already subscribe to AWS and don't want to split their hosting between platforms.
  • When is Amazon AI happening? Amazon AI was announced in November 2016 at re:Invent and is now available for use on AWS.
  • How do I start using Amazon AI? If you have an AWS account using Amazon AI is simple: All that's necessary is deploying an Amazon Linux or Ubuntu instance using Amazon's AWS Deep Learning AMI.

SEE: Free ebook--How to implement AI and machine learning (PDF) (TechRepublic)

What is Amazon AI?

Amazon AI is AWS's deeply customizable machine learning and AI platform. Like other components of AWS, Amazon AI is designed to be scalable to match individual business needs and requirements.

Amazon has structured Amazon AI as three layers that appeal to different requirements and levels of AI and machine learning expertise.

Amazon AI's structure

Image: Amazon

At the top, most basic, level are three AI services. These APIs are pre-trained and pre-tuned so they can be used by those with little to no AI expertise. They are:

  • Amazon Rekognition, an image analysis and facial recognition service.
  • Amazon Polly, a text-to-speech service that Amazon says uses advanced deep learning to sound more like a real human voice.
  • Amazon Lex, which uses automatic speech recognition and natural language understanding to create chatbots and conversational interfaces.

One layer lower are the AI platforms, which are designed for "customers with existing data who want to focus on building custom inference models." These platforms are:

  • Amazon Machine Learning, which contains a series of visualization tools to analyze data without the need to learn complex machine learning code.
  • Amazon Elastic MapReduce (EMR), a managed Hadoop framework designed to perform log analysis, predictive analytics, real-time analytics, and other big data operations.
  • Spark, an Apache product that runs inside of Amazon EMR as a distributed processing system for big data workloads.

Below that are the AI frameworks, all of which run as part of the Amazon Deep Learning Amazon Machine Image (AMI). AMIs are pre-built, ready-to-deploy images for Amazon EC2, which in this case contain a huge array of libraries and frameworks.

The AI framework level is the most development-intensive of the three--it's a basic Linux environment with all the tools needed to do AI and machine learning work, but it's up to the owner to build what they need.

Additional resources

Why does Amazon AI matter?

AWS may be the senior cloud computing provider in the market, but it's the new kid on the block when it comes to AI and machine learning hosting.

Google has been around longer and has more APIs to offer, and Microsoft's cognitive computing offerings go further than what Amazon currently has available, Forrester AI analyst Mike Gualtieri told me.

Where Amazon stands out, Gualtieri said, is in pricing.

Take Amazon Rekognition, for example. It typically costs $1.00 per 1,000 images analyzed. Once you pass 1,000,000 images the price drops to $0.80 per 1,000 images. Compare that to Google's Cloud Vision API, which costs $1.50 per 1,000 images, and then drops to $1.00 after 5,000,000 images are processed. The first 1,000 images are free, with Vision API but that really won't save too much money once you've racked up several thousand images processed.

On top of that, Amazon includes Rekognition and its other services in the free tier of AWS, so those who just want to get their feet wet in machine learning can give it a try without having to worry about paying. The free tier of Rekognition offers 5,000 processed images per month.

The free offering from Amazon is a huge boon to those who aren't sure they want to invest a considerable amount of money in AI or machine learning. In that way, Amazon AI matters to those new to the AI game.

Check the appropriate service sites for specific pricing information.

Additional resources

Who does Amazon AI affect?

Amazon AI affects anyone who wants to get into machine learning without having to make a large investment for hosting or analysis. If those new to AWS manage to stay on the free tier there's a whole year of free API usage to get out of Amazon AI before having to pay a thing.

For those already paying for AWS, Amazon AI can serve to beef up big data processing capabilities without having to worry about splitting systems between different providers.

Additional resources

When is Amazon AI happening?

Amazon AI was announced at re:Invent in November 2016 and its features rolled out in early 2017. It is now fully integrated into AWS and available for use.

Additional resources

How do I start using Amazon AI?

Anyone with an AWS account who is willing to pay the additional fees (if they don't qualify for the free tier) to run Amazon AI components can start using it today. Amazon AI, like other AWS features, is charged on a per-use basis, so costs should be negligible for basic experimentation.

The three Amazon AI services can be used for free for a year, provided you don't perform more than a certain number of requests per month. After that the services are still relatively cheap: Polly, for instance, can process up to 1,000 1,000 character requests a month for only $4.00.

Other components, like the platforms and frameworks, are charged based on hours of use and prices vary per piece.

If you're interested in using any of Amazon AI's features, you'll need to get started using AWS, and then head over to the Amazon AI page so you can pick from the services offered.

Additional resources

Visit TechRepublic