Amazon Adds AI, SQL to Analytics Arsenal
By any measure, Amazon Web Services is already a monster in the field of analytics via pre-built Hadoop, Redshift, Kinesis, and deep learning services. With today’s unveiling of its new Amazon AI offering and a new ad-hoc query mechanism for data stored in S3 called Athena, the company is extending its considerable lead.
According to AWS CEO Andy Jassy, AWS has become a leader in machine learning and deep learning, with many companies like AON, Fraud.net, and Zillow running those workloads on its cloud infrastructure.
“A lot of companies don’t realize the heritage that Amazon has in the machine learning space,” he said during today’s keynote address at the AWS re:Invent conference in Las Vegas. “Part of that is just because we don’t talk about it very much. Tends to be our style. But if you actually think about it, AWS has a very deep heritage in this space. We have thousands of people dedicated solely to AI in our business.”
With the forthcoming launch of the new Amazon AI offering, the company expects to see many more customers availing themselves of pre-built artificial intelligence capabilities. In particular, AWS hopes the new offering attracts smaller firms that lack the deep technical expertise needed to set up and run deep learning setups.
The company today unveiled its Amazon AI brand, and announced the first three services that will be available underneath it, including Amazon Rekognition, Amazon Polly, and Amazon Lex.
Amazon Rekognition is a pre-built computer vision system that will automatically detect features in images. Customers will be able to upload images via an API or an SDK, and the Rekognition service will tell the customer various things about it, such as identifying simple objects, like people or cars. It will also count the number of people in images, determine the sentiment of the people, and perform facial recognition, which Jassy said could be useful for authentication.
“The service is really easy to use,” he said. “It will take million and millions of images in batch form, you can do it in real time. We’ll have the ability, because we have so much data and we have so much customers using this service, to continually fine-tune these models. And because they live in the cloud, you get to enjoy these improvements for free and really under the covers. And it’s very cost effective.”
Amazon Polly is a text-to-speech learning service build on deep learning that will take a stream of text and output an MP3 file. It will have intelligence to be able to convert things like abbreviations into their words. The response times will be fast, Jassy said, and it will come 47 voices in 27 languages.
Amazon Lex is a natural language understanding service designed to allow developers to build conversational interfaces. The service, which is based on the technology that powers the Alexa platform and all Echo devices, will be able to execute functions based on natural language inputs.
“It’s really interesting what we can do with the service,” Jassy said. “It gives you an integrated development console in AWS. You can set up whatever text or audio that you want to actually build conversational applications around.”
The service will feature connectors into Salesforce, Microsoft Dynamics, Marketo, Facebook Messenger, and Twilio. “We have a lot of data that we’re constantly going through that we continue to refine to make the models even better,” Jassy said. “Fully manged for you so you don’t have to do any heavy lifting.”
That wasn’t the only announcement out of Vegas today. Jassy said AWS had a number of customers who said they wanted a simple way to run ad-hoc analytics against things like logs, event files, and clickstream data. So the company developed Athena to be provide a relatively simple new way to run SQL queries against data stored in S3.
“Athena is an interactive query service that makes it easy to analyze data in S3 using standard SQL,” he said. “It’s really easy to use. You can query S3 directly. There’s no need to move the data …You just query directly at S3. There’s no infrastructure to spin up or mange. There’s no cluster to spin up or manage. The results and response times are really fast–seconds or millisecond and you pay only for the queries you run.”
The new offering, which is available now on the AWS website, won’t compete with existing analytic services, like its hosted Hadoop service, Elastic MapReduce (EMR), or its fast column-oriented MPP database, Redshift.
“The reality is that each of these services are different. Each of them serve a very useful purpose,” Jassy says. “There are always going to be a lot of customers who have a need to spin up clusters and mange them and tune them. These are customers who are using very large scale analytics in a very consistent fashion where the heightened performance is worth the effort in managing the cluster.”