Amazon Web Services(AWS) is a cloud service from Amazon, that delivers services in the form of building blocks, these building blocks can further be utilised to construct and deploy any type of application with any size in the cloud. AWS can be broken into two main products EC2(Elastic Compute Cloud) i.e. Amazon's virtual machine service and S3(Simple Storage Service) which is basically Amazon's storage system. These services are outlined to work with each other and resultant applications are efficient, ultra-modern and highly scalable. It is the most cost-effective way to provide you with computational resources, stored data, and other applications.
Last year Amazon introduced its new Amazon AI platform at its re: Invent developer event. It aims to reach out with the concept of NLU, ASR, visual search and image recognition, text-to-speech (TTS), and ML technologies for developers. This new service avails many of the machine learning smarts Amazon has introduced in-house over the years to devs outside the company. It provides good quality, reliable AI capabilities that are highly efficient.
The concept of AI is comprised of three main layers- AI Services, Platforms and Frameworks that constitutes the AWS infrastructure:
At the highest level, AWS provides a collection of efficiently trained and tuned AI services for developers who want to work on AI technologies and does not intend to have any training or originate their own ML models.
Amazon Polly is a service provided that operates through advanced deep learning technologies which transform text into lifelike speech. Amazon Lex allows constructing conversational interfaces into the application through voice and text by providing advanced deep learning functionalities. Amazon Rekognition enables you to add image classification and image analysis to your applications through detecting objects, scenes, faces.
For users with pertaining details and contents intended to work on constructing custom inference models, the AI Platform services eliminate the undifferentiated overhead associated with establishing and handling AI training and model hosting.
The Amazon Machine Learning enables tools and documents that direct you to the process of creating ML models without grasping complex ML algorithms and technology. In addition, Apache Spark on Amazon EMR contains MLlib for efficient ML algorithms.
There have been a number of Frameworks provided by Microsoft, Google and now it's Amazon. As Amazon steps into providing us for Framework, it favours deep-learning framework to facilitate the work. Frameworks such as Apache MXNet, TensorFlow, Keras, Theano, Caffe, Torch, and CNTK extends adaptable programming models.
The AWS Deep Learning AMI is available for Amazon Linux and Ubuntu and allows you to create managed, auto-scaling clusters of GPUs. It is extended and supported by AWS, for use on Amazon EC2.
Deep learning frameworks, like Apache MXNet, use neural nets, which draws in the process of multiplying a lot of matrices. Amazon EC2 P2 instances present a powerful Nvidia GPUs to accelerate the time of computation. Once the training is completed Amazon EC2 C4 compute-optimized and M4 are appropriate for running inferences along with the trained model, also AWS Lambda allows us to untangle operations. AWS Greengrass provides you with a feature of running AI IoT applications on the AWS Cloud and local devices.
AWS are working to get ahead of other cloud competitors and along with the services provided it is expected for them to do really great but they need to work on adding more features so as to sustain the competition by Microsoft Azure and other service providers. But the services are expected to present us with a whole new generation of applications with human-like intelligence.
An Overview of AI on the AWS Platform
Video Source: Amazon Web Services