Overview of AWS AI offerings

To better understand AWS AI offerings, we can group the services into two main groups.

The following diagram shows the subsets of AWS AI capabilities and AWS ML platforms that we will be covering in this book, organized by the two groups:

The first group in the preceding diagram are the AWS AI capabilities. These services are built on top of AWS pre-trained AI technologies. They work right out of the box to provide ready-made intelligence for your applications. You do not need to understand the AI techniques that make them tick, and you do not need to maintain the infrastructure to host them. AWS has done all the hard work for you, and has made these AI capabilities available via APIs. As AWS continuously improve these capabilities, your application will automatically get more intelligent without any effort on your part. These managed services can provide quick lifts to your applications, thus allowing intelligent solutions to be built quickly and economically.

These AWS AI capabilities are as follows:

• Amazon Comprehend: A NLP service that uses ML to find insights and relationships in text. This technology allows your applications to sift through mountains of unstructured text and dig up golden nuggets of information. This service can perform various tasks, including automatic classification of documents; identification of entities such as company names, people, and addresses; and extraction of topics, key phrases, and sentiments within the text.
• Amazon Lex: A service for building conversational interfaces into applications using voice or text. This capability is built using deep learning techniques such as automatic speech recognition (ASR) and natural language understanding (NLU) in order to convert speech into text and to recognize intents within text. This is the same technology that is behind the Amazon Alexa voice assistant, and the same capability can be embedded into your own applications.
• Amazon Polly: A service that turns text into life-like speech that allows you to add a human voice to your application. The text-to-speech technology that is behind this service uses advanced deep learning technologies that can synthesize a voice with different languages, genders, and accents.
• Amazon Rekognition: A service that can analyze images and videos in order to identify objects, people, text, scenes, and activities. This service can also provide accurate facial analysis and recognition for various applications. The deep learning technology behind this service has been trained on billions of images and videos for a high level of accuracy on a variety of analysis tasks.
• Amazon Transcribe: An ASR service that provides speech-to-text capabilities to your applications. This technology allows your applications to analyze stored audio files or live audio streams, and get transcription text in real time.
• AWS Translate: A neural machine translation service that delivers natural and fluent language translation. This service is backed by deep learning models that can provide accurate and natural sounding translations for many languages. You can even configure this service with custom language models that can include brand names, product names, and other custom terms.

The second group in the preceding diagram are the AWS ML platforms. These services are fully managed infrastructures and toolsets that help developers build and run their custom AI capabilities via ML. AWS provides the development constructs and handles the ML training compute resources in order to make developing custom AI capabilities easier. The AI practitioners are responsible for designing the inner workings of these AI capabilities. This might include: the collection and cleaning of training data; selection of ML libraries and algorithms; tuning and optimization of ML models; and designing and the development of interfaces to access the AI capabilities. Leveraging the AWS ML platforms to build custom AI capabilities is definitely more involved than using the managed AI services, but this group of services gives you the most flexibility to create innovative solutions.

The AWS ML platform we will be covering in this book is: Amazon SageMaker—A fully managed service that covers the entire ML workflow. With SageMaker, you can collect and process your training data; you can choose your ML algorithms and ML libraries, including TensorFlow, PyTorch, MXNet, Scikit-learn, and so on; you can train the ML models on ML-optimized compute resources; and you can tune and deploy the resulting models to provide AI capabilities that are specifically created for your applications.