Wednesday, April 21, 2021

The Top Cloud-Based AI Services

Businesses that want to use AI run into a major roadblock: it’s too expensive to develop AI products in-house. Hence the demand for outsourced artificial intelligence products: It’s much cheaper for anyone from a SMB to a budget-constrained large enterprise to use the cloud to get its feet wet with AI.

The major cloud computing providers now offer cloud-based AI products. They leverage their vast technical expertise and deep pockets to provide this next-gen service.

AI services are both a hardware and software play, but the real heavy lifting lies in the hardware. It’s been the exponential advances in processor technology that has enabled the AI revolution, both in custom silicon and the adoption of GPUs normally used for graphics and gaming as AI processors.

As the AI product market expands, all the major cloud providers have rolled out some level of AI services. And clearly AI as a service remains firmly in the domain of the top players due to the expense of building such a system from scratch. So we will take a deep dive on the big players, and a few whose profile is not quite as large.

IBM Cloud: The Most Comprehensive AI Package

Late last year, IBM merged its BlueMix cloud service, SoftLayer data centers, and Watson AI group into one, called IBM Cloud, which offers more than 170 services total. Under the Watson brand for AI services, IBM has no less than 16 services. Much of the emphasis is on analytics, from data to speech and text. And IBM has its Global Services consulting business, which only Microsoft can even remotely match.

The IBM Cloud AI services start with Watson Studio for building and training AI models, preparing data and performing analysis on the data. This is available in one integrated environment. For existing data, there is Watson Knowledge Catalog to do intelligent data and analytic asset discovery, cataloging and governance, and Watson Discovery to find connections and relations.

IBM has made a point of noting that only 20% of the world’s data is searchable, and there is heavy emphasis in IBM Cloud Watson on data processing and discovery. An example of this is IBM Watson Services for Core ML, which allows enterprises to build AI-powered apps that securely connect to their data and run either on-premises, offline or in cloud. These apps utilize machine learning to adapt and improve through each user interaction.

Other data discovery apps include Data Refinery, a self-service data preparation tool for data scientists, engineers and business analysts and Deep Learning, which helps developers design and deploy deep learning models using neural networks, easily scale to hundreds of training runs.

To build AI platforms, IBM has Watson Assistant to build and deploy chat bots and virtual assistants, Watson IoT Platform to provide a cloud-hosted service for device registration, connectivity, control, rapid visualization and data storage.

IBM is also big on language recognition and translation. Watson Speech to Text (STT) converts audio and voice into written text, while Watson Text to Speech (TTS) does the opposite and converts written text into natural-sounding audio in a variety of languages and voices.

The Watson Language Translator translates news, patents or conversational documents, Watson Natural Language Classifier interprets and classifies natural language, Watson Natural Language Understanding analyzes text to extract metadata from content such as concepts, entities and sentiment.

On the more esoteric side, Watson Visual Recognition can tag, classify, and search visual content using machine learning, Watson Tone Analyzer analyzes emotions and tones in written content and Watson Personality Insights predicts personality characteristics, needs and values through written text.

Amazon Web Services: Consumer AI Repositioned for Business

Amazon’s AI efforts are split into two categories: improving its consumer devices such as Alexa, and services on AWS. Much of those business cloud services are in fact built on the consumer products, so as Alexa improves, so will its business equivalent. They break down into four major categories, many of which have parallels in consumer interaction:

A service for building conversational interfaces into any application, Amazon Lex uses voice and text. It has automatic speech recognition for converting speech to text and natural language understanding to recognize the intent of the text. Lex technology is now used in Alexa and allows developers to create natural language-enabled chatbots.

If you want to do the opposite, Amazon Polly turns text into lifelike speech. Many artificial speech applications have a stiff delivery, where you can hear the breaks in the words that were stitched together. Polly uses advanced deep learning technologies to synthesize speech that sounds like a human voice. It delivers fast response times required to support real-time, interactive dialog.

Amazon Rekognition: Amazon Rekognition makes it easy to add image analysis to your applications to detect objects, scenes, and faces in images or to search and compare faces. Amazon uses this service to analyze billions of images daily for Prime Photos. The Rekognition API lets you easily build visual search and discovery into your applications.

Machine learning is on the vanguard of today’ AI activities but can require in-house expertise. In contrast, Amazon Machine Learning provides visualization tools and wizards that guide you through the process of creating machine learning models without having to learn complex ML algorithms and technology. It is built on the same technology Amazon uses to help suggest items to shoppers.

Microsoft Azure: Emphasis on Developers

Microsoft breaks down its AI offerings into three categories: AI Services, AI Tools and Frameworks, and AI Infrastructure. Like Amazon, some of its business AI offerings are actually built on consumer products.

AI Services breaks down into three subgroups: pre-built AI capabilities such as Azure Cognitive Services, which adds intelligence to customer-facing apps like Web apps and chat bots, Cognitive Search, which merges Azure Search with Cognitive Services, Conversational AI with Azure Bot Service, and custom AI development with Azure Machine Learning (AML). Microsoft recently updated its Bot Frameworkto create the next generation of conversational bots with richer dialogs, full personality and voice customization for developers.

AI Tools & Frameworks include Visual Studio tools for AI, Azure Notebooks, Data Science VMs, Azure Machine Learning Studio and the AI Toolkit for Azure IoT Edge. Microsoft recently announced it is opening up the Azure Internet of Things Edge Runtime, which will allow developers to modify and customize applications at the edge. With multiple projections of 20+ billion connected IoT devices by 2020, this will be important. The Azure IoT Edge Runtime also serves as a platform from which all of Azure’s new AI-driven applications will be built upon.

AI Infrastructure includes Azure Data Services, compute services including Azure Kubernetes Services (AKS) and AI Silicon support including GPUs and FPGAs. Azure Data Services are the databases available on Azure, like SQL Server, MySQL, PostgreSQL, NoSQL, and MariaDB. The Kubernetes services involve the popular container service, which is used to modernize and cloud-enable existing on-premises apps. Silicon support is simple acceleration beyond CPUs for high performance apps.

Google Cloud: Accelerated by Special AI Processors

A key differentiator for Google is the TPU, or Tensor Processing Unit. This is a special chip designed specifically for use with TensorFlow, Google’s open-source machine learning platform that all of the major cloud providers offer, but none have the TPU to speed things up. And speed it up it does. TPUs are between 15x and 30x faster than CPUs or GPUs, offering up to 180 teraflops of compute power.

Like Amazon and Microsoft, Google has taken the AI from its consumer-facing products and made it available for business users. The AI power behind Google applications like Images, Translate, Inbox (Smart Reply), and voice search in Android is available to its Google Compute Engine cloud offering.

Much of Google’s AI offerings are reflections of its core search competency. For example, the Cloud Vision API can identify objects, logos, and landmarks within images, specific or explicit content, text within an image, can find similar images on the Web, or detect faces and read expressions. Strangely, it does not provide facial recognition.

Similarly, the Cloud Video Intelligence API allows you to search videos for content, such as images or text. For example, it can search images for specific content and block the video on that basis.

DialogFlow is for building chatbots that handle customer messaging, voice recognition and response. It can build interfaces between mobile apps, messaging services and IoT devices. The Natural Language API offers even deeper insight with syntax, entity, and sentiment recognition and classifying content.

Google also has a Speech-to-Text API that enables speech conversion in real time or from recordings in 120 languages, while the Text-to-Speech API produces natural-sounding audio from text. The Cloud Translation API provides translation services for over 100 languages and can work in conjunction with the above APIs.

Google has specifically focused on machine learning, to analyze data for better decision making, while at the same time offers flexibility and accessibility for inexperienced AI developers through its Cloud ML service. Developers can train high-quality machine learning models, such as customer service, using Google’s existing APIs.

For more experienced ML developers, Google offers Machine Learning (ML) Engine for bringing machine learning models to production, using TensorFlow models that need to be trained for various scenarios. It has a prediction service that takes trained models and uses them to make predictions on new data.

Other Cloud AI Providers: Specialty Offerings

Oracle AI: Oracle’s main pitch here is its support for data sources for mining and extracting data. You have support for its database, MySQL, and Big Data clusters like Hadoop. It comes with popular machine learning tools and frameworks to rapidly build applications.

Salesforce: The company’s Einstein AI platform is fully integrated with other Salesforce cloud offerings to build apps using machine learning and predictive analytics and utilize your Salesforce data. It is used to build apps such as chat bots and sales prediction.

Baidu: China’s clone of Google has also cloned Google’s AI efforts, right down to making its own custom AI processor. It has a mobile service, called Baidu Brain, and a conversational AI OS, DuerOS. But this is only available in China.

Vendor Distinguishing features of its offering
IBM Watson Largest and most diverse set of AI services; established and mature program in Watson; IBM Global Services is unmatched in consulting.
Amazon Web Services Rapidly maturing set of services based around its popular consumer products; runs on AWS, the top cloud service provider
Microsoft Azure Built on Microsoft legacy software; Microsoft’s strong history supporting developers
Google Cloud AI Top performance thanks to custom chips to accelerate AI and ML; specializing in machine learning.
Oracle AI Built on Oracle’s legacy line of business apps, so you can add AI to your Oracle environment.
Salesforce Easy and rapid development of apps; adds AI to Salesforce’s comprehensive CRM offerings.
Baidu Mirrors Google in terms of features and specialized accelerator chips.

Similar articles

Latest Articles

IT Planning During a...

Without a doubt, 2020 changed everything. I like to compare it to a science fiction movie where time travel is involved. Clearly, we have...

Best Data Quality Tools...

Data quality is a critical issue in today’s data centers. The complexity of the Cloud continues to grow, leading to an increasing need for...

NVIDIA’s New Grace ARM/GPU...

This week is NVIDIA’s GTC, or GPU Technology Conference, and they likely should have changed the name to ATC because this year – it...

What is Data Segmentation?

Definition of Data Segmentation Data segmentation is the process of grouping your data into at least two subsets, although more separations may be necessary on...