The Capabilities in Microsoft Azure's AI/ML System: A Mini Primer
Updated: Sep 21, 2023
Ankesh
Editorial Intern for The Indian Learning (e-ISSN: 2582-5631)
Indian Society of Artificial Intelligence & Law
Introduction
Microsoft is one of the technological leaders in the world, which has been involved in almost all sectors of the Information and Technology market. Azure is one of the AI tools which is available to use on a licence basis like other Microsoft product so that users can take advantage of the software according to their needs. Azure is basically a tool mainly for data analysis and decision making. In Microsoft’s own words it is “a portfolio of AI services designed for developers and data scientists. Take advantage of the decades of breakthrough research, responsible AI practices, and flexibilit
y that Azure AI offers to build and deploy your own AI solutions.”
One of the fundamentals while doing market research is to know about what is going to be the target audience for the market, or which type of customers is the research actually aimed at. Microsoft’s Azure is mainly aimed at scientists, and tech enthusiasts, who are in a necessity of using the data collected by them in an organised manner. The platform in addition to this has integrated software such as Jupyter Notebook, and Visual Studio to take full advantage of AI and Machine learning capabilities.
Microsoft describes Azure Cognitive Services as “a comprehensive family of AI services and cognitive APIs to help you build intelligent apps,” and claims to have the “most comprehensive portfolio of domain-specific AI capabilities on the market,” although its competitors might disagree with that assessment. Azure Cognitive Services are aimed at developers who want to incorporate machine learning into their applications.
What is Azure?
The Azure services are a group of services integrated together, to get the best possible usage to any sort of data available for processing on the network. In addition to this these Capabilities are also used in applications such as Visual Studio Code for better optimization. You can finish hours of codes in minutes through these templates and Artificial Intelligence suggestion which are available to the users. Not just this, but these capabilities are slowly been seen in Microsoft’s productivity suite, like Word, PowerPoint, but also in Microsoft’s in-house image and video editing platforms.
Instead of having companies run all Azure services on the Azure platform, Microsoft also offers several Docker containers that allow companies to use local AI platforms. It supports a subset of Azure Cognitive Services and allows companies to run Cognitive Services on firewalls with local data sensitive to companies with strict data security policies and those subject to protection-restricted data. In addition to this, as the debate of responsible AI increases, Microsoft has also, released some Open-Source AI projects from its dockets like FairLearn, InterpretML, and SmartNoise.
These open-source projects have proven quite handy and useful for users and Microsoft too. According to InfoWorld, these open-source codes and APIs can also be integrated with some other released software from Microsoft, in turn increasing their productivity. “Fairlearn contains mitigation algorithms as well as a Jupyter widget for model assessment, and has been integrated into a Fairness panel in Azure Machine Learning.” With regards to the InterpretML, it helps you “understand your model’s global behaviour, or understand the reasons behind individual predictions, and has been integrated into an Explanation dashboard in Azure Machine Learning.” The SmartNoise project, in collaboration with OpenDP, aims to expose discriminatory privacy for future use by providing some basic building blocks for people with sensitive information to use. You can import SmartNoise into your Python notebook by installing and importing your project and adding calls to display sensitive data. Many frameworks and tools are used in the world of machine learning, deep learning and artificial intelligence. Azure AI directly supports dozens of these, but there are hundreds more managed by Azure by providing or allowing integrations. Some, like MLflow, are integrated into Python packages. Others like Pachyderm are often integrated as containers in Kubernetes (AKS).
This set of Services are together called the Azure Cognitive Services. As mentioned earlier that Microsoft defines it as a group of services presents which use Artificial Intelligence and Machine Learning capabilities, to develop better applications for customers, The set of services have mainly four areas of expertise.
Decision Making – Helps you choose or decide what should be the next step.
Language – This is the basic AI are that helps all sorts of language and translate that language into another, it may include any sort of language.
The third one is speaking.
And the final one is vision.
Web Search or the interaction of computers has been given a separate section in Azure Cognitive Services.
These services are ready to use, and trains itself form the plethora of data it gets fed every day, the basic working principle of AI. Azure Cognitive Services Learning is usually not required, at least not as high as you would expect from Azure Machine Learning. Some Azure Cognitive Services are configured, but you don't need to understand machine learning to do so. Almost all Azure Cognitive Services have a free plan to choose from.
The Decision Support of Azure Services
It includes four layers of services to reach the final stage. They are namely an anomaly detector service, a content moderator, a metrics advisor, and a personalized. The Anomaly Detector service integrates anomaly detection into your application, allowing users to quickly identify issues when they arise. No previous experience with machine learning is required. Through the API, Anomaly Detector receives time series data of any type and selects the best anomaly detection model for the data to ensure high accuracy. You can configure parameters to customize the service according to your business risk profile. You can run Anomaly Detector anywhere, from the cloud to smart peripherals.
The content moderator service is designed to manage social media, product review websites, and games with user-generated content and performs image moderation, text moderation, video moderation, and everyone's validation to lower or soften trust prediction contexts. The Metrics Advisor service is based on the Anomaly Detector service and allows you to track your organization's growth mechanisms from sales revenue to manufacturing in near real-time. It also tailors the model to your scenario, provides detailed analysis with diagnostics, and alerts you to anomalies.
Personalizer is an artificial intelligence service that provides a personalized experience for each user. There is a "learning" mode that uses booster learning to optimize the model for a goal and allows the Personalizer to interact with the user only after the Personalizer has reached a certain level of confidence that matches the performance of the existing solution.
Language Platform in Azure
The language area of Azure Cognitive Services includes an immersive reader, a language understanding service, a Q&A platform, text analytics, and a language translator. These services are then integrated to give the best results possible. The reader is basically software converting texts and images in the most readable way possible for a user. “The Azure Language Understanding service, also call LUIS (the “I” stands for intelligent), allows you to define intents and entities and map them to words and phrases, then use the language model in your own applications.”
The Question-and-Answer forum basically lets user create a layer of questions on the text data present like FAQs, etc. Text Analytics is an artificial intelligence service that integrates ideas such as sentiments, units, relationships and key phrases into unstructured text. It can be used to identify key phrases and entities such as people, places, and organizations to understand common themes and trends. A related service, Text Analytics for Health (currently in preview), lets you classify medical terms using pre-trained domain-specific models. Sentiment analysis and text evaluation in multiple languages can help you better understand customer opinions. This also includes a translator feature which is also available as a standalone application for many windows running devices.
The Speech and Vision Platform of Azure
In addition to these, the vision and speech services are also provided as mentioned earlier. The Speech area of Azure Cognitive Services includes speech recognition, text to speech, speech translation, and speaker recognition. Speech recognition works in two use cases. For identification, it matches the voice of an enrolled speaker from within a group, which is useful in transcribing conversations. For verification, it can either use passphrases or free-form voice input to verify individuals for secure customer engagements. The text to speech or vice versa can be used to enter information without the use of physically touching the devices. “Microsoft describes its Speech to Text service as allowing you to quickly and accurately transcribe audio to text in more than 85 languages and variants.” The voice translation service allows you to translate sounds from more than 30 languages and tailor the translation to your organization's specific context.
The vision platform for Microsoft’s Azure is probably one of the most interesting services out there. The future holds quite great for these services as they would be on the frontend services if Virtual Reality has a future to hold here. They work on layers namely computer vision, custom vision, face detection, form recognition, and video indexing. These features basically include analysing the images and videos for extracting data from them to use for the services. Custom Vision uses transfer learning to generate a customized image model from a few tagged images rather than the thousands of images needed. An unaltered image can also help. As more images are added, the model continues to improve. “The Face service includes face detection that perceives faces and attributes in an image.” Also, the Video indexing service allows you to automatically extract metadata such as voice, text, face, speakers, celebrities, emotions, themes, brands and scenes from video and audio files. You can then access data from your application or framework, or make it more discoverable.
Azure Machine Learning and Other Services
In addition to the major services provided by Azure, Microsoft also uses services such as Bot services and Databricks. These services are the various sets that make azure such a comprehensive product to try. The azure bot service is used to Managed service for creating and using chat agents for various application areas and an open-source SDK derived from Cortana development to build Q&A bots, virtual assistants, and more. They also use Access natural language capabilities in the Azure cloud and deploy services across multiple communication channels and messengers.
Databricks are part of MS’s cloud data warehouse ecosystem. These Databricks on Apache Spark allow setup, preparation and training of large amounts of data and they are an essential element for using near-/real-time data or streaming high-scale IoT data.
The Azure machine learning is one of the similar services like the Azure Cognitive services but it uses more end-to-end learning between the computer devices. The features of both overlaps, but there are some clear distinctions too. The Azure Machine Learning is the core data science cloud service to build, train and deploy machine learning models. It also It provides an easy-to-use visual interface for merging open source or shelf models and transforming data by dragging and dropping into data pipelines. As with personal coding in various programming languages such as R or Python, detailed model configuration and customization is also possible.
In addition to all of this some small services such as Web Search service which is currently under Microsoft’s Bing, and services like Data Science Virtual Machines is also there. They have cloud solutions for creating workstations for data processing and analysis. Also, there is no need for a fully integrated data storage environment.
Conclusion
The Azure services are one of the best front-end services with respect to artificial intelligence and machine learning, currently in the market. It is easy to access and is deemed to be customer friendly. The use of these services is more into the simple normal tasks, which a scientist or a data analyst or a coder might face. This although has a wide range of services so more companies can opt for the product. The cutting edge work is not suitable for Microsoft Azure, as it can handle them, but not really in an efficient manner.
Comments