Almost everyone agrees that artificial intelligence (AI) is the next big thing for businesses. But, is there any AI platform ready for you to deploy? While IBM’s Watson is the most widely known AI platform, there are many more popular artificial intelligence tools for you to hire.
Amazon Web Services (AWS) offers a good range of AI services including tools, platforms, frameworks and infrastructure. Amazon Machine Learning is a cloud-based service that makes machine learning technology available to users of all skill levels. Using powerful algorithms, it helps users create machine learning models by finding patterns in existing data, and using these patterns to make predictions from new data as it becomes available. Offering a user-friendly interface utilising visualisation tools, wizards and more, the highly scalable platform enables creation, evaluation and deployment of meaningful machine learning (ML) models. AWS also provides AI services like image recognition, text-to-speech, voice and text chatbots, and more.
This service is widely used in applications like demand forecast, user activity prediction, item recommendation, fraud detection and suspicious transaction flagging, review filtering and so on. Amazon Machine Learning’s secret of ease lies in its automated processing in every step.
Experts suggest that this platform is well-suited for users who are looking for quick and powerful yet low-cost solutions. It is also a recommendable platform for relatively newer users. It can be used for retail, food chain, hospitality, sports and entertainment, and finance among other sectors.
Let’s look at the example of Upserve—a US-based company that delivers cloud-based intelligence and management system for restaurants. Amazon Machine Learning, ingesting data like reservations, real-time payment processing, menu preference histories, customer count and so on, developed over a hundred machine learning models, which helped Upserve’s client restaurants to improve profitability and drastically cut down on wastage based on major predictive parameters.
Google Cloud Machine Learning Engine
Google Cloud ML platform has made a mark very quickly despite being a much newer release than Amazon or Azure ML platforms. This platform attends to developers’ desire of forging their own machine learning model—no matter how complex or how large. This ability comes from the TensorFlow system—Google’s open source deep learning framework—lying at the heart of Google’s machine learning platform.
The massively scalable machine learning engine backed by AI APIs for speech recognition, advanced search, natural language processing, video, image and text analysis, and much more, gives users a machine learning model with benefits like accurate real-time analytics, efficient anomaly detection and correction, prediction, easy search and recommendations, meaningful insight into complex unstructured data, automated processes and so on.
Customers of any stature ranging from search engines, online retail, job search sites, sales or marketing up to bioengineering or even aerospace and defence firms will find Google Cloud ML platform suitable for their requirement. The biggest confidence in using Google’s offerings comes from the fact that the global giant itself uses the same technologies in its own systems as offered to its customers.
In a real-life scenario, Google’s ML platform has benefited the likes of Airbus Defense and Space in detection and correction of satellite images. Even at high quality level, satellite images capture phenomena like cloud formation, which hinder the real motive of the picture like accurate atmosphere or seasonal analysis—something which is used in areas like precision farming, yield prediction or crop health analysis, where data is required to be absolutely precise. At Airbus, the process had been manual, time-consuming and error-prone for over a decade. Google’s solution automated it and delivered to Airbus customers results with absolute accuracy and detailed insight. Thus, Google’s platform is capable of delivering solutions for any scale of complexity.
USA-based IPSoft created an intriguing AI innovation, called Amelia. The interesting aspect of Amelia is its interaction capabilities coupled with conversation and advanced learning talents that push it closer to the true sense of AI. Geared with advanced learning and self-learning tools, machine learning and deep learning capabilities, understanding of entire sentences and conversations with knowledge of 20 languages, and facial expression and emotion detection, the system can extract knowledge from large documents and historical records, interact directly with its employer’s customers, educate human colleagues, oversee and streamline workflow in a management role, and do much more. The applications of Amelia put a major focus towards digital workforce management.
Chetan Dube, founder and CEO, IPSoft, says, “Amelia is neither a platform, nor a solution. It is a digital employee who has been designed with intelligence, reasoning and decision-making skills that help her more closely resemble the way humans think, act, and work. Amelia is a scalable digital workforce that can address as much work volume as needed. She has a direct impact on the key ROI metrics that shareholders and management teams prioritise above all others. Amelia’s ability to learn more quickly, manage more complex dialogues, respond to analytical triggers in real time, and better understand those with whom she interacts with truly sets her apart as a valuable asset for any modern enterprise.”
AI has the capability to take up any data-driven work, some examples being data mining, analysis, prediction, real-time actions and process automation, enabling decision-makers to efficiently manage their workforce by shifting repetitive or accuracy-major tasks to the platform. Currently, Amelia has been taken up by over 50 organisations globally in sectors as wide-ranging as financial services, telecommunications, leisure, travel, healthcare, IT, BPO and government. It is substantially involved in the roles of customer service, finance and banking, insurance, IT infrastructure management, human resources, service desks, or even technical guide or advisor for field-workers or engineers.
Let’s take the case of an online gaming major which needed a sophisticated agent to connect with gamers for support and block phishing attempts or ever-evolving fraud techniques, for which the company needed to quickly adapt in order to maximise customer service and workflow efficiency. Amelia cannot be socially engineered—a feature the company leveraged to interact with human customers and verify accounts. Amelia is helping the gaming company to verify user identities at 100 per cent accuracy, cut fraud-screening time from five minutes down to three minutes and receive an 86.72 per cent customer satisfaction rating—higher than with human operators.
Microsoft’s machine learning platform was introduced for public preview in July 2014, and since then it has picked up good steam. The specialty of the Azure Machine Learning platform is the interactive user interface brought by the integrated Azure Machine Learning Studio, which allows users to set up their machine learning model through simple drag-and-drop actions. Inbuilt algorithm makes building supervised and unsupervised learning models easy. However, the modelling process requires active intervention of users in all development steps. Support for R and Python programming languages is a bonus for users who plan to create or customise their own ML models.
Additionally, Azure ML supports a wide range of statistical functions and methods which are core to Data Science. Supported ML models include regression, binary and multi-class classification, clustering, recommendations and anomaly detection.
All user categories ranging from SMEs up to large industries in different verticals including retail, insurance, finance and banking, healthcare, oil and gas or even manufacturing can benefit from this AI platform. Smarter and swift predictive analytics, project outcome management, re-usable and scalable models, efficient data pipelines, risk and fraud management, and patient recovery and treatment prediction are some of the biggest benefits of Azure Machine Learning.
The Microsoft Azure website showcases a story where Mendeley, a free research content reference and academic network platform for researchers, students and academic readers, implemented Azure Machine Learning to manage its users better, provide more accurate recommendations, and improve interface response and overall user experience. Feeding historical data like user activity trends, views, libraries, searches and data provided by the marketing team, helped the company to build ML models, which eventually improved the company’s recall by 30 per cent.
IBM Watson is an interactive cognitive platform whose analytical basis is the humongous amount of disparate big data structures that the system can engulf. It computes this data at very high speed and analyses to give the results in minimal time.
Talking about its cognitive and interactive capability, Dr Prashant Pradhan, executive director, Watson & Cloud, and chief developer advocate, IBM India/South Asia, says, “Watson understands the nuances of human language, so it is able to bring back relevant answers in context of the question. Watson also gets smarter, learning from each interaction with its users, and each piece of data it ingests.”
Driven on the IBM Bluemix cloud platform, IBM Watson comes with 50 AI APIs including IBM dialogue, retrieve and rank, machine learning, speech-to-text, text-to-speech and, more interestingly, concept and visual Insight that enable an application to expand and relate concepts, drawing on the meaning of a word rather than simple text matching, and allow developers to build apps that reveal insights from social media images and video. As an analytics platform, Watson provides a visual representation of the business drivers and insights into any focus areas. While the major involvement of the user is in loading the correct data, rest of the background computations are mostly automated within the system.
Watson has seen diverse uses in sectors like healthcare (to suggest treatments for cancer), financial services (to recommend personalised portfolios using the latest information), customer engagement, law, retail, fashion and education.
In an interesting application from India, Sriram Raghavan, director, India Research Labs, IBM India/South Asia, explains how Watson is being used in diverse industries like fashion as well. Designers at Falguni Shane Peacock took up IBM Watson to design their line called ‘Future of Bollywood Fashion.’ The system analysed data from 600,000 publicly available fashion runway images of the last decade, 5000 Bollywood celebrity images from social networks and 3000 historical Bollywood fashion images. With the help of visual recognition APIs and tools from the IBM Research Cognitive Fashion project, the system was able to assist significantly in the design process.
If you want to share such case studies with EFY readers, please contact us at [email protected]