Topics In Demand
Notification
New

No notification found.

The Seven Patterns of Commercial AI And AI-as-a-service
The Seven Patterns of Commercial AI And AI-as-a-service

580

0

 

 

Artificial Intelligence is undergoing rapid democratization. Democratization entails deeper and wider adoption, easier accessibility, and faster time-to-value. This is possible due to the rapid developments occurring in the Cloud Computing space – particularly in the areas of storage solutions and compute power (lower costs). Additionally, the global Cloud Infrastructure and Platform providers (CIPS) are offering pre-built Machine Learning solutions which are accessible via APIs and SDKs for easy integration.

 

NASSCOM’s new report ‘AIaaS – Democratizing AI at Scale’ attempts to define the evolving AIaaS market in India by bringing together different voices which make up the constituents of this market. AI-as-a-service is an euphemism for democratization of Artificial Intelligence. Wading through a host of focused interviews, we could identify the prominent threads that has led to the recent developments.

 

Today, the general discourse on Artificial Intelligence and its application hinges on the following:

 

  1. The different patterns of AI that can solve business problems, or bring about societal impact (as opposed to pure academic/research)
  2. And what are the forms of those and what are the underlying technologies?

Answering the questions above can give a better understanding of the landscape and the evolution of the solutions.

 

The Seven Patterns of AI:

 

 

AI Usage Patterns and AIaaS Convergence
Image Model for Commercial Artificial Intelligence 

 

 

  1. Hypersonalization and Recommendation Engines: There is an explosion of consumer-centric data across industries, particularly in Offline Retail, eCommerce, BFSI, Healthcare and OTT media. AI enables companies to leverage the data to create targeted and specific offers for customers with a view to gain better engagement, retain them, improve their buying experience across devices and platforms and create competitive advantage in the market. Solutions driving recommendation engines have matured over the past few years. Today, the combination of customer data, both demographic and behavioral, along with algorithms is among the mainstays of AI adoption globally.

 

  1. Recognition Systems: Recognition systems cut across both Vision (Image, video), and Language (Voice, Speech, handwriting). Algorithms enabling Recognition find wide usage in Manufacturing & Industrial (including Automotive), Autonomous Vehicles, Transportation and Logistics and Healthcare, particularly in monitoring systems and diagnosis. This involves the capability of machines to recognize, segregate and classify which imitates parts of the human cognition apparatus.

 

  1. Conversational AI and Machine - Human interaction – This is another mature area finding large-scale adoption across industries. This involves interaction between Humans and Machines through Conversational AI. Some of the key application areas of Conversational AI is in smart devices such as digital assistants - Alexa, Siri, Cortana and also includes Chatbots in contact centers, B2C businesses etc.

 

  1. Anomalies/Defect Detection – Ideally a subset use of Pattern recognition systems involving images and videos, it is finding traction in the Manufacturing/Industrial/Automotive industries. This leverages pattern recognition to flag items that are not following the norm, not adhering to quality specifications, or identifying potentially hazardous situations or equipment failure. This is most prevalent with the coming of age of IoT and connected devices, equipped with sensors (LIDAR and RADAR)

 

  1. Predictive Analytics – Predictive Analytics has been around for the longest time and has already moved to peak levels of productivity, basis its large-scale adoption. While it started as one of the application areas within Advanced Analytics which also included Descriptive Analytics and Prescriptive Analytics, it has evolved and is counted among the most widely used AI use-cases. This involves self-learning algorithms trained on historical data to predict future outcomes. As newer data flows in, the models are updated which improves the prediction capability. One of the key applications of Predictive Analytics is in Asset Maintenance, also known as Predictive Asset Maintenance. This is successfully used to predict the intervals mandating maintenance of equipment, machinery or heavy physical assets in Manufacturing, Power Transmission systems and the Oil & Gas sectors.

 

  1. Autonomous Systems – Autonomous Systems include Robotics, Drone Tech, Self-driving Cars (also known as autonomous vehicles). In simple terms, Autonomous Intelligent Systems are AI systems/software that have the ability to act independent of direct human supervision. The role of Artificial Intelligence in making autonomous systems work features among emerging roles of AI and is still some distance away from maturity and large-scale adoption.

 

  1. Goal-driven System: This is defined as systems that leverage cognitive thinking capabilities to arrive at optimization decisions. It involves learning through trial and error and to determine the best course of action from a pool of multiple options available. Unlike Supervised or Unsupervised learnings, a goal-driven system learns through discovery of clusters of information and other groupings through Reinforcement Learnings. There is generally no predefined or labelled output or goal-factor, and the system improves itself based on iterations. The application of goal-driven systems is nascent and expected to scale up in the future.

 

The adoption of AI across Industry segments revolves primarily around these 7 patterns.

 

How to define AI-as-a-service?

 

Below are two approaches to define AIaaS - one is a hypothesis, and another is based on the current market trends. One closely held idea is that any AI application/system that can be used to solve a specific business problem can take the form of as-a-service model. Or in other words, every use-case has the potential to become Artificial Intelligence-as-a-service.

 

  1. UI for Data as AIaaS: Traditionally, data has been used to build KPIs, Dashboards, Insights etc. but there is no User Interface for Data. Is it possible that a pre-built module for a use-case be defined as Artificial Intelligence as-a-service? According to this hypothesis, a purpose-built UI for data in an autonomous is AIaaS. For e.g., healthcare data involving patient monitoring activity can be used to identify and detect diseases and to take preventive action. A specific UI for this data can take the form of AIaaS.
  2. Application Programming Interface (API): APIs have emerged as a key contributor to the democratization of AI. This involves packaging the AI modules as APIs; when the API is called it fetches the data. There is growing convergence that APIs can act be termed as AIaaS. One key aspect is that modern AI systems are still within the realm of Narrow Artificial Intelligence. So, whether it is for Voice, Text or Image, it is capable of delivering the specific service based on predefined tasks. Providers of API enabled AI systems are bundling them with their broader portfolio offerings including Platform-as-a-service or storage or compute offerings. Going forward, API based services would be the backbone of AI as-a-service models.  

 

Two Inter-dependent Trends That Will Drive AI In Future

 

 

AI will pivot around Cloud Computing as well as broader adoption of 5G networks. The debate is expected to be around enterprise use-cases for AI Vs. Edge AI use-cases

 

  1. AI on the edge – The explosion of devices and sensor-derived data will act as a catalyst for AI. Sophisticated algorithms are in play here which are self-contained and are bounded by definitions in terms of what they can or cannot do. The service is closer to the device, where the action is. Zero-latency 5G networks would ensure that the services are delivered real-time. The emerging patterns need to be aggregated and conserved to be applied in the future into broader AI cycles to make corrections and improvisations and innovations.
  2. Summarized view on the cloud together with enterprise data, algorithms and patterns. This view is strategic in nature and broader.

 

 


That the contents of third-party articles/blogs published here on the website, and the interpretation of all information in the article/blogs such as data, maps, numbers, opinions etc. displayed in the article/blogs and views or the opinions expressed within the content are solely of the author's; and do not reflect the opinions and beliefs of NASSCOM or its affiliates in any manner. NASSCOM does not take any liability w.r.t. content in any manner and will not be liable in any manner whatsoever for any kind of liability arising out of any act, error or omission. The contents of third-party article/blogs published, are provided solely as convenience; and the presence of these articles/blogs should not, under any circumstances, be considered as an endorsement of the contents by NASSCOM in any manner; and if you chose to access these articles/blogs , you do so at your own risk.


images
Bandev Ghosh
Senior Manager

© Copyright nasscom. All Rights Reserved.