MeitY
    false

    AI-first for live enterprise

    Highlights

    The report is published by Infosys Knowledge Institute. The Institute is known for assisting industry leaders gain a deeper understanding of business and technology trends through thought leadership, subject matter expert and in-depth research. The report exclusively explores the trends surrounding the subdomains which has the potential to upgrade an enterprise from H1 to H3. Under the umbrella of AI algorithms and architectures, the trends that have emerged are improving generalization and accuracy with deep neural network architectures and transition from System 1 deep learning to System 2 deep learning. The report also emphasizes that building an AI model requires a tremendous amount of effort, investment and subsequent challenges.

    Summary

    Introduction

    The report is published by Infosys Knowledge Institute. The Institute is known for assisting industry leaders gain a deeper understanding of business and technology trends through thought leadership, subject matter expert and in-depth research.


    The report starts on a positive note by stating that the ongoing COVID-19 pandemic has urgently necessitated the process of drug discovery and trial processes, which can be made been possible through AI. Having said that, thereby giving hope, the report then quickly points out that any enterprise undergoing AI transformation are moving across the following 3 horizons: a) Horizon 1 (H1) characterized by enhancing fragmented intelligence into existing systems with superior capabilities, b) H2 which are much more complex systems requiring higher-order generalizations, accuracy and learning capabilities such as neural machine translations and c) H3, also known as digital brain, a system that delivers intelligence with distributed learning through well managed and governed AI systems. However, moving to H3 would essentially require enterprises to work on AI subdomains such as computer vision, AI algorithms and architecture, natural language processing (NLP), speech and AI life cycle tools to name a few.


    The report exclusively explores the trends surrounding the subdomains which has the potential to upgrade an enterprise from H1 to H3. Under the umbrella of AI algorithms and architectures, the trends that have emerged are improving generalization and accuracy with deep neural network architectures and transition from System 1 deep learning to System 2 deep learning. The trends driving computer vision are image segmentation, classification and attribute extraction, video insights such as generating video captions, video highlights, content moderation, surveillance and people/object tracking. As far as speech is concerned, the report believes that adoption of neural machine translation and transcription-based systems will mine conversational insights apart from speech biometrics which has gained importance during the pandemic. The subdomain NLP is undergoing huge innovations to improve machine’s ability to understand, generate, process and derive insights. While NLP (H1) was primarily meant for extracting and representing information as a set of words, H2 has the ability to understand custom-named entities (such as currency symbols). Apart from these, there has been a trend towards point-specific contextual learning with edge-based intelligence such as voice recognition and assistance, facial biometrics or autonomous vehicle navigation system. Also, in the recent past, there has been a trend towards integrating AI lifecycle tools to drive enterprise-wide standardization as a typical AI life cycle involves various stages from data collection and analysis to tuning, testing, monitoring and feedback.


    The report also emphasizes that building an AI model requires a tremendous amount of effort, investment and subsequent challenges. Hence, it makes sense to devise a way to reuse the effort invested by sharing the created models and ensuring compatibility and portability across environments. However, amidst instances of growing AI biasedness, the report reiterates the fact that the basic principle that should govern AI should be ethics.


    Relevance of the Report


    AI has become an intrinsic part of our daily lives and as technology improves, consumers are increasingly relying upon AI to perform their day-to-day tasks. Having said that, enterprise adoption of AI has spearheaded and is rapidly evolving to manage multiple tasks such as planning and forecasting, predictive maintenance and customer service to cater to this growing demand. As a result of which there has been a clear shift towards simplifying the science of AI as the transformative journey gradually evolves from an augmented intelligence to a generative and explainable system. The report comes in handy for those enterprises who are willing to transform and manage the AI systems based on the current trends.


    Key Takeaways


    • The AI dynamics is based on 3 horizons: H1, the conventional AI, H2, Deep learning involving higher accuracy and predictability and H3, generative AI that are explainable and interpretable.
    • The report clearly identifies the key trends across AI subdomains and is as diverse from improving accuracy and AI architectures to speech biometrics, model sharing and adherence to AI ethics. 
    • To make the report an interesting read, there has been special mention of Infosys partnered/backed technologies that have implemented these trends to fix a given problem or improve efficiency.

    Report Contributors

    Infosys

    AI-first for live enterprise

    By Infosys
    Read Research Report

    Previous Research Summary

    The global AI agenda: Asia-Pacific

    Next Research Summary

    With AI’s great power comes great responsibility

    Suggested Articles