Jayprasad Hegde, Head of Data Science & AI Initiatives at the National Payments Corporation of India (NPCI), offers a profound exploration of Generative AI from an AI scientist’s perspective at MLDS. His talk, “Gen AI from the lens of AI Scientists,” delves into the intricate world behind the revolutionary advancements in AI, highlighting the transformative power and the challenges that lie ahead.
The AI domain is witnessing an unparalleled expansion, with Generative AI technologies captivating the imagination of both the public and specialists alike. However, beyond the user-centric view lies a complex and rapidly advancing technological front. From ChatGPT challenging the Turing test to Microsoft’s use of tens of thousands of A100 GPUs, the pace of innovation is staggering. Topics such as Langchain, Retrieval-Augmented Generation (RAG), anchoring, Generative Adversarial Networks (GANs), Large Language Models (LLMs), Vision Language Models (VLMs), Diffusion, and Reinforcement Learning from Human Feedback (RLHF) are gaining prominence daily. Even traditional AI themes like “Reasoning” are resurfacing with renewed importance, demonstrating the depth and breadth of the current AI revolution.
One of the central dilemmas faced by AI scientists today, as Hegde points out, is choosing between creating foundational models like BERT or applying these models to new and innovative domains. This decision is not trivial, as it shapes the direction of one’s work and the impact one can have within the AI field. While creating foundational models requires an immense investment in resources and expertise, applying these models to solve real-world problems can lead to immediate and tangible benefits. This juxtaposition underscores the diverse roles AI scientists can play in shaping the future of technology.
In the rush towards leveraging Generative AI for various applications, Hegde emphasizes the importance of not losing sight of the fundamental techniques that underpin these advancements. Techniques such as diffusion, used in image generation, and GANs, pivotal in applications like style transfer, remain critical to understanding and advancing AI capabilities. The emergence of smaller models and considerations for edge computing and federated learning highlight the nuanced decisions AI scientists must navigate in their work.
As AI continues to penetrate every aspect of technological development, AI scientists find themselves at the crossroads of an unprecedented technological tsunami. The relentless pace of innovation, coupled with the expanding toolkit of AI technologies, presents both opportunities and challenges. Hegde’s insights prompt AI scientists to critically evaluate where they can contribute most effectively, whether in creating the next breakthrough in AI or in applying existing technologies to unlock new possibilities.
Jayprasad Hegde’s talk at MLDS serves as a compelling call to action for AI scientists and practitioners. As we stand on the brink of a new era in AI, it is crucial to engage thoughtfully with the technologies at our disposal, understanding their implications, challenges, and the vast potential they hold. Whether one chooses to innovate at the foundational level or apply these advancements to solve pressing problems, the journey through the landscape of Generative AI promises to be both challenging and rewarding.
By delving into the intricacies behind the scenes of Generative AI, Hegde not only illuminates the path forward but also invites AI scientists to ponder their role in navigating and shaping this transformative wave. The future of AI, as seen through the lens of an AI scientist, is not just about embracing the dreamy potential of these technologies but also about making conscious choices that will define the trajectory of innovation and its impact on society.
