In today’s rapidly evolving technological landscape, artificial intelligence (AI) and generative models have become the cornerstone of innovation. Over the past few months, GenAI has garnered significant attention, making this an opportune moment to delve into it in depth.

As the CTO of Chiratae Ventures, a fund deeply rooted in supporting startups that build for population scale, we wanted to have a larger discussion around GenAI and the potential of this technology, its scale and the future we envision it to impact.  

My exchanges with the Google Cloud Team set forth the idea to get these questions formalised into discussion, and hence, we have started a community of Chiratae portfolio CTOs in partnership with the Google Cloud team to table these conversations, and in the process, enrich our knowledge, share challenges and solutions and most importantly, get together a set of CTOs who are hands-on with technology on a day-to-day basis. 

In partnership with Google Cloud, we hosted a CTO Round Table in Bangalore, which saw good participation from Chiratae Portfolio CTOs from companies working on autonomous vehicles to Financial services, AR-VR, Agritech, Gaming, Last-mile delivery and a few others, all of whom have been working on various GenAI use cases.

One of the key highlights was a deep and insightful conversation with Subram Natarajan, Director of Customer Engineering, along with a few key members at Google Cloud India to delve into their insights on the past, present, and future of AI, especially focusing on generative AI, and how it is poised to revolutionise various industries.

The Accelerated Evolution of Generative AI

During our chat, Subram spoke of the remarkable journey of generative AI. He emphasised the incredible speed at which this technology has progressed, with use cases emerging in just a matter of months. Unlike previous AI concepts that lingered in the realm of potential, generative AI swiftly moved into production and real-world applications.

His insights reveal that AI has been on the radar for a while, but its impact was somewhat limited until the advent of large-scale generative models like Transformers. He shared that the paper titled “Attention Is All You Need” by Google researchers played a pivotal role in the development of generative AI, laying the foundation for what we see today. “However, the journey has only just begun, and the full potential of generative AI is yet to be realised,” he shared.

During our chat, Subram was also quick to draw a compelling parallel between the impact of generative AI, the transformative effects of the internet on information dissemination, and mobile technology on communication systems. He envisions generative AI collapsing the adoption cycle of AI, making it a ubiquitous presence in various industries.

Google’s Unique Approach to Generative AI

Google has been a pioneering force in AI, and its deep involvement in AI has generated curiosity. It was fascinating to gain insights into Google’s strategy on GenAI and how it sets them apart from other enterprises in this domain. Subram shared, “Google’s reputation as a consumer-centric company is well-established, with services like Search, Mail, and Maps being household names. Over the years, Google has seamlessly integrated generative AI into these services, enhancing the user experience.”

He added, “For enterprises, Google focuses on critical aspects such as data control, intellectual property protection, cost management, security, and accuracy. These are the core concerns of businesses when implementing generative AI. Google’s commitment to addressing these challenges and lessons learned from the consumer space positions them as leaders in the enterprise adoption of generative AI.”

The Future of AI in Various Industries

Like many of us in technology are dedicated to driving innovation, we collectively share a vision for the future of GenAI and its impact on the overall ecosystem.

Curious about Google and Subram’s take, it was one of the questions that came up in our discussion. Subram’s vision extends beyond enterprise adoption. He foresees AI transforming every facet of our lives. For example, productivity tools could soon have AI assistants co-creating documents and sparking ideas. Moreover, generative AI is slated to profoundly impact content creation, communication, and numerous other use cases across industries. Generative AI’s influence will be far-reaching, from revolutionising government citizen services to offering personalised retail experiences. He emphasised that while AI plays a significant role, human involvement remains central, ensuring responsible and ethical AI use.

Vineeth from Kristal.ai pointed out how the GenAI use cases of financial services helped solve customer queries. Simplifying the customer reports with some insights.

AI Challenges: Trusting Data Sources

GenAI also involves the critical challenge of establishing trust in data sources, which demands a comprehensive understanding of the data’s quality, reliability, and ethical considerations.

While data quality will always be a question, Karthik – CTO, Metadome.ai shared his thoughts on data issues. He said, “Data is a problem when you know what to do with that data”.

Regarding data, Subram’s primary concern regarding AI is data quality. AI’s inferences heavily rely on data, and manipulating data can lead to biased or inaccurate AI outcomes. Thus, ensuring the integrity and trustworthiness of data sources is critical in the AI-driven future.

Privacy and AI Production

One of the other key concerns around GenAI is privacy and speed of AI production. Both go hand in hand, and in the context of GenAI, it is crucial to have it in place. 

Subram also expressed his concern about AI production in the context of privacy. He recommended applying security controls and filters to ensure safe AI usage. Enterprises can also train models on their proprietary data to enhance AI accuracy and relevance while maintaining control over sensitive information.

The CTOs at the RoundTable also discussed Licensing of LLMs and Demystifying GenAI with Google Cloud’s GenAI Studio. 

As generative AI technology becomes ubiquitous, licensing models might become common practice. Google acknowledges the widespread use of LLMs but emphasises the importance of ensuring data integrity and responsible AI usage.

Raghu Dubey, Customer Engineer at Google Cloud India, provided an overview of Large Language Models (LLMs), highlighting their evolution from Natural Language Processing to Natural Language Generation. He explained Google Cloud’s offerings, emphasising the enterprise-level capabilities of Vertex AI for tuning models, data security, and governance, along with a discussion on various industry use cases that can be solved with generative AI.

The Model Garden initiative demonstrates Google’s commitment to open source, providing curated models for various tasks. Google aims to simplify AI adoption for developers, business users, and AI practitioners, catering to varying levels of complexity.

At the Roundtable we also discussed the crucial role of data, explored use cases and best practices, and ventured into the fascinating world of retrieval-augmented generation (RAG). Some key insights from that fascinating conversation are below:

The Future: AI Integration Across Industries

AI-driven solutions are set to revolutionise industries by improving search experiences, enabling conversational AI, automating document processing, enhancing recommendations, and advancing healthcare applications.

Gagandeep – CEO and CTO, Minus Zero, provided a few interesting stats. He shared that the computation power doubles every quarter. Furthermore, computing costs have been reduced to half in the last five years. 

Leveraging LLMs for Various Tasks

LLMs like GPT-3 have incredible potential, with the models already demonstrating their strengths, particularly in handling text and image data. As a look towards future developments in the field, the industry is excited about the potential for growth and innovation in 3D Model Generation, even though the current offerings may be subpar.

The Crucial Role of Data

A recurring theme in the conversation was the pivotal role of data in the development and utilisation of Gen AI. Interestingly, 3D data was highlighted as potentially richer and more accessible than other forms of data, opening doors to new possibilities. However, it’s worth noting that while the data might be available, there are still significant challenges in effectively harnessing it due to the current limitations in AI technology, particularly in handling 3D data.

Revolutionising Content Creation

Gen AI isn’t limited to one field; it has its fingerprints all over the landscape of content creation. In particular, the discussion emphasised its significance in Extended Reality (XR) and game development. The ability to generate 3D models and environments efficiently could revolutionise these industries by eliminating one of their most significant barriers—time-consuming content creation. While steps have been made in the right direction, there’s a shared anticipation for more advanced solutions in this space.

Best Practices in the AI Frontier

In a rapidly evolving AI landscape, the need for established best practices cannot be stressed enough. The conversation yielded some valuable insights:

• Combining Tools and Technologies: To create effective solutions, it’s beneficial to combine various tools and technologies, leveraging the strengths of each.

   

• Maintaining Focus on the End Goal: As one navigates the ever-changing AI landscape, keeping the end goal in mind, such as delivering the highest accuracy possible, is crucial.

   

• Adapting to Regulatory and Compliance Requirements: Recognising and adapting to regulatory and compliance requirements is essential, especially in sectors like legal and finance.

   

• Embracing Change: The AI landscape is dynamic, with innovations happening regularly. Being ready to adapt and evolve is a key best practice.

Data Security, Navigating Compliance, and Privacy Concerns

The room agreed with the critical concerns surrounding data security, compliance, and privacy. Clear data handling policies around personal data and that of law firms and financial institutions, for example, must be established and constantly updated, while maintaining transparency with all parties.

A Dynamic Landscape

A robust AI system has the potential to revolutionise processes in various industries. For example, in the field of financial advisory, a generative AI system has the potential to leverage speech-to-text and crunch reams of data all while safeguarding user privacy and complying with legal requirements. It will take a collaborative, one aimed at continuous improvement, adaptation to the evolving tech space, and addressing present challenges and future potential.

The Fascinating World of Retrieval-Augmented Generation (RAG)

In addition to discussing AI, the conversation delved into the frontier of retrieval-augmented generation (RAG). This approach integrates LLMs with retrieval systems to ensure more factual and reliable outputs. It introduces exciting concepts like “prompt engineering” and vector searches, adding depth to the AI development journey.

As we navigate the transformative landscape of AI and generative models, one thing is clear: the future is bright with possibilities. The collaboration between Chiratee Ventures and Google Cloud and the continuous evolution of AI technologies promises to reshape industries and create a world where AI augments human capabilities and accelerates innovation.