Google’s AI Chief Reveals Secrets to Unlocking Generative AI’s Potential for Enterprises

Yasmeen Ahmad, managing director of strategy and outbound product management for data, analytics and AI at Google Cloud, shares his thoughts on how to unlock the potential of generative AI during the VB Transform conference.

The role of data in improving large language models (LLMs): While increasing the size of LLMs can lead to better performance, it’s not the only factor; domain-specific data plays a crucial role in enhancing the models’ capabilities:

• Smaller models trained on domain and context-specific information can outperform huge models with a large number of parameters, highlighting the importance of data in empowering AI models.
• Fine-tuning and retrieval augmented generation (RAG) techniques are essential for successfully training models on specific enterprise domains, allowing LLMs to understand the language of the business and provide accurate, real-time answers.

The importance of multimodal capabilities in LLMs: Tapping into multimodal data, such as videos, images, and text documents, is critical for enterprises to fully leverage the power of LLMs:

• Google’s study showed a 20-30% improvement in customer experience when multimodal data was used, enabling enhanced understanding of customer sentiment and the ability to bring together data on product performance and market trends.
• Traditional data foundations struggle to handle multimodal data, emphasizing the need for a new AI foundation that can accommodate the future of AI and business data.

The evolution of conversational AI: LLMs must be given semantic context and metadata to provide specific and accurate answers, and they need to be able to engage in coherent, back-and-forth conversations:

• The industry is moving towards the next generation of conversational AI, which goes beyond simple chatbots and acts as a “personal data sidekick” that can ask questions, engage in a chain of thought, and provide query transparency.
• LLMs are beginning to mimic human brains in their ability to break down tasks, think strategically, understand cause and effect, and learn honesty, with real-time capabilities improving rapidly.

Broader implications: The advancements in generative AI are pushing the boundaries of what machines can create and what humans can imagine, blurring the lines between technology and magic. As LLMs continue to evolve and improve, they are spawning new breeds of business and redefining what is possible in the realm of AI. However, enterprises must adapt their data foundations and embrace new techniques to fully harness the potential of these powerful models.