Generative AI (GenAI) can transform the travel and tourism industry, delivering personalized and seamless experiences that redefine customer engagement. With its ability to dynamically plan itineraries, GenAI provides tailored recommendations based on individual preferences while enhancing convenience and satisfaction. Virtual agents powered by GenAI offer 24×7 customer support, resolving queries with remarkable speed and efficiency. Simultaneously, AI-generated content immerses travelers in destination simulations, while predictive analytics optimize flight bookings and hotel stays. By anticipating travelers’ needs, GenAI elevates the entire travel planning process, making it smoother, more engaging, and intricately personalized.
In an exclusive conversation with The Interview World, Dr. Shakti Goel, Chief Architect and Data Scientist at Yatra Online Ltd., underscores how generative AI is revolutionizing customer experiences in the travel sector. He elaborates on Yatra’s innovative use of GenAI to enhance user experiences. Furthermore, he delves into the complexities of hallucinations in GenAI, and shares forthcoming advancements in the AI space. The following are key insights from his enlightening interview.
Q: How do you envision the adoption of Generative AI transforming the travel industry, particularly personalizing the customer experience?
A: GenAI can enhance customer experiences by creating personalized travel packages. For instance, if you’re planning a trip to Udaipur and share your interests, GenAI can book your flights and hotels. Moreover, it will recommend tailored activities based on your preferences.
The information on local experiences is sourced directly from large language models, ensuring rich, dynamic content. Moreover, when searching for hotels on a website with hundreds of options, GenAI streamlines the process by curating a customized list that aligns with your interests.
GenAI’s capabilities extend further, enabling it to respond intelligently to emails. Sans human intervention, customers receive swift, well-structured, and intelligent replies, significantly improving response times and overall communication efficiency.
Q: With Yatra serving millions of customers, how do you see the adoption of Generative AI shaping your approach to enhancing user experiences? Could you share specific use cases where GenAI has enhanced your operations by personalizing and optimizing customer interactions?
A: At Yatra, we have developed advanced tools tailored for the corporate sector, where automation and process digitization are in high demand. Leveraging GenAI, we enable companies to streamline their operations efficiently. For instance, extracting specific details from an expense receipt is traditionally a tedious manual task. While OCR technology can sometimes fall short, GenAI provides a far superior solution, delivering accurate and reliable results.
Additionally, the GenAI-powered bots we’ve created offer real-time, highly relevant information to customers, significantly enhancing their experience and ensuring they receive the most accurate support when they need it.
Q: While leveraging Generative AI for customer interactions, what are some instances where AI hallucinations have occurred, and how do you address these inaccuracies to ensure a seamless customer experience?
A: Generative AI is inherently prone to hallucinations. Yet, the precise reason for their occurrence remains elusive. This stems from large language models (LLMs) that are trained on vast datasets and possess billions—not millions—of parameters. These models function by striving to minimize the error in an objective function, aiming to optimize their parameters accordingly.
The key question is whether the error has been minimized or the model has reached the optimal set of constants and parameters. When you pose a question, the model may sometimes retrieve an inaccurate or fabricated response, as it can tap into a less accurate portion of those parameters. This issue is unlikely to disappear soon.
Achieving a significant reduction in hallucinations requires extensive training to reach the global minimum of errors. One effective method to mitigate this is through Retrieval-Augmented Generation (RAG), where you integrate your dataset with the model’s semantic capabilities. By doing so, the model retrieves answers from your specific data, and you can implement guardrails in your prompts—directing the model not to respond to certain questions. This layered approach helps to minimize the occurrence of hallucinations.
Q: What emerging technologies do you foresee following Generative AI, and how do you envision the world adapting to and leveraging these advancements soon?
A: Generative AI remains in its infancy, yet it is already driving remarkable advancements. Recently, the Nobel Prize in Chemistry was awarded for AI modeling, specifically for GenAI’s ability to predict protein structures. This breakthrough transforms the process of determining protein structures, reducing what once took 10 to 20 years to mere minutes or seconds. Such an achievement is nothing short of extraordinary.
These models, connoted as “super AI models,” excel in specific tasks, outperforming human capabilities in their designated areas. We are witnessing a trend where super AI GenAI models proliferate across various industries, including healthcare and law. This trend points toward a future dominated by specialized models designed to address particular challenges with exceptional precision.
