fbpx

Exploring the Evolution of AI in Search and Contact Centers: Insights from Google’s Vlad Vuskovic

Exploring the Evolution of AI in Search and Contact Centers: Insights from Google’s Vlad Vuskovic 1792 1024 Kane Simms

Vladimir Vuskovic, a key figure at Google in the development of AI, particularly in Google Assistant, search and contact centre applications, recently joined me on VUX World to discuss the journey and future of AI technology.

With over a decade at Google, Vlad has been instrumental in shaping tools like Google Assistant and is currently focused on advancing search AI. This article summarises his insights into the evolution of AI, the integration of large language models (LLMs), and the future of AI in enterprise applications.

Introduction

AI technology has significantly evolved, impacting various domains, including search engines and contact centres, as we know. Google has been at the forefront of this evolution, developing tools that leverage AI to enhance user experiences, as well as some of the fundamental infrastructure that paved the way for the powerful models we see today. Vlad Vuskovic’s journey at Google, from working on YouTube and Google Assistant to now leading search AI initiatives at Google Cloud, offers a deep dive into how AI has transformed these fields and where it is heading.

Challenges Faced in Early AI Development

Vlad’s involvement with Google Assistant began during a time when AI technology was not as advanced as it is today. The initial challenges revolved around the limitations of the existing AI models, specifically in natural language understanding (NLU). At that time, recurrent neural networks (RNNs) and long short-term memory (LSTM) networks were state-of-the-art but had significant drawbacks. These models struggled with understanding long texts, maintaining context over extended conversations, and processing information in real-time. Moreover, they were not efficient in learning from vast amounts of data, often forgetting critical information—a problem known as vanishing gradients.

To overcome these limitations, early versions of Google Assistant relied on manually crafted grammars, developed by linguists. This approach, while functional, was labour-intensive, expensive, and not scalable. Each language required a dedicated team of experts, making it challenging to expand and maintain the system across different languages and dialects.

The Shift to Transformer Models

The breakthrough came with the introduction of transformers, a type of AI model that addressed many of the issues inherent in RNNs and LSTMs. Transformers allowed for parallel processing of text, enabling faster and more accurate language understanding and generation. This technology laid the groundwork for Google’s BERT (Bidirectional Encoder Representations from Transformers), which significantly improved the ability of AI systems to understand context and generate meaningful responses.

At Google, BERT was quickly adopted to enhance intent recognition in Google Assistant, replacing the older, less efficient models. This shift allowed Google to scale its AI capabilities more effectively, enabling more natural and fluid interactions between users and AI systems. The introduction of transformers and BERT marked a pivotal moment in the evolution of AI at Google, paving the way for more sophisticated applications.

Solutions Found: AI in Contact Centers

After successfully integrating AI into consumer products like Google Assistant, Vlad transitioned to Google Cloud, where he applied these advancements to enterprise solutions, particularly in contact centres. The contact centre environment presented a unique opportunity for AI deployment, as it already involved natural conversations between users and agents. This setting was ideal for implementing AI-driven solutions that could enhance user interactions by providing real-time support and automating routine tasks.

Google’s contact centre AI evolved from simple chatbots to more complex systems capable of understanding user intent, retrieving relevant information, and even reasoning through problems. This was achieved by integrating large language models with traditional rule-based systems, offering a hybrid approach that balanced the strengths of both methods. For example, routine tasks like payment processing could be handled by rule-based systems, while more complex queries are managed by generative AI, which could provide contextually relevant responses.

One of the critical advancements in this space was the introduction of grounding in AI models. Grounding allows AI systems to pull real-time, authoritative information from specified data sources, reducing the risk of hallucinations (incorrect or nonsensical outputs generated by the model). This feature is particularly crucial in contact centres, where accuracy and reliability are paramount.

Results: AI in Search and Beyond

Vlad’s current focus is on advancing search AI at Google. Search, whether internal or external, is a fundamental aspect of many business operations. Google’s Vertex AI Search, a suite of tools developed under Vlad’s leadership, leverages the same generative AI technologies to enhance search capabilities across various domains. These tools allow businesses to implement advanced search functionalities, whether for product discovery on e-commerce sites, internal data retrieval for employees, or complex research queries in specialised industries.

Vertex AI Search incorporates both semantic search, which uses vector-based data storage to understand the meaning behind queries, and traditional keyword-based search, offering a hybrid approach that maximises search accuracy. Additionally, features like dynamic retrieval and grounding ensure that the information provided is both relevant and reliable, with traceable sources and reduced latency.

Summary

The conversation with Vlad Vuskovic highlights the incredible strides made in AI technology, from the early days of Google Assistant to the cutting-edge capabilities of Vertex AI Search. The integration of large language models has revolutionised both consumer-facing products and enterprise possibilities, making AI more versatile and powerful than ever before.

As AI continues to evolve, the focus will likely shift towards even greater contextual awareness, multimodal capabilities, and real-time processing. These advancements will open new possibilities for AI applications, from enhancing customer interactions in contact centres to solving complex research problems. Google’s ongoing innovations, under the guidance of experts like Vlad, will undoubtedly play a crucial role in shaping the future of AI.

Big thanks to Vlad for joining me on the podcast. You can listen to the whole episode on Spotify, Apple Podcasts, YouTube, LinkedIn Live or wherever you get your podcasts.

    ☎️ Exclusive session - When AI picks up the phone: How to manage AI agents at scale
    This is default text for notification bar
    Share via
    Copy link
    Powered by Social Snap