On June 5th, we're hosting our free virtual AI & the Workforce Summit, where Jeff Dalton will offer a deeper dive into the AI revolutionizing how we learn and grow at work.

Learn more

[[divider]]

From the invention of the printing press to the birth of the internet, technological leaps have tended to expand both the amount of information we can access and how quickly we can access it. With each passing century, more of the knowledge we use to navigate the world has moved outside of our heads and into libraries—first physical ones filled with books, and then digital ones accessible over the internet. 

As a technologist, I’ve spent my career studying the science (and art) of “information retrieval.” But in recent years, I began to notice that my colleagues and I had become stuck in the system we created. 

We lived by the basic structure of keywords leading to pages. When we wanted to know something, we just “Googled it”—relying on Alphabet’s algorithm to reliably organize the world’s information into a list of text-based links. It all became so familiar that we stopped seeing that those URLs were predominantly static, and frequently so was the page they pointed to. Social media and news sites added immediacy and flexibility. But they remained chained to the old model of media made for many to consume—rather than dynamic information generated specially for you.

AI changes all of that. It makes it time we break out of the search box. 

Generative AI changes the way we access information. It makes it dynamic, something that can be synthesized and generated on command. Just as the internet meant a card catalogue was no longer the best way to find what we were looking for, AI asks us to rethink and reconsider how we organize, access, and apply information of all kinds. 

Notably, I’m not only talking about Large Language Models. I have moved away from thinking of them in that way, and prefer the term “Foundation Models”—because they now do a lot more than just predict the next word. They create images or videos, synthesize voices, and talk to you much like you would another person. They seek out the information we are looking for, acting increasingly like a compression algorithm for the internet. 

This is a transformative moment for “information retrieval”—for all the ways we satisfy our basic curiosities (in life), and the ways we process information to create value from our time (in work). 

In this blog, I look at the biggest ways that AI changes how we access information, and changes the nature of information itself. And I explain what that means for our AI coach Nadia. 

‍

‍

4 ways Generative AI changes information and how we access it. 

1. The future is customized content and personalized conversations.

Google search is based on documents and web pages. Google’s algorithm takes these as the base unit of information, measures their value, and then ranks them in response to a query. New pages are constantly created, but the page content only changes when someone decides to update it. Because the pages are fixed, they are also written for a general audience, not for you and your specific search. 

Generative AI evolves past that, to a new paradigm of nuanced and personalized conversation. It is not just that we have new ways of accessing information, but that the information presented to us is changing. It is itself dynamic. The answers it presents won’t be the same twice. Each answer looks across the library of information available to an AI model and synthesizes it in real-time for that specific conversation. 

2. The future is multimedia conversations.

Search today is built for and structured around written text, and written text has dominated information since the beginning of the printing press. Even when we create a video, it’s largely the written text associated with the video that powers search. 

AI is evolving human language from written text to conversational streams of images, audio, and video (what some of us in the field of information retrieval call “Generated Information Objects”). AI is already rapidly developing the ability to understand and produce this multimedia content. 

3. The future is agentic assistants.

Once we recognize that there is a new species of information out there, generated for us individually, then we can also consider new interaction modalities—way beyond search boxes. Google was a huge leap forward in sifting previously unfathomable amounts of information. But it was still at heart a librarian looking through a stack of information for a specific document. 

AI agents are like having a team of personal assistants that don’t just seek out the information we specify, but expand and act upon your query. In the jargon of tech, we call this an “agentic architecture.” As these agents become embedded into our technology, we’ll be doing a lot less sifting of information, and relying instead on our AI agents to understand and even anticipate what we want, push information and insights to us, and help us act based on our specific request and context. In other words, they are less about retrieving from the Internet, and more about acting on it for our specific use. 

4. The future is integrated devices.

Today we primarily access information through a computer or phone screen, and a lot of our time is spent toggling between various applications to accomplish a given goal. However, with AI, we will state our goal, and the agent will navigate the various steps—no toggling required.

This begs an important question: What will the devices designed for AI look like? Phones and computers are optimized for the world of apps. Fundamentally, they have to be integrated into where and how we work. We’ve heard a lot about augmented reality and wearables, and I believe these are finally coming of age. If smartphones and tablets are any guide, that won’t happen immediately, but it will happen, as the screen and keyboard will yield to voice and other rich forms of interaction. 

What does all this mean for Nadia? 

It was impossible to imagine, let alone build, Nadia, even three years ago. As an AI coach, Nadia is an entirely new class of technology, native to this AI-powered era of personalized information agents. 

Forty years ago, to give frontline managers the best possible on-the-job guidance, Valence might have published a handbook. Twenty years ago, we might have made it searchable, á la Google. But today, we can build an AI coach, offering frontline managers a dynamic guide advising them in the course of their day-to-day work, informed by a deep understanding of who they are and what they uniquely need. 

Perhaps most exciting of all, as we build this we create the ability to model each organization and its unique culture, goals, and dynamics. Nadia is not just your coach, but a guide to help you, your team, and your company reach new levels of performance and purpose.

[[divider]]

Jeff Dalton is Head of AI and Chief Scientist at Valence. He is the author of more than 100 research papers and holds multiple patents in search, natural language understanding, and question answering. Prior to joining Valence, he worked at Google to develop language understanding capabilities for the Google Assistant and build next-generation knowledge graphs for Google search. 

‍