The complex algorithms that power search engines like Google may have changed how we find and sort information online, but LLMs are the next step in digesting and making sense of the information.
We’re learning that LLMs are not just about predicting words or performing tasks (although those are important aspects of their training) but about synthesising and compiling information to answer search queries conversationally. And I believe there’s a key difference in how people think LLMs can be used versus what they’re best at doing (for now).
The development of digital products built around LLMs was primarily based on the premise that people naturally engage in conversations to seek information. By creating a chat interface, OpenAI made it easier for users to access the information they need directly, reducing the need for complex user flows or information architecture. The interaction happens seamlessly within a single screen, making it intuitive for people to interact with AI products.
Interacting with AI through a chat interface means we can reframe responses, learn, and adjust as needed to get better results. Unlike search engines, which provide static summaries of information on web pages, LLMs involve synthesising and curating. This gives us hope that, through the correct use of AI, important information or links to services that might not have strong SEO (and so aren’t readily picked up by search engines) will be surfaced by AI. This would have an enormous impact on, for example, people’s ability to access legal information and justice services.