May 11, 2023
When people use a search engine, they’re usually looking for information. Sometimes they want to buy something, and sometimes they want to go to a website that they don’t quite remember the web address for.
But most frequently, they’re trying to answer a question.
The recent rise of AI-fueled chatbots may change the way we search for information – because these chatbots are very good at answering questions.
Which raises another question: if a user can ask an AI model for information in a conversational format, why would that user rely on a search engine that requires clicking on links and navigating several web pages?
“Search engines will have to interact with human beings as if they were human too; therefore, searches will become like a conversation between humans,” said IEEE Senior Member Euclides Chuma, whose work focuses on the development of human-centered systems.
A Brief History of Search
In the early days of search engines, users typed in a word or phrase, and the engine showed websites containing that phrase. It worked well for a while, but had its flaws. For example, if you misspelled “bicycle” as “bicicle,” you’d get no results unless a website also made the same mistake. Search engines also struggled to understand what users really wanted when they searched. The search engine couldn’t tell if a user wanted to buy a bike or find the history of the bicycles.
Now, 20 years later, search engines are much better, and smarter. They can find bike shops, fix spelling mistakes and understand unclear language. This is all thanks to AI.
A New Way to Find Information
The AI that search engines use to connect users to information isn’t quite the same as the AI used by today’s headline grabbing chatbots.
The chatbots, which are often referred to as “generative AI” because they can create new content, are able to answer questions with original content, rather than showing content made by others. It will also offer a result whether someone has made a website to answer that question or not. And since it’s conversational, it can give information in a way that’s easy for the user to understand.
But search engines today are still heavily reliant on outside publishers to generate content that meets the needs of their users. And anyone that has used a search engine recently knows that, while they are remarkably good, the content they offer doesn’t always provide exactly what you are looking for. In fact, Google says that about 15% of all searches have never been searched for before.
One problem with popular AI models is they sometimes make things up, called “hallucinations.” But IEEE Member Yale Fox thinks this will get better: “I believe that these models will improve in their ability to separate fact from fiction once algorithms specifically designed for this purpose are implemented.”
In the future, AI-fueled search engines might connect photo recognition with text search. Imagine someone wants to start a heart-healthy diet. An AI assistant could give guidelines for the diet, identify healthy foods from a fridge photo, plan meals and search the websites of nearby grocery stores for the ingredients needed.
“AI models will observe user behavior, interpret images and videos on the screen and proactively offer hints or assistance when users encounter challenges,” Fox said. “The next generation of multimodal AIs will seamlessly blend into the browsing experience, revolutionizing how we navigate and interact with the digital world.”
And that may be only the beginning, Fox said, as generative AI systems begin to write their own algorithms and optimize themselves in ways that humans have never expected.
Learn More: If you are wondering how the rise of AI will impact how we access knowledge and information, you aren’t alone. Check out the IEEE Future Directions blog’s thoughts on new paradigms in search.