Unlike the traditional way of online searching, Kagi.ai is designed so that when a user types a question, they get the answer immediately. When the user receives a response, the application provides an opportunity to rate the answer and read more about it.
The idea for this startup was inspired by the reasoning that for several decades now, the usual process of searching on the internet has been to type a term in a search engine which then, once you click on “search”, returns a large number of links related to the search term. It is up to you to choose the one in the sea of sources and find the necessary information.
Although this is very simple, there is an even simpler solution—instead of filtering information from multiple sites or reading the entire text for just a few information, startup Kagi.ai should allow users to get direct answers to questions asked.
Vladimir Prelovac, a San Francisco based Serbian entrepreneur who is behind this startup, is not a new name on the Serbian startup scene. He is most commonly associated with Devana Technologies and ManageWP which were acquired by one of the largest domain registrars in the world, GoDaddy, three years ago.
What makes this app unique?
To make it easier to understand how the application works, Vladimir proposes to take a simple question, such as “How far is the Sun away from Earth?” as an example.
In response to this question, in less than a second we get about 60,000 results on Google and a large number of links leading to an article on Wikipedia and other sites about the Sun.
And this kind of result then requires that you enter the site, scroll through, and search for the information needed.
On the other hand, Kagi.ai’s application uses data from web pages and extracts answers by using natural language processing technologies. Kagi.ai technology is intended to be used in any scenario to retrieve information, including document search and enterprise databases.
At the moment, a team led by Prelovac currently has six members, half of whom work in Serbia.
All of them face the challenge of taking the next big step in the search world:
We have very talented researchers and engineers who have tackled this challenge, the most ambitious one so far. We aim to develop a technology that will drive the next thousand billion searches. The first thousand billion searches happened on a desktop, others on a mobile phone and I believe the next one will come through natural language / voice, e.g. you will ask your car to recommend a restaurant on the way.
Although Prelovac has considerable experience in running companies, he points out that the latest project is something he has not encountered before.
This company is different from all the others I have had, in the sense that it is a R&D (research and development) project in a branch that is developing at high speeds and breakthroughs occur almost monthly. We have no target groups, users and traction—all that matters is the quality of the technology, which is measured by the accuracy of the response. We currently have better results than the AI assistants of most tech companies (Apple, Microsoft, Samsung, Amazon, etc.) and are side by side with Google for the breadth and accuracy of answered questions.
In order to further develop the application and to monitor the accuracy of the answers, the company also conducts independent research on the performance of digital assistants. Based on their latest results, Kagi.ai has been equated with Google Assistant and has almost three times better results than the last-ranked Siri.
The business model we use is licensing the technology to the companies mentioned above once the accuracy becomes high enough (more than 85 percent, currently around 60 percent), or launching our own branch of the product (search engine, AI assistant) in this industry.
As Vladimir further explains, their approach is a paradigm where possible inputs that use artificial intelligence remain as open and as general as possible, working to improve the quality and accuracy of the results.
