Instead of keywords for questions and links for answers, Kagi.ai offers a new way to search, using natural human language as input and direct answers as output.
Kagi will provide an end-to-end solution for answering even the hardest questions. Our approach takes full advantage of every scrap of human knowledge available on the web to achieve state-of-the-art performance on open-domain question answering. It also leads the market in the scope of questions it can handle. Read more about it in our blog post.
Kagi is the only question-answering service that uses the wealth of unstructured data from the web (or any custom text corpus) to answer natural language questions. For computational queries, we partner with third-party services such as Wolfram Alpha to provide accurate results. This hybrid approach (language model, computational model and prior knowledge from Internet) ensures Kagi has widest scope of questions it is able to address. Our demo will give you a good sense of technology capabilities.
The search technology behind Kagi can work in any information retrieval scenario, including for businesses in internal document archives and knowledge bases. Check Kagi Search for Enterprise.
For data geeks, we power our technology demonstration with a cluster of 32 CPUs and 10 GPUs with a processing throughput of 40 questions per second (3.5M/day).