July 2024 Summaries

Exa, an AI research lab, has announced $22 million in seed and Series A funding led by Lightspeed Venture Partners, with contributions from NVentures and Y Combinator, to develop a search engine tailored for AI applications. Unlike traditional search engines optimized for human interaction, Exa utilizes embedding models to convert web pages into numerical data, enhancing the precision and depth of search results for AI. This approach allows AI to efficiently retrieve relevant information, thus serving as a critical data layer for AI systems. The company, co-founded by Will Bryk, has already seen rapid adoption by various companies and developers, with its revenue tripling recently. Exa’s mission is to organize the world's knowledge, facilitating AI applications across diverse industries, and it is supported by a vision to redefine how AI leverages information, backed by Lightspeed's investment in disruptive innovations.

Will Bryk argues that the development of "superknowledge" is a more pressing technical challenge than superintelligence, as the current knowledge systems, exemplified by Google, are insufficient for the advanced AI models like GPT-4, which possess high intelligence but are limited by their knowledge retrieval capabilities. He highlights that superknowledge, defined as a system capable of handling complex retrieval requests, would significantly enhance productivity and information access across various fields, from healthcare to AI research, thereby accelerating societal progress and safety. Bryk believes that superknowledge could prevent inefficiencies and misinformation, which are exacerbated by the scattered nature of existing knowledge systems, and posits that Exa, a company working towards this goal, could achieve it within a few years by leveraging advanced technologies like transformers. He emphasizes that creating superknowledge before superintelligence is crucial to navigate the impending volatile decade safely, advocating for a mature knowledge ecosystem that could guide both humans and AI in making well-informed decisions.