Exa Raises $22M to Develop AI-Focused Search Engine

Exa, a San Francisco-based AI lab, has raised $22 million to develop a new AI-focused search engine. Led by Lightspeed Venture Partners, with NVIDIA and Y Combinator also investing.

Exa's founders, Will Bryk and Jeff Wang, both in their mid-20s, started the company before ChatGPT's launch. They aim to improve search by focusing on links rather than keywords, using vector databases and embedding technology.

The company initially offered a free version and various paid tiers, hosting on AWS. Post-ChatGPT, many AI firms sought Exa's API for integrating with their models.

Exa's search differs from traditional methods by predicting links instead of words, reducing clutter and irrelevant content. It categorizes results by type, such as PDFs or company sites, enhancing specificity.

While effective for tech-savvy users, Exa's search lacks direct answers, requiring manual filtering. It excels in indexing but sometimes includes inaccurate or outdated information.

Perplexity, another AI search tool, offers a more user-friendly, question-and-answer format, contrasting Exa's technical approach.

Embedding technology, used by Exa, converts words into numerical vectors, aiding machine understanding. This differs from models like GPT-4, which generate content directly.

Exa serves as a data intermediary for AI systems, focusing on deep data integration, while Perplexity caters to general users with immediate, concise answers.

Overall, Exa's innovative approach to search, though complex, could revolutionize how AI interacts with web data, especially in specialized, technical contexts.