What's Happening?
Octen, a new startup founded by a former Alibaba AI search lead, has introduced an AI search service utilizing a proprietary model called Octen Pulse. According to VentureBeat, this service employs a generative retrieval technology that can identify webpages
in approximately 10 milliseconds. Unlike traditional search engines that index and re-rank candidate pages, Octen's trained model directly generates an identifier corresponding to an existing URL. The company claims its 4 billion parameter model covers about 2.01 billion webpages, with outputs restricted to identifiers within a URL list to prevent the model from fabricating non-existent URLs. Octen also offers an unlimited web and image search solution for $20 per month. It's important to note that the 10-millisecond figure represents the median server-side processing time, not the total time from enterprise application query to answer. The downloadable Pulse model also has non-commercial use restrictions.
Why It's Important?
This development signifies a potential shift in the landscape of AI-powered search and information retrieval. The rapid generative retrieval technology could offer significant speed advantages over conventional search methods, which typically involve more complex indexing and ranking processes. For businesses and developers, faster access to relevant information could enhance the efficiency of AI agents and applications that rely on real-time data. The subscription model for unlimited searches also presents a new commercial approach to AI search services, potentially challenging existing search engine paradigms. However, the non-commercial use restrictions on the downloadable model suggest that widespread enterprise adoption will require careful evaluation of its applicability, cost-effectiveness, and actual performance in real-world scenarios.
What's Next?
Enterprises considering Octen's solution will need to conduct thorough testing to assess the utility of its search results, cost implications, and whether the promised speed genuinely improves task completion. The company's ability to scale its 2.01 billion webpage coverage and maintain accuracy will be crucial for broader adoption. Future developments may include expanding the model's coverage, refining its generative retrieval capabilities, and potentially offering more flexible licensing options for commercial use. The success of Octen could also spur other AI companies to explore similar generative retrieval approaches, intensifying competition in the AI search market.
Beyond the Headlines
The emergence of generative retrieval technology like Octen Pulse raises interesting questions about the future of information access and the role of AI in mediating our interaction with the web. By directly generating URL identifiers, this approach could fundamentally alter how information is discovered and consumed, potentially leading to more direct and less curated search experiences. However, it also introduces new challenges related to bias in model training, the potential for 'hallucinations' if not properly constrained, and the ethical implications of AI-generated search results. The shift from traditional indexing to generative models could also impact content creators and SEO strategies, as the visibility of information might depend more on a model's internal representation than on conventional ranking factors.













