OpenAI's Search Depends on Indexing, and Google Holds the Index

Originally published in Chinese on July 26, 2024. OpenAI's search ambition OpenAI is getting closer to general-purpose search, and the general search it opens up should arrive soon. This piece looks at what is different about how it would get its data, and what that does to its relationship with Google. The indexing problem How traditional search engines index By the definition of the product, the problem today is that OpenAI cannot build an index as large as the ones built over history. The absolute core strength of a search engine is the index of its data source system. I believe OpenAI's index of data sources is still not as rich as what Google and Baidu have accumulated. These are two different problems, at different layers of technology. OpenAI's new approach to indexing So it uses a newer way to handle the index. It takes the user's question, breaks it down in an optimized way, gets the answer the user wants, and builds the index of the data in the same process. From "index first, serve later" to "search as a service"