I think that at this point Google's primary advantage is that they have a generally up-to-date snapshot of the world as it currently exists. Practically every business is integrated with Google, and practically everybody uses Google to find things in the real world like restaurants or local services. Google doesn't give a damn about whether people use the service to learn about some topic or generate boilerplate code or something - those queries aren't really where the money is anyway. It's essentially become the high-tech modern equivalent of the phone book, and despite what HNers often say, it is damn useful.

I honestly don't see any way that OpenAI will be able to replicate these structural advantages using ML models. How are they going to get up-to-date info about the millions of businesses in the world? How are they going to compete with the very built-out infra Google has for customer ratings, most busy times for locations, etc? Google has been building good will with real-world organizations for decades to achieve what it has now. What, you think every tech-illiterate business owner is going to suddenly jump onto OpenAI?

My prediction is that whatever OpenAI offers will likely have a similar use case to ChatGPT. Good for learning, good for exploration of topics, but bad at giving accurate and up-to-date real-world info. I'd love to be proven wrong, for OpenAI to release a really revolutionary product that makes Google obsolete. But it strikes me that this problem of "matching users to the accurate information they desire" is just as much of an infrastructure / logistics problem as it is a software one, if not even more so. And OpenAI just doesn't have the scale or the capability.

Discussion [0] (40 points, 6 days ago, 37 comments)

[0]: https://news.ycombinator.com/item?id=40235206