Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

GPT4s killer revenue stream isn't selling ads on keyword searches. It's selling access to a general compute engine that can take text and do useful things with it. The API is dead simple to use, and all you have to do is just change a line of text to "upgrade" to the newest models over time. (By the way, the upgrade from ChatGPT -> GPT4 is huge).

Google should be deathly afraid of this. Not because Microsoft is going to replace them with Bing. But because, GPT is going to be in every major software product that's connected to the internet within a few years. By the end of the decade if not sooner agents powered by LLMs will be the primary mode of interacting with the internet. Going to Google.com or whatever to search for an answer is done. You'll just ask Siri 3.0 and the answer will be good enough.



I’ve said this before on HN, but if LLM agents catch on, won’t it be self-defeating in the long term? They need to pull tons of info from the web to be useful and up-to-date. If we all stop visiting websites, then who’s going to keep publishing all of that info?


Well, I would assume a lot of the producers, for example people making products would still have sites the agents could pull from to have up to date information. You'd most likely lose a lot of the middlemen that take that data, add something, and then cover the page in ads.

There are also tons of ego websites that aren't going anywhere from what I can tell. Tiktok will still be around unless it gets banned, and something else will pop up behind it if it does.


Product descriptions are a good example of info that will still be published, because people will still by buying and selling products. The problem is with writing where the writing itself is the product, such as news.

The LLM relies on free access to that writing in order for it to be informed about current events. The news publisher relies on readership to pay writers. If LLMs are too useful to the point that people rely on them alone for news, then ironically they will become worse products over time as the publications they draw from go out of business.


Who says this is true? The AGI will certainly be able to figure it out without depending on some crap medium blog


The ad based revenue model is done, yes. A lot of sites will still exist though behind paywalls.


Doesn’t that mean that search products will also struggle, since they rely on everything being freely available?


The internet you’re describing sounds like hell on earth. I don’t want to talk to an llm.


Out of curiosity, have you played around with ChatGPT-4 at all? While I would disagree with you on v3.5 alone, on v4 I _strongly_ disagree with that sentiment. Interacting with it is significantly better than modern search engines to get a mental model of something, slight mistakes and searching to validate claims and all.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: