alt Hacker NewsThe solution is what Lauren did, she rolled her own. Once that took teams of experts and big bucks. Now a single ML expert can do it for small bucks because she "needed a restaurant recommendation" and didn't trust the available ones. Soon any mild mannered programmer will have the same capability, and then the muggles will get it, in a mass, just for the asking of their favorite chat bot.
If the progression holds, oodles of recommendation engines can bloom, and it'll be trivial to fork and customize a favorite with a prompt. As the friction of doing large analysis jobs tends toward nil, the Google moat dries up and their commanding height subsides. Too optimistic?
Replies
The data is the key though. How did they effectively scrape the data? Does every restaurant have a website? I bet half rely on Google Maps. So IMHO you are too optimistic because regularly and effectively getting the data is the hard part, not the model.
>then the muggles will get it, in a mass, just for the asking of their favorite chat bot
I guess you can do it right now if you tell a llm your preferences.
There's a couple different threads here.
Can we make a decentralized search engine. Which breaks down into two questions, is it technically feasible and is it socially feasible?
(Maybe the word search would be a bit more broad than retrieving web pages. It could be for everything right.)
I don't know but I'm inclined to say that the difficulty will be more on the social side than on the technical side.
The web was very decentralized 20 years ago, and we had all manner of peer to peer systems already. There just doesn't seem to be much appetite for that kind of thing, at least in the mainstream.
Although there might be something to it, with the AI part of the equation.
Like we had self hostable services for a long time, most people just don't want to be a sysadmin.
Well, I gave Claude root on my $3 VPS. Claude is my sysadmin now. I don't have to configure anything anymore. Life is good :)