|
|
|
|
|
|
by seaourfreed
387 days ago
|
|
|
Replacing Google with open source. Imaging this:
1) An open-source competitor to Google Search exists
2) It exists in "Front-end nodes". This is the website user interface. Search query entered here. It then searches cached indexes.
3) Searches fall in two patterns.
a) Frequent searches (top 20m search query phrases),
b) long-tail rare (beyond top 20m)
4) Here is how "Frequent Searches" results are built.
* They are built before. Updated once a month
* There are LLM prompts on how to assess web pages against a search query. This is about taking the top 5,000 website contenders, and ranking those exactly
* There can be contention on these LLM. This is how that is solved.
* Stake holders. Each is a company or an owner of a website with human content. They have the ability to vote. This is towards "fair" LLMs to reason on finding the best match for a term.
* They vote away LLM prompts that skew or overly focused on one area.
5) Long-tail searches may be redirected to a DuckDuckGo like solution. Bing / Google / DuckDuckGo compete to be fair to win the traffic
6) The search front-end gets add revenue for Frequent searches. Their job is to do advertising to get customers to switch from Google search
7) The Frequent Search final results of voting are saved in the blockchain. From the once a month LLM compute.
8) LLM compute may happen with different LLM vendors, to weed out bias.
9) The LLM prompt runs (once a month), may happen at ~200 (or 2,000) different open source servers. Run by different people. What wins, is consensus on what most matches. This weeds out bad actors and bias.
10) Big non-tech companies may pay for this, in order to get SEO to work again, for their revevnue. Organic traffic. Ford, GM, Home Depot, AT&T, SalesForce, Oracle, J&J, P&E, GE, Colgate-Palmolive, General Mills, Kraft, etc. |
|
|
https://yacy.net