| I would pay real subscription money for a search engine that focused on knowledge-oriented results rather than retail and commercial results. When I type “shoes”, it would give me: links for the functional and creative history of footwear, the taxonomy of shoes, methods of construction, current and historical footwear industry data, synonyms and antonyms, related terms and professions, the dictionary definition, and similar links related to secondary meanings (such as any protective covering at the base of an object, horseshoes etc). I’d also hope for a comedy link to a biography of Cordwainer Smith. What I actually get, which I don’t want at all: pages and pages of shoe shopping. The various means to exclude “top X sites” are the roughest possible heuristic in that direction, and throw out the baby with the bathwater (for example, a long-established manufacturer may well have an informational online exhibit) Google has essentially failed me in its primary mission. Bing at least has the grace to admit they are here to “connect you to brands”. And sadly, right now, every other option is an also-ran. In practice I use DDG, directed by !bangs towards known encyclopaedic or domain-specific sources. I am certain that I’m missing out. |
* when you make a query to this knowledge base, it has a history of your prev searches / preferences (not google)
* it can propose variants of suggestions on what is your intent in this particular query - and make much more detailed queries (auto include/exclude keywords websites etc) to multiple sources (not only google, maybe anonymously)
* it can parse results from these sources and re-arrange them (use own rank system) according to the your preferences. In this system, you can explicitly say - I hate that, and I like that, and this will affect the behavior. Yes this is 'information bubble' but it is controlled by you and not by google!
* finally, this system may work in background and handle 'research' search queries. What I mean here: currently, Google is about instant search - it gives you results in milliseconds, and that's all. It cannot spend much computations for more precise, more intellectual check of content in links from the search results - it cannot do reasoning - and you have to do that by yourself: open links from 1-st page, and close most of them immediately b/c they are not relevant for you, go to 2-nd page and so on. It would be cool if most of this could be automated - with modern natural language processing approaches and old-school prolog-like reasoning this is real and not a fantastic from sci-fi.
My vision that this kind of search assistant cannot be SaaS / closed source. It is about the freedom - and thus this should be open source / self-hosted app that can be deployed on PC or on cloud VM - but hosting should be controlled by end-users, not companies.
I don't know if something like this ever exist. If not, maybe its time to create it.