Keyword search is dead?
Keyword search is dead?
Perhaps we might rather say that other search modalities are now showing more signs of life? Though perhaps also distinguish keyword search from fulltext search or with reference to various ways searching is mediated (from [stopwords](/search/about#stopwords) to [noindex](/search/about#noindex) and [search query length limits](/docs/2023/07/06/search-query-length-limits.html)) When is keyword search still particularly valuable? (Cmd/Ctrl+F is still very alive?) How does keyword search have a role in addressing hallucination?
Surely though, one exciting thing about [this moment](/hire-me) is how much people are reimagining what search can be.
Perhaps we might rather say that other search modalities are now showing more signs of life? Though perhaps also distinguish keyword search from fulltext search or with reference to various ways searching is mediated (from [stopwords](/search/about#stopwords) to [noindex](/search/about#noindex) and [search query length limits](/docs/2023/07/06/search-query-length-limits.html)) When is keyword search still particularly valuable? (Cmd/Ctrl+F is still very alive?) How does keyword search have a role in addressing hallucination?
Surely though, one exciting thing about [this moment](/hire-me) is how much people are reimagining what search can be.
@vectara via Twitter on Jun 15, 2023
Keyword search is dead. Ask full questions in your own words and get the high-relevance results that you actually need.
đ Top retrieval, summarization, & grounded generation
đľâđŤ Eliminates hallucinations
đ§đ˝âđť Built for developers
⊠Set up in 5 mins
[vectara.com](https://get.vectara.com/)
See: @duguid2012world for a look at grep and the opening/closing of search. I'm intrigued with how much generative search might support an opening up of searching.
While LLMs are often spoken of as black-boxed, the underlying models are but one part of the larger sociotechnical practice and much of what we may want made transparent exists outside the models themselves. Increased competition (due to lower scaling costs and the blue ocean ([wc?](/glossary#wc)) environment, ["where nobody is an expert"](https://twitter.com/chrisalbon/status/1649096741378195459)), contextually-relevant explanation may introduce massive shifts in each of the three forms of opacity raised in @burrell2016machine:
While LLMs are often spoken of as black-boxed, the underlying models are but one part of the larger sociotechnical practice and much of what we may want made transparent exists outside the models themselves. Increased competition (due to lower scaling costs and the blue ocean ([wc?](/glossary#wc)) environment, ["where nobody is an expert"](https://twitter.com/chrisalbon/status/1649096741378195459)), contextually-relevant explanation may introduce massive shifts in each of the three forms of opacity raised in @burrell2016machine:
(1) opacity as intentional corporate or state secrecy, (2) opacity as technical illiteracy, and (3) an opacity that arises from the characteristics of machine learning algorithms and the scale required to apply them usefully.Beyond changing the surfaceâshifting away from ten blue linksâsearching is being reappraised and it may become shorn from the hegemonic notions of the last twenty years. (I'll acknowledge that automation bias may foreclose an opening of some searching, but that story hasn't been written.)"
I do really appreciate the phrase "full questions" (highlighting added), and created a tag of it.