The announcement of LaMDA & Bard is not surprising; there are claims outside Google that ChatGPT tech is a threat to search, hence this reaction. My perception is that this is a trend comparable to the gradual demise of tech support we saw at the end of the last century. It used to be that one could reach highly knowledgeable developers to assist in problem solving. That turned out to be too expensive, so companies added layers of intermediaries that would discuss and triage, thus filtering calls to protect developers. For some problems, the added layers degraded tech support. So now, with LLM search, the agent will talk endlessly about your query, shielding traditional keyword/specialist search engines. For some customers, this will actually improve matters because their searches can be answered with a few superficial facts. For more technical queries and expert users, this approach will be wasteful.
@muratdemirbas Productivity hack that Cal did not try (I think). https://simonberens.me/blog/i-hired-5-people
@muratdemirbas This almost makes vi usable. https://github.com/Exafunction/codeium.vim
The dog's name is "Question Hound" (of the "this is fine" meme)
https://www.npr.org/2023/01/16/1149232763/this-is-fine-meme-anniversary-gunshow-web-comic
RE - the visualization of AUCs and the problem of needing a single metric (like "the one ring to rule them all"). I think you walk a tightrope in that being precise and correct, stipulating confidence and adherence to best practice, you can overdo the formality and make the reader need to work hard to understand the point. Would need to be written carefully - and with the right kind of citations.
Computer Scientist, mineral collector, gardener, bike enthusiast; un peu de français; nederlands maar erg roestig.