zlacker

>>i13e+(OP)
> Once trained on this giant corpus of natural language, the chatbot can, when prompted by users’ instructions, produce a fluently written answer rather than merely serving up a list of links.

A word salad of untrustworthy bullshit instead of the pages you were looking for is not going to upend search.

It's not a replacement for search in any way.

Now if the search engine could find pages which satisfy complex queries in natural language, referencing content and relationships among content, that could be something.

>>kazina+xr
But google (nor bing) return the pages you're looking for even close to as reliably now as they did before. If Microsoft can somehow work out the issues with factual accuracy in the gpt interface (and salvage the bing brand) they're going to have a winning product on their hands.

>>friend+1G
Users sometimes want a factual answer. Usually they want to go to a page. Pages are not just sources of information. People interact with pages to do real stuff connected with off-line activities. E.g. chatty paragraph won't connect me with the government/medical/banking/restaurant/shopping/whatever page I'm too lazy to search for through its own URL and portal.

At best, you can have this: if it looks like the user is typing a question into the search box, an AI-generated side bar can complement the answers.

The web search side of it cannot go away, otherwise you just have a sophisticated Eliza.

If I'm going to see a factual answer, I want external references for those facts before I trust it. I'm likely not a subject matter expert, if I'm asking about something; I can't tell when the AI is bullshitting.

Now if the AI-powered search engine had an intelligent index of all the texts of all the crawled web pages such it could do some basic reasoning about it to serve up those pages based on sophisticated questions (and not simply serve up original prose based on those texts), that could be a game changer.