@emilymbender Yeah. This is sort of how the WAIS system from Thinking Machines was meant to work, *forty years ago*. An interesting experiment at the time, but there's a reason it was supplanted by other tools.

@emilymbender I would like to add that i have always been of the opinion that spouting orders in natural language has never been a good interface for querying any system, machine or human. There is a reason why ANSI standards begin with phrases like "in this document, MUST means exactly X". Rather, voice assistants earlier and LLM chatbots now are attempting to market to customers a "feeling", reproducing the social power dynamics of being a successful person who has underlings to order around.

@emilymbender people sending me LLM suggestions in a field where I would probably count as "Expert" and me having to explain that it doesn't work because of xyz has become a big part of my work day.
LLMs have helped me personally learn things adjacent to what I already know though. Or do tasks that I personally find tedious but can check the correctness quickly. 1/2

@emilymbender @jqheywood

Very true. Sadly the industry has conditioned people (often students) that search results = understanding. You’re so right …. the CONTEXT of searching is often more informative (and accurate) than just the quick search result ‘answer.’

Answers alone =! wisdom, insight, or intelligence. As a consequence, many functional idiots are created each year - with obvious ramifications for humanity.

@emilymbender provenance! Amen

@emilymbender "hold high expectations regarding provenance --- i.e. citing of sources."

we all need to expect and demand that search results include links to the source and we should not boost any statement that do not include the source, images should include the creator and CC license

Info access is going to become critical. Just as land ownership in the middle ages was controlled by the wealthy, information access will be the next battle.

#democracy is more than how we vote!

@emilymbender thank you for this great thread. I can see use cases where there is a correct answer, but those will be a smaller subset and require considerable curation. Perhaps they’d be appropriate on a site wide level, but not the entire vast internet

@emilymbender For code, it's often trivially easy to confirm whether you got the 95% of correct answers or 5% of incorrect. You plug the snippet of code in, execute it, and see if it passes your test or not. Chatbots have revolutionized data retrieval for programming languages. A chatbot can surface an esoteric piece of documentation much faster than a developer can find it and present the documentation in a more understandable way, e.g. by presenting a worked example.

@emilymbender I've been on the internet for 30 years. My bullshit detector is finely tuned, and I've learned to treat the majority of information as suspect until it can be corroborated by sources with good reputations. I fear for people who haven't developed that skill yet.