Skip to content

Let’s delve into medical studies

Published on 15 Comments on Let’s delve into medical studies

Some food for thought:

Wait. ChatGPT overuses the word "delve"? Apparently so. According to a dataset of 50,000 ChatGPT responses, its ten most overused words are:

  1. Explore
  2. Captivate
  3. Tapestry
  4. Leverage
  5. Embrace
  6. Resonate
  7. Dynamic
  8. Testament
  9. Delve
  10. Elevate

I guess if these words start showing up on my blog you'll know that I've died and been replaced by a robot. Sort of like all those books that Tom Clancy keeps writing.

15 thoughts on “Let’s delve into medical studies”

  1. csherbak

    But would this be a terrible thing? I'm not an academic or have run studies, so maybe I'm out of line, BUT (asserting my cis white male privilege to pontificate on anything) ... isn't most of the work crafting the study, getting it funded and then running it? Assuming the authors of the study at least reviewed what the AI thought of it and agreed, is it all that reasonable and provide a more consistent presentation of the study methodology and summarization of the results. And potentially created and published sooner? And then make it easier to replicate because of the more 'standard' nature of the presentation?

  3. Dana Decker

    Kevin should Delve into the rich Tapestry of Large Language Models - Embrace and Explore the Dynamic that has Resonated with, Captivated, and Elevated so many, and which is a Testament to its ability to Leverage seemingly random data into something much larger that its parts.

      1. Chondrite23

        Emergent properties are those that arise from a group of items but which are not visible in studies of the individual items. A single molecule of water doesn’t have a melting point or boiling point like a collection of billions of water molecules do.

  4. Jasper_in_Boston

    I can personally attest that ChatGPT is a big fan of "delve" and "tapestry" and several others on that list.

  5. cephalopod

    NYTimes recently ran a piece that talked about how peer reviewers are having AI review the articles for them.

    I really doubt that most examples of AI use are actually cases where people read the output carefully and very selectively choose what they will add to their own papers. I remember when people said that access to the Internet meant people would fact-check things for themselves. People just don't do that. It's the same with AI. I don't think for a moment that the author of this article looked closely at his AI-generated images (neither did the reviewers or editors, apparently): https://arstechnica.com/science/2024/02/scientists-aghast-at-bizarre-ai-rat-with-huge-genitals-in-peer-reviewed-article/#:~:text=Those%20figures%E2%80%94which%20the%20authors,text%20label%20of%20%22dck.%22

    Whenever we test AI at work we end up with all sorts of stuff that is actually wrong or poorly written. Even summaries of text we hand it goes off the rails fairly frequently.

    One colleague has found that AI does a good job of creating the wrong answers for multiple choice tests, though. Hard not to get a good laugh about that one.

  7. golack

    Sounds like corporate speak...or motivational speakers on overdrive. Maybe LLM's have overdosed on "training videos".

  8. Kit

    It depends on what you mean by ‘written’. People increasingly use AI to rewrite their text, correcting the grammar and applying an appropriate tone. Perhaps that’s what we are seeing.

  13. Coby Beck

    While I have yet to delve into this dynamic tapestry of an X-post, it does resonate with me and seems to captivate and elevate your other readers as well. It's a testament to how one can leverage an audience that likes to embrace and explore data.

Comments are closed.