Wednesday, January 10, 2024

SearchResearch Challenge (1/10/24): How do you use LLMs in your SearchResearch?

 It's a New Year! 

P/C Dalle3. [an evocative picture of data, data tables, line charts, histograms]

Let's reflect on what SearchResearch is all about... 

I started this blog back in January, 2010 with About this blog--Why SearchReSearch?  That was 5,093 days ago.  

So far, there have been 1,374 posts, with 4.75M views and 13,700 comments.  We're running around 40K blog views / month, and that doesn't include the various syndicated versions of the blog.  If you include those, I'm Fermi Estimating the readership at 50K /month.  That's a decent number (roughly 1,660 people / day).   

As longtime SRS readers know, one of my goals was to use the blog as an effective prompt to write a book.  That book--The Joy of Search: A Google Insider's Guide to Going Beyond the Basics--came out in September 2019 and has done reasonably well. At least well enough to be translated into Korean and Chinese, as well as go into a paperback edition.  

As my Regular Readers also know, I'm working on another book ("Unanticipated Consequences") which I'm trying to finish up in the first part of this year, 2024.  

Writing a regular blog is a serious investment of time and energy, at least it is for me.  On average, I spend around 4-8 hours each week bringing you the most interesting SearchResearch tidbits I can find.  

But it's a big time commitment, and I'd like to use those hours to work on the new book.  

So I'm going to try an experiment--scaling back a bit, trying to do each blog post in two hours or less.  

To do that, I'm going to try to involve you Regular Readers a bit more in the search for insights.  

As we've discussed, the advent of Generative AI is promising to radically change the SearchResearch process.  It has the ability to give us new insights quickly (an example), and also to generate convincing-sounding nonsense with ease (an example).  

I know that many of you are using ChatGPT or Bard or Llama to answer questions, so I want to tap into the collective intelligence that SRS Regular Readers can bring to the discussion.  

I've asked this before in general (How can we use LLMs to search even better?) and for the specific case of medical searches (How might we best use LLMs for online medical research?).  But if there's one thing to know about LLMs and Generative AI in general, it's that this field is changing fast--really fast--fast as a flash rifling down the electric blue gun barrel into a sea of new-age synth-pop psychedelia.  


Our SearchResearch Challenge of the week is this is to revisit this: 

1.  How have you found yourself using an LLM system (or more generally, any GenAI system) to help solve a real SearchResearch question that you've had?  


I've told you before that I've used ChatGPT to help me with some data table manipulation. I've solved a few problems in a minute that would have taken me an hour or more to do with my regular programming skills.  Those are huge savings. 

I've also gotten LLM help in finding answers to SRS questions that I couldn't figure out how to frame as a "regular" Google search.  Bard has answered a couple such questions for me (which I then fact-checked immediately!).  

We all want to know what you've found to be useful.  What has worked for you in the past few months?  

Alternatively, what has NOT worked out for you?  

We're all ears here at SRS! 

Keep searching. 

P/C Duet AI (Google Slides) [kangaroo on top of piles of text listening ]


  1. Happy New Year, Dr. Russell.

    I don't use yet LLMs. I tried Bard sometimes and didn't find it useful. My searches are too easy maybe. I have seen these tools being used in medicine thanks to Mr. Topol MD. and also with you. That is a good use of technology. Others allegedly predicted game results or even best players in a tournament. That, I think is just fun and without any science.

    About Challenges, I have been thinking and trying without any luck, as an example, to find a way to discover new videos on YouTube that have multiple options on subtitles.

    Other was to find how people in the past used cabañuelas and to read if they worked. Nothing to get my answers at the moment.

    Maybe LLMs could make weather predictions? I'll try that.

  2. "fast as a flash rifling down the electric blue gun barrel into a sea of new-age synth-pop psychedelia. "
    enigmatically specific/esoteric
    I B Skippy - too many kangaroo holes to go down
    almost 12 hours -
    kinda like Lake Geneva…
    Niro Knox
    … no, the LLMs (large 'lectric music) were not useful or time- savers in this how could they be?

    1. an unnamed LLM made me find this - not just my love of bagpipes…

  3. As a retired librarian I use LLMs to provide a basic understanding of new topic or to provide further insights into a topic I am researching and have some knowledge of. One of the features that helps me in researching is the conversations, because it requires me to think what else I should consider in my reference query. But always having that librarian mindset, I do check multiple search tools since often one finds different and helpful information to complete one’s research query and these can include general search engines or research sites. This also helps to avoid misinformation from LLM hallucination by comparing one to another. One unique LLM tool is Poe where you can search multiple AI sites and tools.

    1. Can you query multiple LLMs at once? (I though Poe did one or another, but didn't blend the results together.)