Bayesian Believers in a House of Pain?

Posted on by

This post is by Lizzie

A colleague from U-Mass Amherst sent me this image yesterday. He said he had found ‘multiple paper copies’ in an office and then ruminated on how they might have been used. I suggested they might have been for ‘a group of super excited folks at a conference jam session!’

This leads back to a rumination I have had for a long time: how come I cannot find a MCMC version of ‘Jump Around’? It seems many of the lyrics could be improved upon with an MCMC spin (though I would keep: I got more rhymes than there’s cops at a Dunkin’).

My colleague suggested that there is perhaps a need to host a Bayesian song adaptation contest ….

24 thoughts on “Bayesian Believers in a House of Pain?”

  1. I kind of suspect ChatGPT wrote this? I just gave it a similar prompt and it gave me this chorus instead:

    I’m a Bayesian, now I’m a believer,
    In a world of probabilities, I’m a receiver.
    From hypothesis testing to posterior delight,
    I’ve changed my mind, I see the light.

    Definitely seems in the realm of things it can pull off. And I’m sure it would be happy to write you the MCMC version of Jump Around …

    Reply ↓

    • Quote from above: “I kind of suspect ChatGPT wrote this? ”

      I recently came across a youtube video about the possible role of computers and Chat GPT-like things concerning writing song-lyrics. It reminded me of this blogpost on here.

      I like writing lyrics, and music, and stories in songs. I have wondered whether certain aspects of lyric-writing might be too much for computers to write, and one of the things I wondered about was story-telling in songs.

      Take the song “One Piece At A Time” by Johnny Cash for instance. That song depicts a story and builds up the story over several verses. I wonder whether a computer is capable of writing such lyrics to a song, and my guess is that that is not possible at this point in time. Sure, a computer may be able to use words that often go together in other songs, and can use a certain lyric-format, and can rhyme words, but I wonder whether it can put this all together in telling a coherent story.

      Another thing that might be too much for a computer in lyric-writing might be using words to tell a story without making it explicit what the details of the story are. With certain lyrics you have to read between the lines to interpret the story, which may also be something that a person is able to achieve in song-writing but a computer is not able to do.

      Reply ↓

  2. Great suggestion! I just asked it “Do you know of any songs about Bayesian statistics?” and it wrote, “Yes, there is a song called “Bayesian Rhapsody” that humorously explains the concept of Bayesian statistics. It is a parody of the famous song “Bohemian Rhapsody” by Queen.”

    We spent a while confirming this song does not actually exist (“I apologize for any confusion, but “Bayesian Rhapsody” is not an actual song. I provided it as a hypothetical example to illustrate the concept of a song about Bayesian statistics. As of my knowledge cutoff in September 2021, no such song exists.”) so then it wrote me a version ….

    (Verse 1)
    In a realm of probabilities, data caught my sight,
    Navigating uncertainties with Bayesian might.
    Prior beliefs, they clash with evidence anew,
    Bayesian inference, my method tried and true.

    (Chorus)
    Oh, Bayes, my guide, with priors I confide,
    Updating beliefs as data does provide.
    From prior to posterior, the journey I embark,
    Bayesian statistics, shining like a spark.

    (Verse 2)
    Like a puzzle unfolding, calculations I hold dear,
    Posterior distribution, the answer will appear.
    Likelihood and prior, Bayes’ theorem sings so bright,
    Magic of updating, wisdom shining in the light.

    (Chorus)
    Oh, Bayes, my guide, with priors I confide,
    Updating beliefs as data does provide.
    From prior to posterior, the journey I embark,
    Bayesian statistics, shining like a spark.

    (Bridge)
    MCMC and checks predictive, hidden truths they find,
    Bayesian modeling, unveiling the grand design.
    Quantifying uncertainty, decisions we make wise,
    Bayesian reasoning, a treasure in disguise.

    (Chorus)
    Oh, Bayes, my guide, with priors I confide,
    Updating beliefs as data does provide.
    From prior to posterior, the journey I embark,
    Bayesian statistics, shining like a spark.

    Please keep in mind that these lyrics are intended for personal use and are inspired by the structure and melody of “Bohemian Rhapsody.” If you were to use them for any public or commercial purposes, please be aware of copyright considerations.

    Reply ↓

  3. And then of course I asked for Jump Around:

    Here’s a playful take on a song about MCMC (Markov Chain Monte Carlo) to the tune of “Jump Around” by House of Pain:

    (Intro)
    Pack it up, pack it in
    Let me begin
    MCMC gonna make you grin
    Jump around, jump around
    MCMC, it’s time to get down

    (Verse 1)
    Markov chains, they’re the key
    Exploring space, oh so free
    Monte Carlo sampling, we’re in sync
    Drawing samples, link by link

    (Chorus)
    MCMC, jump around
    Burn-in steps, shake the ground
    Accept, reject, converge we will
    Sampling’s fun, just gotta chill

    (Verse 2)
    Mixing chains, making moves
    Exploring distributions, grooves
    Step by step, we’re on our way
    To the land of posterior play

    (Chorus)
    MCMC, jump around
    Burn-in steps, shake the ground
    Accept, reject, converge we will
    Sampling’s fun, just gotta chill

    (Bridge)
    From Metropolis to Gibbs, we thrive
    Swapping states, feeling alive
    Auto-correlation, we’ll fight
    To reach that independent light

    (Verse 3)
    Thinning chains, thinning out
    Decorrelating, no doubt
    Efficiency, we strive to find
    Inference game, one of a kind

    (Chorus)
    MCMC, jump around
    Burn-in steps, shake the ground
    Accept, reject, converge we will
    Sampling’s fun, just gotta chill

    (Outro)
    Pack it up, pack it in
    MCMC, we’re gonna win
    Jump around, jump around
    MCMC, it’s time to get down

    But it would not help me with including any lyrics about cops and Dunkin’ because “Writing lyrics that involve potentially sensitive topics like law enforcement and specific commercial establishments may not be appropriate or respectful to everyone.” But I did get it to include HMC:

    (Intro)
    Pack it up, pack it in
    Let me begin
    MCMC gonna make you grin
    Jump around, jump around
    MCMC, it’s time to get down

    (Verse 1)
    Metropolis, we take the lead
    Jumping states with our fiery speed
    Gibbs sampling, we’re in control
    Unraveling the parameters, soul by soul

    (Chorus)
    MCMC, jump around
    Metropolis, Gibbs, shaking the ground
    HMC, exploring the space
    Sampling methods, setting the pace

    (Verse 2)
    Hamiltonian flows, elegant and smooth
    HMC sampling, finding the truth
    Momentum and position, we intertwine
    Navigating the landscape, oh so fine

    (Chorus)
    MCMC, jump around
    Metropolis, Gibbs, shaking the ground
    HMC, exploring the space
    Sampling methods, setting the pace

    (Bridge)
    Jumping from state to state
    Exploring distributions, we won’t wait
    Autocorrelation, we break free
    Independent samples, the goal we see

    (Verse 3)
    Mixing chains, dance with grace
    Decorrelating, we embrace
    Efficient sampling, our aim so high
    MCMC, reaching the sky

    (Chorus)
    MCMC, jump around
    Metropolis, Gibbs, shaking the ground
    HMC, exploring the space
    Sampling methods, setting the pace

    (Outro)
    Pack it up, pack it in
    MCMC, we’re gonna win
    Jump around, jump around
    MCMC, it’s time to get down

    Reply ↓

    • Fine, but I was hoping it would be inspirational, encouraging someone to write something better (as I would, if I were good at such things).

      The photocopy by the way of Bayesian Believer is much older than most chatbots I think.

      Reply ↓

  6. Sunshine of your Data:

    I just got some data
    I’m gonna crunch it now
    I downloaded R
    I’ll know The Truth in about an hour

    The Truth is the sky is green
    p-value point zero fourteeeeeen

    (da da-da)
    I’ve been waiting so long
    (da da-da)
    for the light of knowing (emotional riff)
    (da da-da)
    Now The Truth is flowing from dataaaaaaaaa

    (highly emotional riff, Job’s faith in Data is restored, gathers pieces of garments and tearfully sews them back together)

    Reply ↓

    • This feels a triumph of comments help. Thank you! I see the Youtube rotates next to a ‘The Boxer’ rewrite, which I appreciated also. And thanks to Tom Metzler and others for the info (and good song ideas).

      Next up I need some help understanding dendro stats, but first I need to finish more reading.

      Reply ↓

  8. This is from one of the (famous in some circles) Valencia meetings on Bayesian statistics. I’ve never attended, but I understand that they were known for these sorts of skits and musical performances. I have no doubt that it really was written and performed by Brad Carlin!

    Reply ↓

    • From https://bayesian.org/isba-history-and-meetings/ :

      Brad Carlin has recently reminded me that at the ISBA and Valencia meetings “..one of the most popular features (after the sun and free wine) is the “cabaret” performance which traditionally takes place on the last night following the conference dinner. Acts over the years have included jugglers, magicians, jokesters, and even the occasional male striptease (the now-infamous “Full Monte Carlo”). Still, the cornerstone of the cabaret has always been the singing of new and often humorous Bayes-related lyrics to popular songs, a practice dating to the landmark work of Box…”*, namely, “There’s No Theorem Like Bayes’ Theorem, ” sung by him and Herb Solomon with everyone joining in at the first Valencia meeting. Many other “smash hits” were written and presented at subsequent ISBA and Valencia meetings by outstanding Bayesian stars. For a list, see The Bayesian Songbook, edited by Brad Carlin at https://www.biostat.umn.edu/~brad/cabaret.html. Bon reading. *Quote from Abstract, The Bayesian Songbook, edited by Bradley P. Carlin, June 14, 2001.

      (The link to “The Bayesian Songbook” appears to be dead.)

      Reply ↓

  10. I like lyrics and music!

    I have recently concluded that I like including wordplay, and humor in my scientific writing. It’s amusing to me to re-read, and I think it might also provide some sort of way to deal with frustration or annoyance concerning some topic I am writing about. I do think one has to be careful not to do this too much, or that the focus shifts from the scientific content to wordplay/humor. I dislike reading scientific papers that sound like the author really wanted to be a poet or a novelist. However, I think wordplay, humor, playing with the structure and naming of headings, etc. adds an entire new dimension to writing papers, and it might even have a useful function. I think it’s one of the things I have enjoyed the most concerning writing my recent scientific essays/papers.

    I enjoyed it so much, that I attempted to write some lyrics to some possible songs, as I reasoned that there might be some overlap between writing scientific essays/papers and lyrics in this way. There is a certain structure in most lyrics I think, and there are lots of opportunities to play with words, etc. I have found some overlap between writing essays and lyrics: I have some general idea, make a basic structure with headings or verses/chorus/bridge, and then fill in the rest with ever more specific information and details.

    Just like with my scientific writing, I reason it’s important to make sure the wordplay in the lyrics don’t interfere with the story I am trying to depict. I view wordplay/humor as something that may only become apparent when reading the paper for the Nth time, or listening to a song for the Nth time. Or maybe it isn’t noticed at all by many of the people who read it or listen to it.

    Reply ↓

    • I’ve tried to write some lyrics fitting to this post. No robots or computers were involved (as far as I’m aware), so I hope it’s okay to share the following:

      Psych10none

      I thought you might be the one,
      like a light in the darkness,
      like something to build upon
      But of these things,
      I now feel you are none

      Most of what you told might be a lie,
      contributing to more darkness,
      instead of illuminating the sky
      Concerning these things,
      I now only ask why

      (Chorus)
      From Psych101 to Psych10none
      The more I learned, the less I understood
      But perhaps it is darkest before the dawn
      Class is dismissed, class is done

      (Bridge)
      There are times when I think,
      that I wasted a lot of my time
      But just before I start to sink,
      I remember that I have to climb

      Maybe you weren’t the light,
      but your darkness made clear
      that other things are bright
      and can better be used to steer

      In that way I can now see,
      that a loss might be a win
      In some way you did help me,
      perhaps the light shines within

      (Chorus)
      From Psych101 to Psych10none
      The more I learned, the less I understood
      But perhaps it is darkest before the dawn
      Class is dismissed, class is done

      Reply ↓

  11. I’m so pleased this group has discovered my foundational role in Bayesian cabaret ;)

    It’s hard to imagine that things this fun ever happened at conferences. But they did, and for many years. Unfortunately, ISBA has killed the Bayesian cabaret (and pretty much all other fun things) at their meetings, so it’s not clear when performances like this might happen again. But it would be great if someone would resurrect them. It’s clear from this page that there is still a lot of Bayesian song parody talent out there!

    Thanks for finding the Bayesian Songbook in the wayback machine; I should repost that to my website. For now, just type “Bayesian cabaret” in the YouTube search window and be astonished at what you find: jugglers, tumblers, singers, and something called “The Trumpet Sketch” that you absolutely will not believe (even I didn’t, at the time).

    I’d also like to mention that besides the Imposteriors (my cabaret band with Mark Glickman, Jennifer Hill, Mike Jordan, and Don Hedeker) there was a student band at NC State called the Fifth Moment that wrote several very good songs, performances of which are also out there on YouTube.

    Reply ↓

    • * Unfortunately, ISBA has killed the Bayesian cabaret (and pretty much all other fun things) at their meetings, so it’s not clear when performances like this might happen again. *

      Awkward

      Reply ↓

Leave a Reply

Your email address will not be published. Required fields are marked *