As every aspiring computational linguist on this planet, I’ve been playing a little bit with ChatGTP. While I’m not tremendously impressed with ChatGTP’s answers on any slightly-less-than-straightforward questions with verifiable answers, asking it to do something that requires some creativity occasionally results in a display of a mixture of genuinely fun ideas and absolute incapacity to make good use of them in a coherent text. So, I decided that this weekend we should try to cooperate on writing a silly poem like the ones I sometimes post on this blog. What follows is the result of our joint efforts.

Conference or journal, which way to go?
Let’s see the differences, row after row.

A conference paper is short and concise,
Meant for the event, with less time to revise.
With results preliminary, work sill ongoing,
It gives you a chance for networking and showing.

What lands on a journal’s in-depth and refined,
Detailed and polished, with printing in mind.
It’s subject to long reviews, usually blind,
And ends up more visible, easier to find.

So, conference or journal, which one to go for?
Choosing’s on you now that you know more.
Both are of value in their own unique way,
So just pick the one that fits best today.

The process

It all started with a relatively straightforward question I asked ChatGTP:

harisont: What is the difference between conference and journal publications?

As expected, I got one of ChatGPT’s typical long, plain, but otherwise fine answers:

ChatGTP: Conference and journal publications are both important ways to share research findings with a wider audience, but there are some key differences between the two.

A conference publication is typically a paper or abstract that is presented at a conference or other academic event. Conference publications are usually shorter and more concise than journal articles, and may be more focused on preliminary findings or work in progress. Conference papers are often presented in a brief presentation or poster session, and the author may have the opportunity to discuss their work with other attendees.

In contrast, a journal publication is a more in-depth and polished version of a research study that has been submitted and accepted for publication in a peer-reviewed journal. Journal articles are typically longer and more detailed than conference papers, and may be subject to more rigorous review and editing processes. Journal articles also tend to be more permanent and widely available than conference papers, as they are typically published online and/or in print and are often indexed by scholarly databases for easy access and citation.

Another key difference between conference and journal publications is the time frame in which they are published. Conference papers are often presented at events that occur on a specific date, while journal articles can be submitted and published at any time, and can be read and cited by scholars years after their initial publication.

In summary, conference publications are shorter, more preliminary versions of research findings that are presented at academic events, while journal publications are more polished, detailed, and permanent versions of research studies that have been submitted and accepted for publication in a peer-reviewed journal.

At this point, I asked ChatGTP to give me the same answer in rhymes, and it produced this:

ChatGTP:
Conference or journal, which way to go?
Let’s see the differences, rhyme by rhyme, row by row.

A conference publication is short and concise,
Presented at events with less time to revise.
It may be preliminary or work in progress,
But it’s a chance to network and discuss.

A journal article is in-depth and refined,
Polished and detailed with a more permanent design.
It’s subject to rigorous review and editing,
And widely available for scholars’ reading.

Conference papers are presented at a specific time,
While journal articles can be read years down the line.
The former may be more focused on preliminary finds,
While the latter presents more developed designs.

So, conference or journal, which one to choose?
It depends on the study and the path you want to pursue.
Both are valuable in their unique way,
So choose the one that fits best today.

It’s not very good, but it’s not too bad either.
The second verse is too long, but other than that I quite like the opening. The meaning of the answer in prose, and even its structure, are more or less preserved. There is some actual rhyming, some things that look like rhyming (cf. progress/discuss), and some other that sound at least a little bit similar to each other (cf. choose/pursue). The division into stanzas is perfectly logical, and I like that equal space is given to defining conference papers and journal articles. The fourth stanza seems a bit useless, but the closing, apart from another disproportionately long verse, is cool, especially the part where ChatGPT addresses the reader, and the ring composition is on the spot.

The main problem is the metric. The poem looks like something that should have some rhythm to it, but even if I tried to read it out loud as best as I could, but there’s no way I can make it work. This was my best attempt:

Disregard my mispronounciation of the word pursue. Even without that, this doesn’t sound like anything more than a first draft.

So, I started proposing changes. Some changes to make were obvious to me, and ChatGTP had no problems satisfying requests like:

harisont: I think the second verse is too long. Can you remove “rhyme by rhyme”?

or

harisont: Replace “publication” with “paper”

or even

harisont: It depends on the study and the path you want to pursue. -> It shure depends on what you want to pursue

where ChatGPT even corrected my typo (“shure”) in the revised version of the poem.

I was getting lazier and lazier with explaining the edits I wanted ChatGTP to make, and I was pleasantly surprised when ChatGTP understood that

harisont: their OWN unique way

meant “add the word own (lowercase) between their and unique”, even though that didn’t work very consistently:

harisont: so YOU choose
ChatGTP: I’m sorry if my previous response wasn’t helpful. As an AI language model, I cannot make decisions for you. Choosing whether to submit a paper to a conference or a journal depends on various factors such as the purpose of your study, the stage of your research, and your career goals. You will need to consider these factors and make an informed decision based on your individual circumstances.

Sometimes, however, I wanted some more sophisticated help, which ChatGPT wasn’t really able to give me. When I asked

harisont: “Progress” does not rhyme with “discuss”. The accent is different. Can you find some better replacement for the latter?

It replaced the whole

ChatGTP:
…
It may be preliminary or work in progress,
But it’s a chance to network and discuss.
…

with

ChatGTP:
…
It may be preliminary or a work in motion,
But it’s a chance to network and foster communication.
…

where rhyming kinda works (even though the expression “work in motion” doesn’t sound too convincing), but the metric is way off.

Not to mention that ChatGTP would occasionally get creative and make some unsollicited changes. At some point, “pursue” became “peruse”. I wasn’t entirely happy with that:

harisont: Who told you to change “pursue” to “peruse”!? ChatGTP: I apologize for the mistake. It seems that I misunderstood your previous message. I will make sure to pay closer attention to your requests in the future. Thank you for bringing this to my attention.

Seeing that I couldn’t really trust ChatGTP with any nontrivial editing of the poem, I started editing the text on my own asking it for suggestions, like:

harisont: What rhymes with “read”, in this context?
ChatGTP: In this context, you could consider using words like “spread”, “shed”, or “led” as a rhyme for “read”. …

…which is disappointing, because this words are not only semantically unrelated, they also don’t rhyme with “read” at all, unless you interpret that as a past tense, which was not what ChatGPT was doing:

ChatGTP:…For example:
“It’s subject to rigorous review and editing,
And widely available for scholars to be reading.” …

(fair enough)

ChatGTP:…or:
“It’s subject to rigorous review and editing,
And widely available for scholars to spread.” …

(what the hell!?).

At some point, ChatGTP finally forgot about the poem and started answering my questions without any in-context examples, again with mixed results:

harisont: rhymes with progress?
ChatGTP: Here are some words that rhyme with “progress”:

  • Congress
  • Address
  • Success
  • Transgress
  • Oppress
  • Egress
  • Express
  • Redress
  • Digress
  • Confess

harisont: Which ones rhyme with the NOUN “progress”?
ChatGTP: My apologies for the confusion. Here are some words that rhyme with the noun “progress”:

  • Success
  • Access
  • Distress
  • Duress
  • Excess
  • Finesse
  • Impress
  • Oppress
  • Redress
  • Stress

I ended up continuing editing the poem my myself. ChatGTP might have felt a little bit offended, but it tried to hide it:

harisont: I think I finished touching up the poem
ChatGTP: Great! I’m glad I could help. If you have any more questions or need further assistance, feel free to ask.
harisont: Don t’ you want to read it?
ChatGTP: Sure, I’d love to read the updated version of the poem! Please share it with me.

…And in the end it seemed to like it:

ChatGTP: Great job! The poem flows well and effectively captures the main differences between conference and journal publications.

Well, thanks ChatGTP! To be fair, your poem was already pretty effective, but I do think that my version, while it isn’t perfect, flows better:

But there was one more question I needed to ask ChatGTP:

harisont: Can you propose a few titles?
ChatGTP: Sure, here are a few possible titles for the poem:

  • Choosing Between Conference and Journal Publications
  • The Pros and Cons of Conference and Journal Papers
  • Navigating the World of Scholarly Publishing: A Guide to Conferences and Journals
  • Deciding Between Conferences and Journals: What You Need to Know
  • Weighing Your Options: Conference Papers vs Journal Articles

I found it hilarious that ChatGTP’s propose a scholarly-sounding titles such as “Navigating the World of Scholarly Publishing: A Guide to Conferences and Journals”, which I ended up choosing after seeing that “playful” options just seemed blog post titles. When I reminded ChatGTP that we were writing a poem, things got slightly better (with a solemn “In Search of Publication” as option 7), but in the end we agreed on my first choice:

harisont: I think I’ll go for a paper-like title and make a fool of everyone.
ChatGTP: That could be a fun twist! How about “Navigating the Academic Landscape: Choosing Between Conference and Journal Papers”?
harisont: That’s exactly the one I was thinking of!

Some quick final considerations

  • Was this fun? Yeah.
  • Was this more fun than writing a poem of my own? No, but it was interesting. It gave me a better understanding of what I can expect from ChatGTP.
  • How long did writing the poem take? About 1.5 hours.
  • Is that less of what it normally takes me to write one of my silly poems? Not necessarily, but I have never tried to write a poem on a random subject chosen by someone else. I usually already have some idea that I think is promising when I start writing.
  • Was ChatGTP a good poetry slam partner? It did provide an interesting first version of the poem, but the interactions following that were not too satisfying. It was able to perform some simple edits, but they were all things that I could have more easily done by myself: explaining ChatGPT what to do takes longer than actually removing a stanza or replacing a word in a text editor or on a piece of paper. Anything more sophisticated wasn’t really possible. Some titles ChatGTP proposed were fun, but I don’t suppose they were meant with irony. In fact, they were not meant at all: ChatGTP is yet another stochastic parrot.