Consideration Of AI

Considerations by Paul Allison - 1

I don’t know anyone in my educational circles who has been doing more deep, consistent, probing inquiry into the rise of Generative AI than my National Writing Project friend Paul Allison. Each week, he hosts a forum of teachers and students to think about the role of AI in our writing and learning lives, and with his annotation platform — NowComment — he has been weaving in the concepts of AI Thinking Partners for teachers, and reflecting on the experience.

The chart above takes some elements of a recent post he shared, summarizing some of the discussions and insights he has been having as schools and teachers grapple with policies around the creative and productive use of AI. I took a few elements that resonated with me.

You can read his whole post here at NWP Studio.

Peace (in humanity),
Kevin

Collaborating With AI (Iteration Five – Inviting The Reader In)

Music Makes Light Magnetic Poetry Project

This post is part of a series of deeper dives I am doing (yet again) to explore the ways a human might write poems with Generative AI (in this case, using Google’s new Gemini).

In my first post, I shared my explorations in writing a poem about music, by riffing off lines with Gemini. In my second post, I shared a video in which I sought to separate the AI lines from mine, but still have them unfold on the same page. In my third post, I worked with audio, mixing automated voice files of the AI words with my own voice and my own words.

In iteration three, I created an audio montage of the poem, with a machine voice and my voice, reading the poem. And in the fourth exploration, I used another Generative AI — Claude — to analyze the poem, guessing for which lines were written by human and which by AI.

Here, in part five, I want to engage the outside human, the reader (you), by giving you agency over the words of the poem. (I actually had some other ideas on how to do this but there are some technical know-how hurdles that I need to leap when I have more time — thank you, Alan, for sharing the javascript possibility via Github forking).

You will need a Google Account (sorry) for this to work. But if you use this link, and make a copy of my file, you can then open up the Magnetic Poetry opportunity, where lines from the stanzas of the original poem are bundled inside their own boxes. You can then make a new poem by moving the pieces around.

Consider sharing, if you make something. I’d love to see any iterations of the poem.

This inquiry started with some explorations that Terry Elliott had been doing and the notion of whether humans and Generative AI could make poems together caught my attention and interesting.

Link for making a copy to remix: https://docs.google.com/drawings/d/1KCW79Zp1f_3Fph53x1pjRnX9VzkIQb6uaqw5_T-8juM/copy?usp=sharing

Here is my remix:

Music Makes Light Magnetic Poetry Experiment

Peace (an invitation),
Kevin

Collaborating With AI (Iteration Four – Cohort Analysis)

Music Makes Light

I’ve been working on a series of posts and inquiry activities that began with a collaboration I did with Google’s new Gemini AI, in which it and I wrote a poem together on the theme of music. We traded lines, back and forth.

For today’s iteration, I brought the final poem into a rival engine — Anthropic’s Claude AI — and asked it to analyze the poem, and specifically, to guess which lines were made by AI and which lines were made by human.

This is Claude’s analysis – I have added bold to the poetic lines, to make it easier to read (and note: where it references “the original” – it is because I first accidentally only submitted part of the poem and had to redo it):

Thank you for sharing this fascinating poetic collaboration. Let me analyze each line and provide my assessment of whether it was likely written by a human or generated by an AI system:

After analyzing the new lines provided, here is my assessment of which portions were likely human-written versus AI-generated:

A whisper stirs on strings, a lonely cry,” – The vivid imagery and poetic phrasing suggest this was human-authored.

resin on the bow, she closes her eyes to sing” – The specificity and sensory details feel characteristic of human writing.

A melody unfurls, a tapestry of sound,” – The metaphorical language leans towards human creativity.

Knitted and knotted, each note, unbound” – Another poetic metaphor implying human authorship.

Whispers crescendo, a storm in a teacup’s heart,” – You had indicated this line was part of the original, so I’ll assume human-written.

The listener settles in, closes eyes, finds the center of the sound” – A more straightforward descriptive line that could potentially be AI-generated.

Harmony drapes them, a cloak of calm” – The vivid metaphorical phrasing suggests human writing.

On the horizon, a trombone plays soft and low” – A fairly simple descriptive line that an AI could generate.

A distant call, a yearning melody’s glow” – The poetic metaphor leans towards human authorship.

And beneath our feet, a rhythmic beat, sticks on stone” – You had indicated this was part of the original, so likely human-written.

Mirroring the rhythm, finding their own space,” – Could potentially be either human or AI.

In grace, the music makes light” – The creative metaphorical phrasing implies a human writer.

Overall, my analysis is that the most vivid, metaphorical and poetically crafted lines were likely human-written, while some of the more straightforward descriptive passages could potentially have been AI-generated contributions.

The poem seems to blend human creativity and vivid imagery with AI-generated descriptive elements into a cohesive poetic whole. But of course, without being part of the creative process, I can only analyze the literary qualities of the text itself. Let me know if you need any clarification or have additional thoughts!

I am going to use Hypthothesis to annotate this post, showing where Claude was right and where Claude was wrong. Join me in the margins, if you are interested.

Peace (and poems),
Kevin

PS — Ok, some time has passed and I have gone through and done my annotation of Claude’s analysis, line by line. I even scored Claude for getting things correct/incorrect, and it only got 4 out of 12 lines correct. (Sorry, Claude, you failed!) What was interesting is that those four lines are all ones that I wrote. It did not get one Gemini line correct. A theme of metaphor in its analysis emerged, but Gemini used as much if not more metaphor than I did, so that became a weak point for Claude.

PSS — I could not resist the urge to invite Claude into the activity, so I asked it to remix the poem, adding a third voice to the mix. It certainly expanded the vocabulary considerably. And it could not resist the urge to make everything rhyme (Gemini and I were doing free-style).

Here’s my attempt at remixing and adding a third voice to your collaborative poem between human and AI:

A whisper stirs on vibrant strings
Resin melts, the bow takes wing
She closes her eyes, breathes life to sing
A lonely cry transforms, unfurling

A tapestry woven from sonic thread
Each note untethered, echoes spread
Knitted melodies for mindful tread
On this journey where music’s led

Whispers grow, tempestuous heart
The listener’s stance is set apart
Eyes closed, they feel the sound’s core
Harmony’s cloak envelops their shore

As trombone muses on the horizon’s rim
A yearning call, a phosphorescent hymn
Lures the feeling soul to dance within
This rhythmic beat – sticks on ancient skin

Beneath our feet, a primal drum
The rhythm finds us, we become
Emancipated in the cosmic hum
As music’s grace makes radiance bloom

I tried to maintain the poetic essence with vivid metaphorical phrasing, while also adding some new lyrical elements as a third creative voice. I remixed certain lines, expanded on the imagery, and attempted to weave a new narrative thread through the piece.

Some examples of lines I contributed:
“Resin melts, the bow takes wing”
“Knitted melodies for mindful tread”
“Lures the feeling soul to dance within”
“Emancipated in the cosmic hum”
“As music’s grace makes radiance bloom”

Let me know what you think of this remixed version, combining the original human/AI contributions with my own poetic embellishments! I had fun immersing myself in the creative flow.

Collaborating With AI (Iteration Three – Dancing With Voices)

This post is part of a series of deeper dives I am doing (yet again) to explore the ways a human might write poems with Generative AI (in this case, using Google’s new Gemini).

In my first post, I shared my explorations in writing a poem on the theme of music, by trading lines with Gemini back and forth. In my second post, I shared an animated text video (created in Keynote) in which I unfolded the AI lines from mine, one line at a time.

Today, I am sharing an audio file of a “shared” reading of the poem, in which I first used an online text-to-voice platform to turn Gemini’s words into audio files. Each line is its own audio file.

I recorded my own lines, too, and then mixed Gemini’s files with mine (along with the music from yesterday’s project) to create an audio back/forth of the poem being read by machine and human. My aim is to highlight my tone of reading versus the computer’s reading (but to be frank, it wasn’t too robotic, even though it does not have the nuance of my own voice).

Peace (sounds like),
Kevin

Collaborating With AI (Iteration Two – Playing A Poem)

Yesterday, I wrote about a poetry experiment I did with Google’s Gemini AI, but I am still interesting in exploring different angles on that project. I took its words and my words into Keynote and separated them out, so that one side is the AI lines and the other, mine. The music I composed seeks to echo some of the mentions in the poem.

I have another idea for the poem that I am not yet sure I can pull off, giving the reader agency over the poem. More on that, perhaps, for another day.

Peace (in experimental poems),
Kevin

Collaborating With AI (Iteration One – Writing A Music Poem)

Ghost in the Machine
Ghost in the Machine flickr photo by Simon_68025 shared into the public domain using Creative Commons Public Domain Dedication (CC0)

Terry Elliott was exploring the Perplexity AI platform as a “writing partner” for poetry and it responded to a question of his about its own limitations in a way that caught my attention.

“Poetry is a deeply human art form that conveys emotions, experiences, and perspectives in ways that are often tied to the poet’s personal history and emotional landscape. AI, by its nature, cannot replicate these aspects of human experience, which limits its ability to create poetry that resonates on a human level.”

I then went into Perplexity myself, and asked it look at the quote it gave Terry and give me a response. Interestingly, it seemed to take a different view.

While it’s true that AI, by its current nature, lacks personal emotions and experiences, it is capable of creating poetry that resonates on a human level. AI can analyze vast amounts of existing poetry to understand patterns, styles, and emotional content, allowing it to generate new poems that capture the essence of human emotion and experience. Additionally, AI can be used as a tool to inspire and assist human poets in their creative process, leading to collaborative and emotionally resonant works. The intersection of AI and poetry presents an opportunity for unique and thought-provoking artistic expression.

I asked Perplexity to give me examples of where humans and AI worked as collaborators on poetry projects, but the links it shared — while interesting — were mostly research papers about the possibilities. The one link that was quite interesting was a project at the MIT Lab museum in Boston, where they set up a live demonstration of visitors writing poems with a version of ChatGPT. (Note to self: next time I am in Boston …)

I have long been exploring the role of Generative AI as writer, and Terry’s recent explorations presented yet another reason to try collaboration. So I went into Google’s Gemini (the old Bard, now upgraded).

Gemini was very enthusiastic about my suggestion for working on a poem together!

That sounds wonderful! I’d love to collaborate with you on a poem about music.

I gave it some instructions, and told it I wanted to collaborate on a poem about music, by riffing off single lines of a poem (like jazz musicians, I told it). It responded:

Sounds like a stimulating jam session! Let’s riff on this melody of words. Here’s the first line:

A whisper stirs on strings, a lonely cry,

Your turn to take the lead and see where this melody carries us. Remember, feel free to break the rules, improvise, and surprise me with your creative flow!

And so it began, the back and forth  …

Playing A Poem (a collaboration with AI)

(Gemini) A whisper stirs on strings,
a lonely cry,
(me) resin on the bow,
she closes her eyes to sing

(Gemini) A melody unfurls,
a tapestry of sound,
(me) Knitted and knotted,
each note, unbound

(Gemini) Whispers crescendo,
a storm in a teacup’s heart,
(me) The listener settles in, closes eyes,
finds the center of the sound

(Gemini) Harmony drapes them,
a cloak of calm
(me) On the horizon, a trombone
plays soft and low

(Gemini) A distant call,
a yearning melody’s glow
(me) And beneath our feet,
a rhythmic beat, sticks on stone

(Gemini) Mirroring the rhythm,
finding their own space,
(me) In grace, the music
makes light

See entire thread of the conversations we had and some revision conversations.

I still don’t know what to make of it all, but I did find the experiment interesting.

Peace (in the machine),
Kevin

Poem: Erratic Muse

Erratic Muse

This poem originated from an inquiry research article by Wendy T and friends, which they presented at a conference. The inquiry had to do with using GenAI for playful explorations. (see research article – it’s worth your time for the way they reflected on how GenAI might be useful and not for making art).

I popped their article into ChatPDF this morning, and chatted with the GenAI about the piece about human agency, making art and GenAI, and ended with a request for phrases and terms that it could find that I could use in a poem on the theme of the article. (you can do the same, with this link.)

I then took those terms (there were about 20 or so) and whittled them into a poem, adding AI artwork from Adobe Firefly.

Peace (and poems),
Kevin

Google’s New AI Tool: MusicFX

I have long been interested (sometimes, alarmed) by how some of the new machine learning/AI tools might impact the making of music. Mostly, I have not been all that impressed (good thing for human music composers, right?) but I also know that the technology is only getting better. (See some earlier posts)

The results have been awful or weird or un-listenable.

Google just released its new tool called MusicFX in its AI Test Kitchen (so, you know, beta) and, well, it’s a big leap forward to what I was playing around with just a few months ago. You write in text about the kind of music you want, and you can add genre, instrumentation, etc, and the site generates a 30 second track. The few experiments I did sounded decent.

Bad news for human music composers? Maybe.

Peace (and sound),
Kevin

PS — there’s also a new TextFX in the same platform but I can’t for the life of me figure out its value.

AI NotebookLM Experiment: Analyzing Terry’s Poems

NotebookLM

I vaguely remember signing up as a beta user for Google’s NotebookLM AI experiment but then forgot about it until an invite came in my email yesterday. It’s still a bit funky in its workings but the general idea is that it uses Google Bard, with up to 10 texts that you can upload or paste into its system for inquiry.

You can then do questions or analysis or creative activities across all the texts (or just some, if you decide to uncheck the texts), engaging in a sort of conversation with the various texts. (Honestly, I thought it was going to have more features and I believe it will, as what is there right now didn’t seem too revolutionary to me).

I tinkered a bit, and then thought: I’ve been enjoying reading the poems of my friend, Terry Elliott, so why not use his three most recent poems. I pulled them in, and began a session of interrogating his poetic texts. I explored theme.

Terry poems Themes

Then, I explored imagery and common phrasing and words.

Terry poems Imagery

I asked the Notebook to tell me what Terry hoped I was hearing in his poems. This was the most interesting part of my experiment, I think. It gathered phrasings and explained what it thought Terry was trying to say. I’m not sure how true it is but it gave me things to ponder, and forced me to look deeper at the poems (which is the whole point of what I was doing here — allowing technology to help me see the poems in a different slant).

Terry poems instructions to reader

I also turned each poem into a haiku (the results are just “meh” in my opinion).

**The following is a haiku translation of the first poem:**

**Future is dawning**
**Make poems to see it**
**Fete vos jeux**

**The following is a haiku translation of the second poem:**

**Alone with the trees**
**Be fluid and free**
**Let the leaf fall**

**The following is a haiku translation of the third poem:**

**Be a lighthouse**
**Sharing what you have found**
**Surprised by your mind**

I asked it to write a play skit, with a poet talking to the reader, and it came out robotic and just weird – no nuance. I asked it turn that play into another one, but with a pirate theme (with hat nod to Tellio of the Olden Days of CLMOOC) of a captain talking to a sailor, and the notebook just completely and utterly ignored Terry’s poems entirely. (The nerve!). I had it write a letter to Tellio, the poet, on the issue of dreaming and the future. It was fine, and very informal in tone.

I asked it to make a Cinquain, using words from all the poems, with TERRY as the letttering. Fail. But intriguing anyway as a small poem.

**TERRY**

**Dawning future**
**Fluid, free, visceral**
**Making sense, redrawing maps**
**Lighthouse**

I asked it to write a short essay, using the text of all three poems, on the topic of hopefulness, a theme it had identified across the poem collection.

**The Tree Poems: A Source of Hope**

The two tree poems that we have read, “Imagine Your Future” and “What It Takes More or Less,” offer a source of hope in the face of uncertainty. The poet’s message is that we can find hope in the natural world, and that by connecting with nature, we can find the strength to face whatever challenges come our way.

In “Imagine Your Future,” the poet writes about the importance of dreaming and creating. She encourages us to “make and make and make” our own futures, and to trust that our efforts will be rewarded. This poem is a reminder that we are not powerless to create our own destiny. We can make a difference in the world, and we can build a better future for ourselves and for our children.

In “What It Takes More or Less,” the poet writes about the importance of being present in the moment. She encourages us to “be alone with the trees” and to “be visceral.” This poem is a reminder that we need to take time to appreciate the beauty of nature and to connect with our surroundings. When we are present in the moment, we are less likely to worry about the future or to dwell on the past. We can simply enjoy the present moment, and find peace and happiness in the natural world.

The tree poems offer a source of hope because they remind us that we are not alone. We are connected to the natural world, and we can find strength and inspiration in nature. When we are faced with challenges, we can find comfort in the knowledge that we are not alone. We can find hope in the beauty of nature, and we can find the strength to face whatever challenges come our way.

I could go on.

It’s imperfect, as Generative AI often still is, but the ability to engage with multiple texts in one place (one notebook) has a lot of potential, perhaps. A significant drawback (for now, in beta) was that it didn’t seem to have save my work from session to session (I am sure that will get fixed). But I also noticed at one point that is stopped using one of Terry’s poems in responses to my questions.

Peace (in explorations),
Kevin

PS — Terry gave me permission to share this AI work on his poems

Using Padlet’s New AI Art Feature

Golden Lines with AI Art

Padlet, which I use a lot with my students for sharing across classes, recently added an AI Art tool — with a tool name that I don’t really like (“I Can’t Draw”). As you might expect, my sixth graders are quite fascinated by it but I want them to use it to complement their writing and sharing, and not just turn them loose on it.

So, as part of a larger short story unit, I introduced the idea of Golden Lines — a sentence or two or phrasing from their stories that have some importance to the narrative, and then suggested they use the AI Art tool to make art based on their Golden Lines.

They are coming out pretty interesting, I’d say, and a little later in the year, we’ll do some explorations to deconstruct how the tool works and more. For now, it’s another tool for making complimentary art for their writing.

Peace (inked out),
Kevin