One Step Further: Collaborating with AI Open

Over the weekend, I wrote about using Text to Transformer to start a poem and see where the AI Open-infused text generator — Talk to Transformer — might take my words.

Then, I started to think about how to find a poem inside the text generated by another poem. Could I surface something from inside of something else, inspired by something else altogether? Another nested poem? I’d find out.

Here’s what I did (in case you want to ever do your own):

  • I went into Google Slides (but any slideshow program would work because when you move across slides, it looks like animation) and began to cross out words (blackout poem style).
  • Then I removed the excess words (I cheated, by turning the font color the same color as background, so white text against white background is no longer visible; otherwise, it would have a long formatting exercise of adding spaces where words had been).
  • Finally, I pulled the remaining, revealed text into another poem. I used transitions and animations to make the process more visible in the slides (the whole thing is as visual hoax, really, using different slides layered on each other to seem like the text is being animated).

Sort of odd. I like that kind of weird writing and weird writing processes.

Peace (in the poem),
Kevin

 

Writing Collaboration with OpenAI: Context and Constraints

What happens when you hand off your poem to a “modern neural network”? Something strange, with a hint of interesting. I was using a site called Talk to Transformer, which is built on the back of some neural network mapping of OpenAI and which is designed to complete your text, using its signifiers and databases.

The site explains that it is:

… an easier way to play with OpenAI’s new machine learning model. In February, OpenAI unveiled a language model called GPT-2 that generates coherent paragraphs of text one word at a time … While GPT-2 was only trained to predict the next word in a text, it surprisingly learned basic competence in some tasks like translating between languages and answering questions.

So, of course, I could not resist feeding it some words to see what would happen, starting the lines of a poem about context and constraints, and in the image above, you can see what it spit out for me. There is something beautiful surfacing there, in the juxtaposition of my poem starter and its story extension, although I am at a loss to really understand how it made the leap from my words to its text.

For example, the point of view shifts from third person to first person, and suddenly, the narrator is talking of their mother’s love (or lack of) in a world fallen apart. But look at the last three lines it generated … it’s almost like the start of something else altogether, maybe a new poem generated by human hand … Maybe the game turns to me to continue onward with the AI’s idea ..

I am what I am when I’m no longer
something that mustn’t be forgotten…
a person so beautiful

So remember me; you must remember us,
as I remember this wasted Earth
when love was nearly lost

and all we had left to hold was each other,
in the days after fallen trees
and warming seas

I still carry the bones of my mother,
that which the soil would no longer hold:
I am young; I am old

The image is a layered gif that I made in Lunapic because I wanted to do something more with the writing. I purposely added non-digital writing tools to contrast the use of AI to make a piece of writing.

Peace (in texts, transformed),
Kevin

 

Do Algorithms Dream of Improvisation?

Here is a convergence of two texts in a feldgang — first, a shared reading of The Art of Is with other CLMOOC friends and then, an article in this week’s The New Yorker by Hua Hsu entitled “Machine Yearning.” In The Art of Is, I am in a place in the text where writer Stephen Nachmanovitch is exploring improvisation and mistakes, and how one (mistakes) often lead to the other (improvisation) to make art. In The New Yorker piece, Hsu focuses in on artist Holly Herndon’s work to create music through a computer-generated voice program she calls Spawn, training it to sing to Herndon’s music on her new album, PROTO.

Hsu then connects Herndon’s work with computers and algorithms to some other movements on the music landscape, including rapper Lil Nas X’s successful maneuver to manipulate the algorithms of pop music charts to create “Old Town Road” (an analysis of top charted songs led the making of this song) and it, of course, has reached the top of the charts.

Further, Hsu notes that a company called Endel is now developing music created only by algorithms to “personalize” a space or environment that, according to Hsu, take ” … into account everything from their (listener) heart rate to the weather and thje time of day.” Endel got signed to Warner.

And Hsu notes that jazz pianist Dan Tepfer is releasing a new album soon called “Natural Machines” in which he plays duets with computer algorithms, in which the algorithms, according to Hsu, are “… designed to respond to his improvisations, producing a more dynamic range of sound than his two hands could make on their own.”

Huh. So, now connecting what I have been reading in the magazine piece to what I have been reading in The Art of Is, a question that keeps popping up into my head is this:

Can algorithms improvise? Will they be able to improvise? What will that mean?

Or is this one of those fault lines between human and machine, where improvisation is an art form that is truly and only human — something that draws from the heart, soul and mind? Or is that statement my own naive thinking, and machines will, if not now then soon enough, come to to learn how to be improvisational machines, creating art in the moment with no pre-programmed instructions? And if so, will we be able to tell the difference?

I’d like to think the first is true — that improvisation is a skill only humans have — and that we will be able to hear and recognize an improvisational solo by machine (Computer vs Coltrane? Saxophonist wins every time … right?) but I am less and less confident on that declaration as the years progress and I see projects like these unfolding.

What do you think?

Peace (I wrote that myself),
Kevin

 

It’s There In the Mistakes Where the Real Art Might Happen

Book nibblers

I’ve been reading, with other CLMOOC friends, the book, The Art of Is by Stephen Nachmanovitch, in an annotation site called NowComment, which has allowed us to engage with the text and have conversations together in the margins.

I reached a section last night about mistakes and improvisation, and how mistakes fuel creativity, but also, how so many people are taught to fear mistakes, how they avoid mistakes at all costs, and how many experience shame when things go awry (partly due to our education systems that teach failure vs success as a zero sum game).

“Mistakes” in improvisation are hard to define, but people recognize when something works and when it doesn’t. — from The Art of Is by Stephen Nachmanovitch

Nachmanovitch suggests otherwise, arguing that the possibilities of art also emerge from places where things are imperfect or gone awry but that finding the art in the mistakes requires your ear to be attuned to the possibilities emerging from the unknown, from the mistakes. This is embedded into improvisation, he suggests, and I agree.

For this is how I write songs. I start at a place where I know I will begin (common chords and common patterns) and then listen for my fingers fumbling on the fretboard of the guitar, knowing that sometimes, something happens — and I need to expect the unexpected, and use the mistake for a new path forward. In those moments, I am often scribbling lyrics like a madman, to capture the idea before it gets away from me.

In my rock and roll band, on my saxophone, when I am soloing, as improv, this, too, is what happens — I often don’t know where I am going with my playing on solos, and sometimes, my fingers take me in a direction that is, well, news to me. This nearly always sparks some internal panic known only to me, Making a mistake in a gathering with other musicians, particularly with a live audience in front of you — that’s tricky business, and stressful, and exciting, too. I’ve learned to trust myself in this high wire act, though. I trust I will know where to go even if I am not aware of it as I’m thinking on it.

Thinking on this concept of mistakes, I wrote a poem for the margins of our shared reading experience …

On Making Mistakes

No one ever noticed,
for more than the second
it takes one to forget,
the perfect song:
the magic of listening
to remember resides
in the blemishes —
the transposed chords,
the slightly off-kilter
phrasing, the slip of the
voice, the chipped reed, the
spit-filled tube, the broken
drumstick, the snapped piano
string, the panic that produced
something to ponder
when the music’s echoes
have since long been over

Peace (where the path goes, follow),
Kevin

Buried in the Feldgang: A Poem from a Quote from a Book from an Idea

Quote: The Art of Is

I am forever overlapping
you; your notes cascading
upon me; where shadows
loom, you hold the light

We meet in the middle,
at the bridge – at the break –
at the moment of unexpected
surrender to the moment of
story and song

I am melody: nothing, but
for the harmony that spans
its wings beneath

Note: this is a #smallpoem, written in the margins of a community feldgang, with this line as anchor:

“Making art, whether you do it solo or in a group, derives its patterns from everything around us, in an interdependent network.” — Stephen Nachmanovitch, The Art of Is

Others have been leaving poems, too, in the book we are reading together in NowComment, and finding them in the margins of the text is a beautiful moment — a dance along the contours of Nachmanovitch’s ideas, made visible for shared experiences.

Peace (along the margins),
Kevin

Digging Ever Deeper Down into The Art of Is

Book nibblers

Terry has us tunneling into the book The Art of Is by Stephen Nachmanovitch, a book with the tantalizing subtitle of “Improvising As A Way of Life” that caught my attention. The introduction has my attention, for sure, as Nachmanovitch weaves in the concepts of improvisation to all sorts of ideas — music, art, text, collaborations, etc. I like the scope of it.

We’re inside NowComment as an annotation space (contact Terry if you want an invite), I am working to make art out of my reading experience. The comic above is a play on Terry’s invitation on Twitter and Mastodon, about “nibbling” at the edges of the work.

I then made this comic on my first reading start, trying to reframe the cover of the book as a piece of art and trying to explore the strange wording of the book’s title.

Comic reading

I’ve also been writing poetry — some of it found right inside the book —

found poem inside The Art of Is

Who knows where this improv will lead … following threads takes faith that the unraveling leads to understanding.

Peace (inside, outside, beyond),
Kevin

 

More Play in the World of Words: Tinkering with Text Animation Apps

Playing with Words: Animation Apps(Created in the Plays app)

I am still tinkering around with different apps that animate words. This week, I explored the apps Plays, MOTT and then came back to Legend (re-found in the Google Play store after it disappeared but not found anymore in the Apple App store). None of these fit exactly what I am looking for but some come close enough to have fun with. Some of these examples here are riffs off others work (Terry, in particular) and others are just isolated word play or riffs off my own poems. I explored some others in an earlier post.

Playing with Words: Animation Apps(Created in the MOTT app)

(Created in the Legend app)

Playing with Words: Animation Apps(Created in the MOTT app)

Playing with Words: Animation Apps(Created in the Plays app)

Playing with Words: Animation Apps(Created in the Plays app)

(Created in Legend app)

Peace (in the movement),
Kevin

Words, Image, Texts and Google’s Generation of ePoetry

“You are invited to donate a single word.”

This is how it begins. An invitation to write. It knows my weakness.

“Your word will be instantly incorporated into an original two line poem generated by an algorithm trained on over 20 million words of 19th century poetry.”

Call me intrigued.

I arrive at this Google experiment (privacy hackles, dutifully raised) in poetry via Terry entitled Poem Portraits, and so I dig in, and learn that it is a collaborative poetry that is “ever evolving” as people add words and Google’s AI system culls through a myriad of texts it has in its data banks. They call it “An experiment at the boundaries of AI and human collaboration.”

As you add a word (my donated word: Harmonize), it uses your contribution to generate new lines of text, adding to an ever-expanding ongoing poem collaboration between human and machine. The AI asks for a selfie (but you don’t need to do one to add a word), and this is where I paused but then decided to do it and go further.

I had seen Terry’s, and then Sarah’s, and then Charlene’s, and then Sheri’s, and the fact is, I was still intrigued by the mix of poetry, text, words and collaboration.

AI poem-portrait

The result is your word, and the words of your part of the poem, projected and mapped on your face, so that you become part of the poem. (Who knows where all those selfies go .. I suspect it becomes part of Google’s facial recognition data base. I’m sure I am already in there, but I would not likely bring students to this kind of poetry experiment).

I wanted to do more with the photo that gets generated. When you get to this step of your poem on your face, you can also read the larger, collaborative, AI-generated (with your word now added) unfolding on the page (You can access the scrolling poem without participating if you stop before adding a word).

So, I relocated my poem-selfie into the mobile app Fused, and began to layer in some visual static, working to deliberatively create a sort of fuzzy overlay of the selfie poem, as a means to represent some discomfort with how I willingly gave my image to Google.

Then, I wrote a short piece of music in Thumbjam, keeping the idea of my word — Harmonize — in mind, and working to layer three different musical sounds that work in harmony, and a bit of disharmony, too.

Finally, I took all of those pieces into iMovie and wove the media together, with a vocal reading of the text that filters across my face as part of my stanza of the poem.

The result of my playing is the video above … which starts as AI machine but ends with me, the pesky human, taking control of the image and poem again. (Or so I imagine).

Peace (in poems),
Kevin

PS — this is how Terry played with his results, calling it his “ghost”

 

Upon Reflection: A Month of Unexpected Poems

Random Access Poetry

During April, every day, I woke up, not knowing what I was going to write. As part of my Random Access Poetry activity, my goal was to use a few different tools and sites to find an unexpected image that could spark a poem for the day. So, for 30 mornings, that’s what I would do — grab a cup of coffee, go to one of my image-finding spaces, land on an image and write small poems.

Here are some of the places I went to for random photo inspiration:

  • John Johnston’s Flickr Promptr (which he set up after I asked if anyone had anything that would generate a random image for poetry, and I so deeply appreciate that he took that idea and built something in Github)
  • John Johnston’s Flickr Stampr — which is as Creative Commons search engine
  • John Johnston’s (he’s great, right!) Flickr Blendr site, which randomly grabs two images and blends them together
  • Alan Levine’s Don’t Look At My Photos — designed to surface photographers in Flickr — a new photographer every hour (although I noticed some repeats and not all of the photos were Creative Commons designated — this is not Alan’s fault. By the way, I regularly use Alan’s Flickr Creative Commons attribution tool)
  • Internet Archives Book Images collection on Flickr — I follow a bot of this site, that sends out random images, and I often found neat things to inspire writing
  • Bud Hunt’s mostly-daily posting of images for poetry during April. Bud has been doing this for years, and I appreciate that he takes the time. Sometimes, I feel like he is doing it just for me, since I am often the only poet posting there
  • Big Huge Labs Random Photo Browser — not all of the images are Creative Commons, but there is a good variety and a search engine tool

Looking back over the 30 poems from April, there were some decent writing days, more than a few mediocre days and a couple of blah days with the poems. Some poems just worked and some poems just didn’t. Some poems seemed to write themselves — I would start and the lines would flow, and I’d try to figure out where the poem was going as it was being written. That’s an awfully strange and interesting experience. Other days, I’d get stuck mid-way into the piece, force myself to plow through and get to a good-enough stopping place.

What I found, as I was about to start writing each morning by calling up a photo with one of the tools above, is that I was searching for a hook in the visual image — something that grabbed my attention, a spark of a hidden story, or a character on the edges, or a small moment, or an emotion. I didn’t know what I was looking for as I was looking but I was fairly confident I might find it if I looked close enough with my writing eyes. Only once or twice did I not use the very first image I found and reset the process. Mostly, I let the random nature of my search become the inspiration, and just went with it.

The thing about poems is that they are designed to evoke, and photos can do the same. Evocation is also a tricky business for a writer in a rush — I wrote poems in a short span of time — and that’s why they don’t always work in this format. There was often a tension between what I saw, what I wrote, and what I aimed to accomplish. But I often left the writing with a phrase or line or stanza on the screen that I found worthy of the page, and for that, I was always inspired and confident as a poet.

If you bothered to read any of the poems, thank you. I hope you were writing, too.

Peace (in poems and more),
Kevin

Random Access Poetry: Day Thirty


Silky – Soyeux flickr photo by monteregina shared under a Creative Commons (BY-NC-SA) license

Random Access Poetry: Day Thirty

Some words just drift
upon air, a cluster
of potential
from a translucent
flower seeking soil for
root, all with hope
that a new poem
might yet emerge
from where another
nurtured —

one idea seeds
a
n
o
t
h
e
r

(NOTE: OK, so the letters of “another” as the last line of the poem are meant to be scattered across the page but the blog keeps formatting it to flush left and I have given up making it work. Imagine those letters scattering to the wind …)

Peace (in the flowering),
Kevin

PS — this is the last poem for this month’s poetry adventure I called Random Access Poetry, in which I used different paths to find images to inspire poems each morning. Thanks to Alan Levine, John Johnston, Bud Hunt, Sheri Edwards, Terry Elliott, Kim Douillard, Raymond Maxwell, Algot Runeman, Margaret Simon, and others for all of the places I have used to write poems and leave poems. Some of those pieces ended up here, as daily poems, and some just drifted into the comment bins of blog posts. Thanks, too, to all the photographers whose images helped inspire me. I tried to leave notes of appreciation where I could.