Category Archives: Artificial Intelligence

ETMOOC: The Emergence Of Deep Fake Music

Date
Posted by
Comments0

AI Music by StableDiffusion

I’ve been paying attention to the ruckus over the deep fake music that has showing up on social media in the last few weeks, where an AI-influenced song – Heart On My Sleeve — using the voices of Drake and The Weeknd has shaken the music industry. The song in question was different from some of the earlier memes and remixes that were taking root in social media. In this case, it was an entirely new song by AI with the voices of the two pop stars singing new words generated by AI. And it was pretty authentic sounding, too, in terms of song theme and vocal intonations and phrasings of the two artists.

This podcast episode — The Daily from New York Times — is a good listen for an overview of the situation.

As a musician and songwriter, I find myself conflicted on the emergence of AI in the field of making music. On one hand, it opens up some interesting doors for creative composition, using sounds and voices and techniques that might not otherwise be easily available to musicians. I’ve done some explorations of AI music sites but haven’t yet been all that impressed by what I’ve found. Clearly, though, there’s more out there that I haven’t yet discovered and played around with.

On the other hand, the legal and ethical issues of copyright and intellectual property use and infringement are huge, as it has been for the AI art generation field (lawsuits are already underway over the scraping of online content to feed the AI databases) and it feels like another reminder that the AI companies will need to find mechanisms (or be forced to, via lawsuits) for recognizing the contributions of musicians to any field of AI Music. I don’t know how that will be done, but it seems important for the tech folks to figure it out.

This all reminds me a bit of an earlier post I wrote a few months back about using loop tracks to construct songs — which is something I do along with traditional guitar-in-hand songwriting– and whether that is “songwriting” or not, since I had not created the original files. The AI revolution takes that idea and pushes it about 100 notches further, in my opinion.

I’ll keep an eye on this field of AI Music, as songwriter, musician, fan.

Peace (and Sound),
Kevin

PS — Update — I wanted to try out making a song via AI, so I used UberDuck’s tool to create this rap: https://app.uberduck.ai/rap/song/6908cc42-58d3-4f3f-8411-5f3faf844270 

I can’t say I am all that impressed but it was interesting. I also used the free account, which is clearly limited in many ways.

ETMOOC2: The Potential Energy Cost Of Generative AI

Date
Posted by
Comments1

[Explored] Mairie de Londres - City Hall of London
[Explored] Mairie de Londres – City Hall of London flickr photo by Jopa Elleul shared under a Creative Commons (BY-NC) license

One interesting thread (via Kate T) that emerged in a discussion in ETMOOC2’s inquiry into Artificial Intelligence and the emergence of ChatGPT, Bard and others was the impact on the environment. I hadn’t really thought much about it (I know, I should have) but when we think of how cryptocurrencies and NFTs all came to have a rather outsized impact via energy use and computing power, it made sense that generative AI would likely be doing the same.

An article in my May 2023 Wired magazine entitled “Generative AI’s Dirty Secret” by Chris Stokel-Walker also explores the topic, and while data from companies like Microsoft and Google and others releasing AI into the world is hard for outside researchers to come by, the article notes that “AI-fortified search engines are likely to demand far more computing power, which means a massive increase in energy use and carbon emissions.” (Stokel-Walker).

I suspect these companies are already working on data center efficiencies and other alternatives to the way the algorithms work to combat this — at the least, conserving energy will save them money, but also, many of these companies have environmental, climate-focused plans in their mission statements, although how much we should trust those mission statements in a business model designed for profit over the common good remains to be seen.

Kate shared these three articles that might be of interest on this topic:

This diagram — which widens the scope a bit — is from the article at Environmental Science & Technology:

As with all things, technology-related, we have to keep an eye on the picture of the world at large, and consider the impacts on the environment and climate, and us, as we dive into new tools and consider the pros and cons.

Peace (And This Planet),
Kevin

ETMOOC2 Webcomic: It’s Only AI (New Bot)

Date
Posted by
Comments0

It's Only AI 1 (New Bot)I was enjoying a week-long series of comics from the Zits comic strip (April 17-21) about Chatbots and thought, I should try to do a series of comics myself, using what I am learning about through ETMOOC2 and inquiry.

Today is the first of what may be only a handful of my own AI-infused comics, mostly focused on George.Paul.Thomas (aka GPT) as a bot with personality. I can’t write the kind of deep humor that the creators of Zits can (they’re so good at it) but I’ll do my best. Some of the comics may work better than others. Just putting that out there in advance, y’all.

I’ve often turned to comics as reflective commentary, mostly using a comic strip app on my old iPad for the art and framing. My Panels From The Pandemic comics kept me sane during those strange times. I have created comics about writing. And then there was a month-long exploration of making poetry comics. And many moons ago now, I first tried my hand at a comic called Boolean Squared. (note: that comic site needs a lot of work, so apologies in advance)

So, you know, here I go again ….

Peace (Framed, Funny and Free),
Kevin

 

ChatGPT Plays A Mad Lib Game It Made

Date
Posted by
Comments0

MadLib ChatGPT combined

(This is the fourth inquiry experiment this week with Artificial Intelligence for ETMOOC2. The first was a poetry exchange between AI platforms. The second was an Interactive Fiction project with ChatGPT and Twine. The third was generating code for a website about saxophone design).

This was a strange idea that came to me during the week, but I wondered how ChatGPT would do if I asked it to make a Mad Lib-style story (with prompts for words that a reader/player would suggest and fill in before reading the actual story) on the theme of ChatGPT. This concept became a starter point for my inquiry activity.

I actually ended up asking it compose me four different stories — one about The Day In The Life Of ChatGPT. Another is about a conversation between Google’s Bard and ChatGPT (which it wrote, without my prompting, emerged in the style of Shakespearean English — sort of). ChatGPT came up with that title all on its own: A Hilarious Conversation Between Google’s Bard AI And ChatGPT. (Note: the “hilarious” is an oversell, in my opinion). The other two stories involve a day without technology called When Nothing Worked and finally, Walking A Robotic Dog.

MadLib ChatGPT Title Screen

You can play the four Mad Libs.

I then reversed course and had ChatGPT play one of its own Mad Libs, asking it via a series of queries for random Parts of Speech words for each of the prompts it had just generated for the story (Such as: give me a random “noun”), plugging those words into the Mad Lib story, and the final result came out predictably strange (but perhaps no stranger than any regular human Mad Lib game).

Process Notes: How I did this experiment was that I took the story generated by ChatGPT with a simple prompt — Make me a Mad Lib style story about … —  along with the list of word prompts that were generated by ChatGPT from my query and I moved those over into a Mad Lib generator hosted over at Flippity (which I use with my students). The formatting on the Bard/Chat story dialogue got funky on me and I can’t figure out how to fix it, so it became an odd run-on conversation. I added some character line breaks to help but it still looks mangled to me.

By the way, when I took the same prompts for Mad Lib stories into Google’s Bard, the results were awful and not at all useful. There really is a noticeable difference between the two AI platforms at this point for the kinds of inquiry that I am doing here, where ChatGPT has much more flexibility and creativity built in (or maybe it is more allowed to flourish, likely) than Bard has at this point in time.

Reflections: When we do Parts of Speech in class, we often play and create Mad Libs as a way to play games with language. I suppose ChatGPT could be a collaborator in the story element of these activities, but its sense of humor is only iffy, and what Mad Libs really need is a subversive streak to them. No need to even invite Bard to the party, though. One thing I noticed: The AI was limited in the Parts of Speech it would leave out of the stories.  It was all nouns, adjectives and verbs, and that got pretty boring as a reader/player. A human collaborator would have to expand the choices.

Peace (Filling In The Blanks),
Kevin