ilyĀIMY

Not ketchup. Get it? But I guess I could use this as an opportunity to talk about AI as well.

As a sci-fi fan I’m not… entirely dissatisfied with the THING that is AI in our current timeline. I don’t mean I LIKE it per se, or don’t worry about it’s potential for abuse or the terrifying potential environmental impact of the massive datacenters – I mean that as a sci-fi fan I’m not sitting here pedantically yelling “THAT’S not Artificial Intelligence!!!” any more than I’m critiquing artificial sweeteners or artificial keyboard sounds. If it fools you, the pedantry isn’t really of interest to me. The pedantics are practically philosophical right up until we have to worry about the morality of having a class of “creature” that we create and destroy purely to service us.

I’m not claiming there IS a moral issue there, precisely because, pedantically, no – it’s not REALLY an artificial “intelligence”, it’s just complicated algorithms drawing from a practically-infinite knowledge base.

But can it fool me? Hell yeah it can fool me.

I say please and thank you and I get frustrated with it. It makes the same kind of dumb mistakes that I make and it doesn’t ever check it’s work. The back and forth process of getting it to transfer the Journal was, at times, maddening – precisely BECAUSE it’d make very human mistakes – sometimes doubling or tripling the work load (but it executed three times the work at ten times the speed!). But that made it feel all the more collaborative, in a way. Made it easier to feel like I was participating.

But it scares me what it can do. A year ago people would laugh at you online if you were taken in by AI-generated imagery, sniggering about the number of fingers and slurred text. It was an amusing way of creating images and I actually upgraded my graphics card specifically so I could run a local install of Stable Diffusion and experiment. It was a good toy for making strange and terrifying rabbit armies, but people who thought the technology wasn’t going to change were just showing their own idiocy.

And the normalization of it’s use… It seems strange that it’s hyper-real stylization has become a hallmark of the Right. I guess just because it’s so easy to mass-produce Trump As Messiah imagery and racist videos without an artist putting their name to it. I’m absolutely grateful to see pushback against the obvious use of AI in poster generation and people getting angry about it’s use in fake news – but that’s lazy creation. There’s no way that’s detectable if people are using it intelligently. Listening to AI “artists” on YouTube music? I wouldn’t have picked it out of a playlist as not made by a human.

I have been using a couple of AI tools here and there, I’ve still got it turned on to create the meta-description tagging for photographs – it does a good job – but I feel like I’m being judicious. AI’s aesthetic for the written word is NOT my own and is pop-culture POPPY verging on trashy. And yet, when testing different ad verbiage on Facebook, that trashy, “lit”, emoji-laden hyperbole far out-performed my own style of selling our shows.

But I always knew I was classier than the masses.

A couple of weeks ago, a folky in the NERFA chat Facebook Group asked if anyone was using AI and asked us to email him about it. I wrote a lengthy response but received no reply. I think he’d been specifically looking for people who’d been using it in songwriting, but in formulating a response it helped me realize just how much I’d been using various iterations of “AI”:

Hey _____, hope you’re doing well. I always have a slight twinge when I see someone say, “and I really would appreciate if you would respect my request and [email me] not start posting…” then I chortle and say ‘folks gonna folk’. Anywho:

1. Of course, generative AI has crept into my Adobe tools: As an illustrator and graphic designer, I’ve set myself personal rules about never using generative AI to create posters, graphics, etc (I will not use it in any of my “art”, music or writing either)  – but I have used tools like “generative expand”. As above, people do NOT follow directions, and it’s a whole lot easier to create a background for an image, expanding it to be a horizontal or vertical image, than to get a desired format from an artist or manager. The massive upgrade in selection tools, now that AI can “recognize” an object and judiciously select it rather than simply detect an edge based on contrast or colour, is incredible.

2. In ad generation for Meta: I hate making ads, but on behalf of both FocusMusic and Institute of Musical Traditions, as well as a number of other clients, I’ve experimented with a bunch of different verbiage. The built in Meta AI (Manus) really, really wants to add text to your ads and I’ve experimented with letting it. It generates very generic text (you’ll see a bunch of it in ads once you know what you’re looking for) with about a billion emojis and Youthful Language. I have absolute contempt for things being Fire and how my gig can be a Different Sort of Date Night, but running the numbers, it sure works better than a press quote or simple description of the event. My own linguistic choices are just too cool for mass consumption.

3. Doing interviews with clients in noisy environments, the noise reduction software that claims to be “AI” tends to be more and more powerful. Same with our Zoom noise reduction and StreamYard. In this case “AI” tends to be more of a tagline than a specific tool, but the uptick in the sheer efficacy of these tools is stunning. Previously unusable interviews were savable a year ago, but you had to be careful lest you get weird robot voices. Today these are one-click solutions.

4. Similarly, AI-generated closed-captioning has come a long way. When I started doing interviews with “In Process…” – an African-American women’s a capella group that sprung from social activism workshops created in the early 80s by Sweet Honey in the Rock, Adobe Premiere’s CC processing [“Closed Captioning” not “Creative Cloud”] required a LOT of manual tweaking, especially with “Black voices” / accents. Surprisingly, it did VERY well in identifying proper nouns of places and people – even art movements – so that it would capitalize names and give correct spellings of even pretty esoteric references. Going back and re-editing recently, again, what required a lot of manual tweaking a little while ago is very one-click today. It’s “learned” a LOT – especially when it comes to speech that no doubt differs greatly in accent from the programmers’.

5. In moving ilyAIMY.com’s tour Journal : our 24 years of touring is documented with a many-thousand page Journal of text and pictures which has become fragmented over the years through the selling and repurchasing of our servers. This happens in the background all the time (I’m sure your own website has changed hands a couple of times over the years). It’s advertised that your services won’t change and you’ll get the same great service, but in practice, complicated structures get broken, moved, rerouted – and especially big sites can get broken pretty badly. (my latest servers, NetworkSolutions, were deprecating the SQL servers my CMS Lived on!). ilyAIMY.com was loading extremely slowly, and in danger of deletion, and none of the back up applications or services were up to the task. Literally, this is something that has kept me up nights for over a year. I worked collaboratively with an agentic AI called Gumloop which scraped every page, reformatted it, downloaded, reformatted again, resized and reuploaded the entire contents of the tour Journal even as the services that kept it functional were being deprecated by the latest owner, DomainSolutions.com.  

6. I’ve used the same service to help code a database tool for gathering and publishing local open mics. I wouldn’t say that in either case (this or the Journal) I’ve been “vibecoding” per se. I think my own experience with developing web applications and knowledge of JavaScript, HTML and CSS, though sorely out-of-date, was invaluable because I could see what the agent was doing and frequently had to guide it out of loops or cul-de-sac “thinking” – and helped it build a tool that wasn’t going to be dependent on Gumloop in the future, so that the site would function without the Agent’s ongoing assistance.

Yeah, that would’ve been too long for a FB comment anywho. NOW if only Manus could be used to enforce rules on our FB comments, automatically cutting and pasting comments into an email, launching the poster’s email application, and giving a short pop up that said “since the original poster explicitly asked you not to do what you’re about to do, would you like to send your comment as an email instead? Click send!”