Not the word I'd use…

Tattoo ideas.

I think there's good reason to believe that small-scale implementations of machine learning will remain as time passes and the giant investment frenzy ends. Useful tools scarcely evaporate, and manageable projects that promise no more than reality are sustainable.

Many of these models are open-source, and whether or not venture capital and the fantasy of infinite growth continue to fund them, that groundwork is already laid, and history shows people willingly volunteer to develop utilities useful to them.

The significant problem to overcome here is that people refuse to accept that the problem lies in perverse motives for using technology rather than the mere existence of that technology itself. This isn't a new mindset, of course. Look at attitudes on nuclear power tainted by the atom bomb.

The thing is, this is a case where it literally is the thing they're complaining about. Speech synthesis to make virtual actors is the same as what's used for text-to-speech tools. It is the same machine which makes the slop they loathe, but they take issue not with the operator, but the machine.

What irks me is that there are people who over and over insist "there is no ethical/good/acceptable use for AI" and moralize anyone who objects. It's like suggesting modern chemistry is bad because the government made Agent Orange or because pfizer patents medication. The issue isn't chemistry.

I shouldn't have to deal with people telling me I'm evil or brain-poisoned or betraying their trust or kissing techbro ass or any other insulting things just because I am happy with an accessibility tool.

It's not stealing anyone's data or anyone's job, it's not boiling oceans or setting fire to the planet, and it's performing a task that makes my life better, objectively.

Querying the model wouldn't incur any resource load that my computer couldn't handle, and it would be spent quite decidedly performing a useful task: helping me consume written text that I otherwise couldn't easily. This should be seen as a good thing!

I could, in the course of a week, take a model with no training, provide it samples of my own spoken voice, train it on that data, and expend no more energy than I typically do playing video games, and have a tool to on-demand produce easily intelligible speech, hosted on my own computer.

Insisting that we cannot support the use of AI to generate speech because AI is somehow ontologically evil just makes no sense to me. It's an effective innovation on the techniques and practices of making artifical spoken words, and its resource costs are reasonable for the benefit.

Older methods of speech synthesis are pretty subpar, can be difficult to listen to for long periods of time, hard to parse, and very poorly articulate tone or inflection. I've been using a screenreader recently and I've gone through a few different utilities for speech synthesis, so I am familiar.

It's bizarre to me how wanting to be correct is immediately treated as bootlicking or grifting. You would think people would agree that mis- and disinformation are bad things, and that one can oppose something on grounds of informed and accurate assessments.

Seeing people doing wildly incorrect "I made up a guy wholesale to get mad at" shitflinging about "AI". There are things to hate about a lot of the commercial implementations of it and how it's used, but people continue to parrot complete lies about technology just because it wins them cool points.

This isn't about one particular thing, because it's a long-standing phenomenon people have exhibited probably since the advent of the foundations of epistemology and rhetoric, but I am definitely seeing it a lot more these days about one specific topic.