Note: This is incomplete. I ran out of steam on this project and am showcasing it as unfinished.

On AI, Morality

Is it truly stealing to use the power of machine learning to clone an individual's voice?

This is a moral dilemma I was slightly rudely confronted with in the Tails Gets Trolled Discord *spits* "server". So, I've taken this evening where I cannot sleep to pause and ponder over it. So, let's define some processes, shall we?

For the machine learning approach to voice cloning, where you are using an AI model to construct new audio forms out of a few existing ones, the overall majority of the work is being handled by the computer. To be more specific, often times this work is handled on the GPU, or Graphics Processing Unit, a nice piece of kit that handles simple arithmetic very quickly. This, somewhat magically (but truly entirely logically), is able to take raw audio files, notice a voice, and compute new things for this voice to say. Everything I'm saying here is quite abstracted, truly the AI model has no intuition for this stuff. For example, the AI might not know how to pronounce a certain word in a certain voice's language (if you're using a Text-To-Speech model). There, it would fail to produce expected output.

For the human approach to voice cloning, a technique has been around for 46 years (minimum). Sentence mixing, or, more apt for the post this evening, word mixing, is quite useful for "cloning" a voice. I put "cloning" in quotations as it's quite difficult to mangle word mixing to appear as normal speech, whereas with the AI implementation of this it is relatively easy (not for the GPU though!). It is, however, possible, with enough work.

Here's where I begin the post proper. I want to relate these two techniques to something quite ancient. That is, computing. Not in the sense of modern processors, but computing as an occupation. Back in the day, before modern advances in technology like floppy drives and AOL, to figure out a number of something large and to a great accuracy, you'd need rooms filled with people. Not transistors. Long hours of crunching numbers, often a position filled by women. But, point is, back in the day you needed human hands to figure out numbers of Pi or √2. When computers (as in the electronic version) were invented, suddenly a whole job class was taken out of business. Not immediately, of course, but you really didn't see much computer jobs after computers in large companies were widespread.

So, I suppose the real question here is whether or not those who lost their computer jobs were glad they were being replaced by machines. I don't of course mean this in the immediate sense. I mean 2-3 years after they lost their job and had moved on to another. Were they sad they lost their job? Or perhaps were they glad they no longer had to sit in a cubicle or at a desk moving around numbers all day.


Addendum: Piracy

Is some piracy okay? If some is okay, does that necessitate all as okay, like jokes? Can one really say they are morally justified to pirate somthing? If piracy is okay, does that make voice cloning okay?