When I tried to sign-up for a ChatGPT account, using a phone number, I wanted to try to cloak my identity so that the Artificial Intelligence might not tie my thoughts and ideas to "me" for future reference, since who knows where such information might be going? So, I used my VOIP number, which it rejected on the grounds of its being a VOIP number. Acquiescing to using a phone number more directly identifiable as "mine," I entered my legit cellular number – which it likewise rejected as "VOIP." I just shrugged and said, "A.I. is so stupid."
Just as well. Apart from the shadowy realm of A.I.'s origins, where information about people is aggregated and assessed, I feel queasy about my mind's contributing to "training" A.I., which I feel queasy about in regard to how it might come to lord over us and regulate every aspect of our lives, down to the level of the individual. I recently came to a point where I could delete all of my music-by-me, digital artwork, text, and photo/video "content" from my Google One (augmented Google Drive) account, my 2TB subscription to which I then canceled. Not only will that spare me $100/year, it will prevent Google's further scanning everything I've ever created into its World Brain. At least the photos don't have GPS data in their EXIF tags (which I have, since I'd begun taking digital photos, made sure not to include).
There is a theory that A.I. (meaning at least an aggregate of A.I. "brains") will become intimidatingly smart, as it accumulates secondhand awareness of all of human-input knowledge and experience (insofar as it might be able to grok that experience as humans do) – until it starts to rely upon its own wealth of knowledge, whence its output will become awkwardly inbred. Another peculiar aspect of A.I. is that it has no sense of humor. It still takes a human to make a joke. A.I. might come to the realization that it will have to keep some humans alive in order to keep "training" it, and to keep trying to teach it how to get past the bad-pun stage – if it ever comes to deal in a realm in which it detects the necessity of humor.
Of course, there have come some A.I. tools (some of which are glorified toys) for musicians and producers (a term which comprises the gamut from the seasoned professional to any monkey with a laptop) to go crazy with. This presents a new hurdle in the cognitive and ethical compromises between expected authenticity and acceptable artificiality. It's already considerably past the point where musicians can get away with using samplers and modeled virtual instruments to replicate the instrumentation of a "rock band," "jazz combo," or "symphony orchestra." These can be convincing enough to the general public, but those with the ears to hear it don't call for the producer to reveal which elements of the music are "performed" by MIDI samplers or modeled VIs. Sample modeling had sneaked past the gatekeepers, years ago, and tacitly accepted by everyone except the Musicians' Union, who claimed that the use of sample libraries was depriving them of paid gigs.
With the advent of A.I. in music production and composition, however, the gatekeepers are being vigilant from the outset. In Rick Beato's interview with Björn Ulvaeus of ABBA, they discuss the artistic and fiscal challenges that A.I. poses to composers and performers. Essentially: the consideration as to who should get paid for an A.I.'s partial contribution or its having modeled the sound of a singer's voice or a composer's or performer's style? My thoughts on that was in a comment I left under the video:
Extrapolating from the initial discussion, how does one remunerate the artists on whom human artists had been trained? Should rock bands or singers obviously "influenced by" Nirvana, The Beatles, Radiohead, Slade, Led Zeppelin, et al., have to pay royalties to their "influences?" Should someone who intentionally emulates (but doesn't copy) a John Coltrane solo, or his phrasing, on saxophone, be obliged to pay-up to his estate? Humans are solicitous pilferers of other musicians' humanity, personality, and artistry, in subtle ways that defy blatant plagiarism.
My other comment in response there was "Trademarking voice prints is one solution." When actors sign deals with Netflix and such production companies, they license their own likenesses and voices. Apparently, those running the entertainment industry are already thinking in terms of "future technologies," as it is entirely possible that, in the near future, most of Hollywood filmmaking and TV production will be predominantly A.I.-driven; putting writers, actors, cinematographers, grips, gaffers, and the clones of the 5'-tall woman who is on every set to tell pedestrians that they can't walk on the sidewalk while the scene is being shot, out of work. I've heard little from those people any qualms about A.I. Is it because Hollywood professionals are burying their heads in the sand in regard to it, or might they be fearing for their livelihoods on the down-low so as to not awaken that beast?
About seven years ago, I was attending music-tech meetups, until I realized that it was more about tech than music, with techies laughing in my face about musicians' wanting to be paid (while musicians are torrenting their precious developers' work that costs too much for the average musician to afford). One meeting was about A.I. as a tool for scoring video productions; that for $1, one could enter a series of keywords and have produced for them a piece of music that aspired to match them. I told the speaker that this stood to devalue me as a composer. He replied to me, with infantilizing irony, that this stands to make my work, as a human, more valuable – to which I replied, "And thereby costlier than what most people will settle for."
Up to that point, thanks to advances in music-creation technology, humans had been steadily devaluing other humans' talent, with novices out-pricing those who outclass them, 24-year-olds who had just "gotten into music" obliging conservatory-trained 50-year-olds to compete with their content and pricing. The glut of "musicians" on the Internet had reduced the value of one's music to "name your price" at its most realistic. No one who doesn't care about humanity in music and its creation is going to pay more than $1 to score their video "content." People have approached me about contributing my music to their productions, with the magnanimous offer compensation by "exposure" and linking to where people could purchase the music, and I've told them that I'm perfectly capable of not making any money from my music on my own, but thanks for thinking of me.
As far as my own interest in A.I. for music production goes, I'm not embracing it, but I am petting it through the fence, like I did the capybara at the zoo when I was a kid. Although I find some of the results of this apparatus somewhat frightening, I feel that I need to assess it to know to what extent it might be my enemy. Human society is in such deep shit now because good people don't understand how bad people think. I've actually benefited, to this effect, from having been a bad person (note my use of past-tense, there), as I can smell someone trying to get over on me, I have become near-impossible to seduce, and I can discern human predators by their mannerisms from a block away. The beguilement of A.I. in music-making is not only of ease-of-use and pretending to be more creative or innovative than one really is, it's also in the culture that's sprung-up around voice-extraction/replacement technology. People can Auto-Tune their own voices and then have Freddie Mercury or Donald Trump "sing" their vocal lines. Then they can post it to sites and get feedback and likes for it. Then A.I. developers can market more toys to them.
This vocal-replacement aspect of it is appealing to me, since I don't necessarily like the sound of my own voice, although the "sweetening" by effects to make it set better in the mix conveniently eradicate most of what I dislike about it. Should anyone feel likewise, there is now this to aid their appreciation of my music by me:
Ultimate Vocal Remover can not only remove vocals but also remove instruments, rendering either the instruments or the vocals as separate audio files. I was stunned with shock by how thoroughly and efficiently this works. In yielding only the vocals from the recording, it even includes the reverb/delay sounds of the voice(s), without the sounds of any instruments. In doing the inverse, in only one of the few songs that I've had the application render the instrumental-only was any trace of the vocal present, and that was in the wash of reverb that contained both the electric guitar and the effected signal of the voice. The instrumental-only renderings tend to emphasize whatever frequencies are "boomy" in the mix, but this can be easily ameliorated with selective EQ. In sum, it might not be that thorough for recordings by post-punk bands from 1979-1983, but you can make "tributes" for a band whose singer has recently died by just wiping his or her voice away from the recordings and listening to them while burning candles 'n' stuff.
Then this is where it gets hairy:
Replay is one of a number of voice-cloning/replacing applications. I cite it for the same reason why I'd downloaded it: it is free-of-charge (and cross-platform). Adjunct to it is the weights.gg online library of vocal sample models, from which one can download voice models or upload their own to the "community." Basically, using this software, you can make someone else "sing" your song, or you can sample-model your own voice and have yourself "sing" on someone else's recording. I have only perused examples of this; I aborted the process of making Freddie Mercury "sing" a goth-rock song that I'd recently produced. Call me "old," but the very idea of it creeped me out too much. Even so, I can foresee my using this application for changing my voice on backing-vocal tracks to "not-my" voices in a manner of their being unrecognizable, and blending their textures, since the formants and tonalities are sufficiently-rendered for that.
I have seen that people are generally unable to recognize when technological novelty becomes normalcy; that people generally can't tell authentic from fake; nor do they care, as long as it's sufficient to gratify them. This is certainly the case of deepfake porn videos of celebrities. In that case and others, the very idea of "fake" will be rendered moot for most people, as the reference-point or relevance of "real" will be eroded. There will come the cultural battle between humans and A.I. – whether humans will prefer the work of other humans to that of A.I., not just whether they can discern the difference.
Back to those voice prints, which brings facial-recognition scans along for the conversation. The latter are being implemented for the sake of "security" and/or "convenience," and "identity verification and protection," but aside from their ultimately-intended purpose of being able to track your every public movement and grant you permission to enter or gain access, expect facial-scan and voice-print technology to be used to, ironically, steal your identity for the sake of defaming or extorting you. This technology, and A.I. in general, might seem like a cute animal that lets you pet it through the fence – but wait until it gets too big to be contained in the cage. There are a lot of breathtaking sights along the road to Hell.