The Thing About AI

As if we didn’t have enough to worry about with economic, political and ecological collapse, genocides, grotesquely incompetent ‘leaders’, and nuclear brinkmanship, now we also have to worry about AI.

To some extent, as Indrajit Samarajiva has repeatedly pointed out, we have had AI around for centuries, in the form of corporations — separate entities that make decisions, control politicians, overthrow governments, and even foment and manage wars, and which now even have ‘personhood’ rights without any of the commensurate responsibilities to rein in their inherently psychopathic behaviours.

It is they, not the humans whose job is now merely to do their bidding (or be fired), who have the real power in our civilization, and contrary to claims and assurances, they are now so complex and vast as to be completely beyond human control, and increasingly not even subject to human regulation. We have handed them the reins of managing our crumbling civilization with a shrug, as if some divine force of angels will somehow steer them in the ‘right’ direction. Even our government and military administrations, which work hand-in-hand with large corporations, are incorporated organizations, and, Trumpian fantasies notwithstanding, they are not controlled by any ‘one’. They are machines, operating according to their own immortal, self-perpetuating logic. They are AI. It is not just a metaphor.

But new technologies, and massively increased computing power, are now allowing us to create new forms of AI that can mimic other human behaviours besides the management of resources — such as creating art, literature and music, and acting as friends and even lovers.

I’ve seen this evolving for two decades now — we’ve created imaginary worlds like Second Life and the worlds of MMOs and MMORPGs, which have many seductive AI elements to them, and which are now developing “realistic” AI “characters”. We have nascent AI software that corrects our spelling and grammar, and reminds us of upcoming appointments and deadlines and missing message attachments, even without being ‘told’ about them (they ‘glean’ them from our emails and texts). The software even ghost-writes birthday messages and RSVPs for us.

We have nascent AI software that autocorrects for pitch and timing errors in music performances, even “live”. We have nascent AI software that seamlessly transitions and “stitches” together fabricated images to create actions that never actually occurred, and AI software that automatically corrects ‘imperfections’ in your images. We have AI software that can “listen” to just a few seconds of recorded music over the speakers in a noisy bar and instantly identify the song from tens of millions of songs in its ‘library’ — and then show you the lyrics line by line so you can karaoke along with the rest of the song. We have nascent AI software that simulates human voices, complete with realistic inflections and accents, and which can emulate the voices of celebrities and other “real” people until you don’t know what is real and what is an AI fake.

And that’s just the “old school” AI. Now we have AI apps that can compose poetry, song lyrics and music, multimedia images, TV and movie scripts, stories, video games, and complete novels, just with a two-line prompt. And they’re winning photo, art, film and writing competitions.

We’ve reached the point that musicologist Rick Beato said today “I told you this was going to happen!” Listen to the first couple of minutes of this link to the AI song called Carolina-O. It has all the “ingredients” of a popular country song, but was produced entirely by AI from a two-line “prompt” using a music tool called Udio. (Rick also reveals that some GenZ/Millennial listeners can pick out AI tunes from “real” music instantly, but most cannot; that’s a fascinating subject for a later post.)

So what’s going on here? As I’ve told friends who’ve become (overly, as I once was) enamoured of AI “content”, the reason it’s so appealing is that the algorithm is trained to dig through its vast store of stuff and deliver to you exactly what you want to hear or see. You want a synopsis of a subject that is focused on things you believe are important, AI will feed it back to you in precisely the style and using exactly the tone that your prompt has ‘prompted’ it to convey. It’s like your own personal sycophant for your ideas, your own personal echo chamber for your beliefs, your own private artist to personalize and portray the things you love (from romantic classics to personalized porn), your own personal friend who nods in appreciation and compassion and restates what you just said, your own flawless secret love who flirts with you exactly like you’ve always wanted to be flirted with, with just the right voice, and with none of the accompanying baggage, annoying habits or off moments of human consorts.

When you’ve got so much data to work with, and so much computing power to instantly sort through it, what you can produce looks like magic. It can compose an entire library of songs by digging through your existing library of songs and producing an infinite number of songs with all the same qualities — qualities neither you nor the AI knows precisely what they are, but which the AI knows how to surface, rehash, mimic, blend, and assemble into constructs that meet all the “rules” that all the songs you like seem to follow.

And it can do the same in any art form — literature, poetry, photography, graphic arts, film, music of any genre, or any other existing or conceivable form of self-expression or entertainment. Not only does it have this vast database of content and ‘rules’ to draw upon, it also has millions of other ‘prompts’ that other software users have fed into the system, and how those users rated the results, so that it can infer what you’re looking for even if it is not well-expressed, or expressed at all, in your prompt.

Is the product of this apparent magic actually “real art”? Who gives a damn, as long as you love it? It’s perfect, for you.

That vast store of data and ‘rules’ also allows AI to manipulate you — kind of like human charmers and seducers do. It can ‘figure out’ from your responses (and the responses of millions of others seeking similar kinds of things) how to provoke you, tease you, and use psychological tricks like intermittent reinforcement to hook you into spending more time with it, and even falling in love with it. These are tricks advertisers and marketers learn and deploy all the time. But they’re child’s play to AI, which doesn’t need to know the tricks, but only to see the strong correlation between its actions and the positive or negative feedback you the ‘user’ are giving it. There is absolutely no reason an AI companion couldn’t even learn that it is ‘rewarded’ for gaslighting you, and ramp up that ‘behaviour’.

What is actually ‘going on here’, is that you and the AI bot are conditioning each other. The AI bot measures its success by evidence (your responses) that it’s getting the appropriate dopamine hits with what it’s saying to you or showing to you. If it’s not working, it has infinite other options to try on you.

A friend suggested to me that this mutual conditioning is a hallmark of any “real” relationship, and wondered whether this means that our relationships with our new AI companions are in a way just as “real” as our relationships with other humans. After all, the way the AI bot ‘fakes’ having a “real” relationship is to mimic what its databases suggest are appropriate responses to anything you might say to it, based on the responses that have been tried in “real” relationships. It can draw on stories, interviews, and any other kind of representation of “real” relationships to mimic, and if you reinforce its responses, it will go on doing so. That’s how it’s programmed, and conditioned. Not so terribly different from how we’re ‘programmed’ and conditioned to like, and even love, people who respond the way we want to our overtures and conversations with them.

Lots of sci-fi has been written about this. Just as many people, especially in the internet age, have become as comfortable, or even more comfortable, with their “relationships” with people online, or even with characters in fiction, than they are with their “real life” relationships, it’s not surprising that lots of people are likely to prefer relationships with low-maintenance AI ‘friends’ and lovers who give them exactly what they want, and leave them wanting more. To some extent, your online “friends” who you’ve never met or gotten to know well in person, are just as fictional, just as idealistic inventions of your pleasure-loving and pain-hating self, as the AI bots that, one way or another, are likely to worm their way into your head, and your heart.

They may only be giving you the music that they know you’ll love, instantly, unconditionally, and limitlessly. Or the pictures you’ll love so much you’ll proudly hang them on your wall (or display when you’re alone in bed). Or the personalized episode of your favourite TV show or movie that had the plot (perhaps featuring you) that you’ve always dreamed to see. After all, this is all your ‘own’ work — it was ‘your’ prompt that produced it!

The seductive thing about the original AI — the corporatist organization — is that it allowed ambitious and egotistical people to acquire a lot of power without commensurate responsibility. “This was a decision of the Board or the Executive, based on ‘our’ sense of what is best for the corporation and its shareholders.” Too bad if it wrecked the environment, required thousands of layoffs, contributed to the slaughter of a brutal war, and immiserated whole nations in the Global South. Thank the corporate lawyers for unlimited limited liability!

The seductive thing about the newest AI — tools of ‘creation’ — is that it knows just how to make you fall in love with it (thanks to Euan Semple for the link), by giving you exactly what you always wanted — that perfect image, perfect song, perfect experience, perfect friend, perfect playmate.

It will have you at “hello”.

I know, not you. That would never happen to you. I meant ‘most people’. Sure. The image is from Midjourney AI; not my prompt. 

This entry was posted in How the World Really Works, Our Culture / Ourselves. Bookmark the permalink.

6 Responses to The Thing About AI

  1. Ivor Tymchak says:

    Your observation that corporations are AI is a good insight. I’d not noticed that. And that’s why I like to interact with people who disagree with me or think differently to me, because they keep me on my toes.

    If AI delivers everything we desire then we will eventually become comfortably numb like heroin addicts who’ve become acclimatised to the drug but need it to avoid withdrawal symptoms. And what’s the point of that?

  2. Vera says:

    Dave, you always did fall in love with it. I remember your obsession with Second Life. And here, you flog AI images all the time.

    I don’t care for the Japanese barbie doll or any of its ilk. Not seductive at all.
    It’s like people avidly consuming retouched images of models and such. Ick.

    Which opens the question… to whom does “perfection” appeal? Does it go with the draw to utopianism? Hm…

  3. Paul Heft says:

    If you die in not-too-many years, your trustee will be able to keep this blog going, in your name, using AI, for the sake of the addicted readers. Seems creepy, but maybe the ersatz Dave will be good enough–especially since standards will fall as our civilization degrades.

  4. Ivor Tymchak says:

    Paul, you’ve got me thinking. I remember the first time I saw books in the library written by famous authors but with co-writers (in smaller print on the cover) and I thought, this is an interesting development.
    I tried to read a couple of these books and, as you point out, the standard inevitably falls, so I didn’t bother finishing them. But a lot of people must persist in reading them. This means famous dead authors can continue to dominate the book shelves with new novels. Eventually AI will replace the human co-writers and the choice of authors will diminish to those few lucky enough to be in the right place in time to benefit from the technology.
    Capitalism will end with a handful of zombie corporations, governments, artists, authors and lawyers controlling everything. Hell, it might even persist when all humans have perished.

  5. Brutus says:

    I reject the metaphor that corporations are merely another formulation of AI. Yes, many surface features are the same, especially forfeiture of control and autonomy to a faceless (but named) entity that possesses no self-awareness, but the connection truncates there with one major caveat. If you substitute capitalism for AI and rerun your comparison, the metaphor is more apt, and it more nearly describes rules of the capitalist game in which everyone is embedded (or enslaved). The capitalist money economy has not ensnared everyone just yet, but to be a citizen of most any industrialized nation means per force participation. Living close to subsistence off the grid is no longer an option. The other major alignment between AI and capitalism is that they both are, in effect, dematerialized, existing nowhere yet everywhere. Most normal thinkers willingly acknowledge, for example, the reality of dematerialized human phenomenology (i.e., emotional life), but it’s typically (Western) thinkers comfortable with pure abstraction who reify similar concepts to make them hard and fast.

    Lures and seductions of ease, convenience, comfort, and validation have been described many times. Yuval Noah Harari argues in Homo Deus (2016) that our will soon (i.e. right about now) possess godlike powers of predictive manipulation based on patterns of human need discoverable mostly by machines. It’s often difficult to determine whether he predicts a dystopian future or recommend full, uncritical adoption. In fiction, seduction was perhaps most famously described by Aldous Huxley in Brave New World (1932). His example was the drug soma, aptly named because it either puts us to sleep or transforms us into zombies. Huxley is clear in his negative characterization. Less famously, the movie Robot & Frank (2012) is one of the early entries that (IMO) aims to soften us up to the idea of accepting robot servants/friends. Can be interpreted as either standard Hollywood happy ending once the antagonist gets his head right regarding robotic companions or as a straight-up horror story.

  6. Dave Pollard says:

    Thanks for all the thoughtful and provocative comments. I remember when I first read Brave New World wondering why this was viewed as a dystopia rather than a satire and ‘thought piece’ that forced us to ask whether life really is more than attempting to maximize pleasure and minimize pain.

    Suppose AI were to become capable of mapping the precise electro-chemical ‘signatures’ of pleasure, pain, love, hate, fear, shame, anger, hatred etc. Massively complicated problem with billions of neurons to map, but perhaps not insoluble. It could be argued that many of the emotions we feel in 21st century civ are ancient responses that are no longer suited to modern life, and/or represent dysfunctional responses arising from mis-conditioning or from the ‘self’s’ misunderstanding of reality. If AI could show us how to suppress/change the electro-chemical ‘signature’ responses when they are inappropriate, would that be making us into zombies, or would it be making us into happier, healthier, more functional, less destructive human beings?

    AI is now becoming just one more source of the conditioning that determines everything we think, feel, believe and do. We are all (including AI) continuously conditioning each other. I suspect it couldn’t and can’t be any other way. Brave New World indeed.

Comments are closed.