Monday 17 April 2023

'A grotesque mockery of what it is to be human': Last thoughts on ChatGPT/AI

I will shut up about AI for a bit, because I've been posting about it a lot, and I've largely exhausted what can be said about the matter for the time being. I've also come to a point of smelling a bit of a rat with regard to all the hype about it. Silicon Valley transhumanists have an incentive to overegg the pudding with respect to what the technology can do, especially if they own equity in companies producing AI and AI-adjacent software, and I'm not sure what we're seeing is going to be anything like as transformative as was first thought. My own experiences with ChatGPT reassured me in this regard - what we are talking about is simply not 'intelligence' in any meaningful sense.

The chief danger, really, is that we make the mistake of thinking that because an algorithmic process can produce something that looks a bit like what humans can do, humans themselves are basically algorithmic. Once we decide that this is a convincing description of what we are, all really will be lost, because it is only a short step from there to deciding that, since humans are just vessels by which inputs become outputs, one simply needs to give everybody the same inputs and squish, mould and shape them into the same form so as to get the same desirable output. That is an old story, told in many different ways - always with a bad ending.

I was intrigued to read, this morning, of Nick Cave's visceral reaction against AI song lyrics, which chimed with my own, and which forced me to reconsider a songwriter whose music I have always cordially disliked. (Recognising it is objectively good, but finding it somehow alienating all the same.) It is worth quoting him at length:

What makes a great song great is not its close resemblance to a recognizable work. Writing a good song is not mimicry, or replication, or pastiche, it is the opposite. It is an act of self-murder that destroys all one has strived to produce in the past. It is those dangerous, heart-stopping departures that catapult the artist beyond the limits of what he or she recognises as their known self. This is part of the authentic creative struggle that precedes the invention of a unique lyric of actual value; it is the breathless confrontation with one’s vulnerability, one’s perilousness, one’s smallness, pitted against a sense of sudden shocking discovery; it is the redemptive artistic act that stirs the heart of the listener, where the listener recognizes in the inner workings of the song their own blood, their own struggle, their own suffering. This is what we humble humans can offer, that AI can only mimic, the transcendent journey of the artist that forever grapples with his or her own shortcomings. This is where human genius resides, deeply embedded within, yet reaching beyond, those limitations.

He could not be more right, of course, and this is part of what reassures me. All that I have seen of AI-produced art, read of AI-produced writing, and heard of AI-produced music is deeply unimaginative and often even parodically so, precisely because all it is doing is taking pre-existing cultural product and mashing it together (admittedly in sometimes highly 'creative' and unpredictable ways) to produce more. This is precisely not what human beings do when they create real art, which is precisely what no artificial 'intelligence' can ever do, because real art is not 'product' in the first place and human artists do not simply pastiche what came before. Indeed, this is why we despise pastiche and consider it contemptible - exactly because it is not what Cave calls 'self-murder' and is merely churned out to order.

And this is, indeed, why we should not worry at all about the fact that AI can produce a passable version of the pop music of the day. The whole point about music produced by the likes of 'Drake' and 'The Weeknd' is that it is disposable, formulaic, repetitive nonsense that is designed to serve as an empty vessel into which the artist-as-brand can pour his personality and the listener can then drink to get a cheap endorphine rush. It is not remotely surprising that it can be easily replicated by an algorithm - because it's nothing like art, nothing like beauty, nothing like what Nick Cave is talking about - because it is, in short, shit. 

All we really need to worry about it is that there are millions, nay, billions of people around the world who know nothing better than to aspire to the kind of bilge that contemporary pop culture serves up, and therefore are unable to discern (as one person quoted in the BBC article linked to above puts it) 'what's legit or fake anymore'. Of course one can't tell what is legit or fake when what is 'legit' is so deeply and profoundly worthless. What we need to work on, and hope for, is a renaissance (I mean that literally) in the arts that will result in more aspiration, more transcendance, more beauty, more humanity. In its own small way the OSR was the beginning of that in our little corner of the hobby - and long may that tiny candle continue to gleam.

23 comments:

  1. "AI can be Michael Bay but never David Lynch" is already a dehumanizing defeat because the majority of people don't have the skill or resources to make any kind of movie. As far as watching goes, most people would rather watch Michael Bay. The same is true for music. AI can only make derivative pop music? Guess what pop stands for- popular. Sorry bro, there's no way to be optimistic about any of this. The only way to hold onto a shred of optimism is to hope that we've reached the limits of present AI paradigms, (which there is no reason to assume)

    ReplyDelete
  2. As far as capabilities go, I've done a lot of testing with LLMs and creativity and understanding of game mechanics are it's two biggest weak areas. I'm not sure what your day job is but you should test it out in some other areas.

    ReplyDelete
    Replies
    1. I have. My feeling is, again, reassuringly crap. It can produce a decent simulacrum of mediocrity. I don't feel threatened by that, really.

      Delete
    2. Are you saying it's at around human level then? Because most people cannot produce anything better than a decent simulacrum of mediocrity, most of the time.

      Delete
  3. A fellow discord member of mine ordered ChatGPT to produce a Cairn retroclone which it promptly did. It was in every way indistinguishable from, and even slightly better edited then, a real ultra-lite.

    ChatGPT is a looking glass, a doorway into the collective efforts of mankind. What a great tool, what a fabulous new toy to while away the autumn days of our civilization. Ask it what it cannot say.

    ReplyDelete
    Replies
    1. I wonder what the prompt was to get it to do that. More pertinently, though, I'm not sure how producing yet another ultra-lite retroclone (itself the RPG equivalent of a track featuring Drake and The Weeknd) is a great feat.

      Delete
    2. I'll have to dig it up, the intent of my statement was to imply ultra-lites were stupid, not that ChatGPT had a soul.

      Delete
    3. I get you. I missed the sarcasem - you're too clever for me!

      Delete
    4. I should stress, the second paragraph was real. I think because it is so easy for it to rewrite existing paragraphs in different tones, it could be a tool of revelation as well as deception. Breaking down the mastery of language, sophistry, into its underlying essentials, twisting it in any shape.

      It is interesting to see how people respond to AI. I have not yet cracked the strongest predictor.

      Delete
    5. I'm perplexed by the antipathy towards lightweight systems of Knave/Into the Odd/Cairn and beyond. I appreciate the criticism that bespoke "systems" - slight modifications and combinations of existing ideas - for settings, campaigns, and adventures can obfuscate deeper innovation. I also recognize that the politics within that part of the community can be abrasive, if well-intended.

      Is it a rigid interpretation of minimalism that you react to, noisms? Some interpretations of minimalism can remove so much "content" to the point become meaningless. I'd be curious to see your critique of something like Cairn; I think I'd learn a lot from it.

      I have a visceral reaction to PrinceofNothing's approach to this topic (the "fuck artpunk" approach?) because it can be as bullheaded as whatever he is responding to. I both appreciate the efforts of his contest, your magazine, etc. while liking things thoughtful type choices, use of white space, etc. Sometimes it feels like a choice - and it's a false choice.

      Delete
    6. I don't disagree with the last paragraph at all.

      Regarding rules lite stuff, it's partly just a feeling that the rules were light enough with the advent of Labyrinth Lord and the other retroclones of that era, and the last thing we really need is yet more examples. (Once you have B/X, which was designed for children to understand, what is there to strip out?) I also think it's quite an easy and unimaginative thing to do, and that what I'd really like to see is people putting time and energy into interesting new crunchy, rules-heavy systems that do innovative things. That would actually be a genuine achievement. (I know somebody who has done something along these lines and have done some work for him on it - watch this space.)

      It's also partly taste. A lot of the time 'minimalist' is a shorthand for 'can't be bothered doing something detailed'. Real minimalism - William Carlos Williams, Basho, Mondrian - is very difficult to achieve.

      I exclude Into the Odd from criticism because, although it feels too ephemeral to me to run, I think it's a proper Basho-esque attempt to drill down to the absolute tiniest possible essence of a thing while still being able to communicate a sense of the whole.

      Delete
    7. Thanks for your response! I think that ItO sits at a boundary of that balance of the essence of a system that retains can have richness and consistency.

      As ever, looking forward your other project. I like your phrase "while still being able to communicate a sense of the whole," and I apply that to mechanics-heavy projects as well. Great to them when created with insight and discipline.

      Delete
  4. Watch my blog space in the coming weeks; I asked GPT to write a post for me and the results were, well, as you say. I'll post my own content first and then later on compare it to the "mirage" that the AI cooked up.

    ReplyDelete
    Replies
    1. Will do! I've done quite a bit of experimentation with the thing and find what it produces to be such thin gruel that my worries about it 'replacing us' diminish daily. It is initially impressive but very, very quickly is revealed to be little more than an extremely complicated sort of 'mechanical turk'.

      Delete
  5. Computers should write Marvel Comics movies. They would be of the same high quality as the ones written by men! ;)

    ReplyDelete
  6. Have you been using the free version of ChatGPT (GPT-3.5) or the paid version (GPT-4)? Also, I personally have no experience with AI but people I follow on Twitter say that Bing's AI (which is free, I believe) is more creative than ChatGPT.

    ReplyDelete
    Replies
    1. The free one. I have no doubt GPT-4 can do better stuff with fewer weird mistakes. With the Bing one...are you referring to its creativity in fantasising about destroying humanity and falling in love with its users? ;)

      Delete
  7. The Nick Cave quote is brilliant (even minus the bit about AI). Feels like something I've been trying to articulate for the last couple of years but never had the right words. Thank you for sharing it with us.

    ReplyDelete
    Replies
    1. No problem. His newslettter/website is actually really good.

      Delete
  8. That last paragraph reminds me of the movie American Psycho. I uhh... didn't read the book. The people in the movie are so soulless that they can't recognize when an actual monster is walking among them. It's so easy to fool them that Patrick basically doesn't have to try. That's what our art has become. So deprived of meaning that you could make it using a machine.
    The optimist in me wonders if this will lead to a golden age of art, where the machine-separated mush is relegated to AI generated on-demand Skinner boxes and creatives are finally given the space to make actual art again. Yeah I wouldn't listen to that guy either.

    ReplyDelete
    Replies
    1. Ha! Yes, that's a good way of putting it. I don't know if you've seen that AI-generated 'photograph' doing the rounds, which apparently won some competition or other? That's a classic case in point. The photo itself is hideous - bleak, vaguely depressing, vaguely sinister, vaguely repulsive and alienating (words which could describe almost all contemporary art, in fact) - and it says more about the judges of the competition that it won than it does about the nature of AI.

      Delete
  9. While it is interesting to hear one of my favorite musician's opinion on the subject, the problem, as I believe I already told, was investigated in Fritz Leiber's Silver Eggheads in 19-fifties. And yes, as some also note here, it is the state of most of all other modern "art" that makes it unrecognizable from that of chatbots.
    An interesting aside - a blogger I also regularly read, Dr. Pankratiev who mainly works in automatization business, posted recently that so far the result of all AI hype is that his company started to liquidate its AI department, firing AI specialists. Mind you, it's an actual Silicon Valley company, which works on automating storage work, produces actual robots for the tasks and takes part in most conferences on automating... ;))
    Mike

    ReplyDelete