An end to copyright?

I suspect that AI generated content is (eventually, not any of the current versions) going to destroy copyright as a concept.

The generally-stated reason for copyright is to incentivise the creation of more works: artificial scarcity, which drives up prices, introduced around the time the printing press was invented (but the idea goes back much further). But, when a creative work can be made for less than the cost of just the food needed to feed a human for just long enough to do no more than merely ask a machine to make it, the economic value disappears and only tradition remains as a reason for copyright. (Trademarks are a totally different thing, consumer protection not creativity, which is why Apple (Computer, the tech giant) coexisted with Apple (Corp and Records, the Beatles’ companies) for so long, and also why they fought every time the computer company did more music stuff).

When will this time come, that creativity is no longer economically relevant or seen as worth protecting?

That’s hard to predict — for all that DALL•E 2 and Stable Diffusion change things (and GPT-3 for text, and probably something for music but that’s not my scene), these AI are not quite at human level yet, and the failure modes can be substantial, and so I don’t know how much computing power a genuinely human level creative AI would consume. However, I can do this comparison with the current image generators:

For the AI, maximum power draw of Mac mini M1 times the 90 seconds it takes Stable Diffusion to create an image = 3,510 joules; at $0.1/kWh[1] that’s $0.0001

For the human, 10 seconds to type in a prompt times the minimum cost of keeping a human alive (for which I am assuming the $1.90 per day abject poverty threshold) = $0.00022

The cost of the AI (with current generation software and slightly old hardware) doing the creative work is about half the minimum possible cost of getting a human to no more than merely ask the AI to do the creative task. Even though more expensive electricity and including the cost of the hardware can make it the other way around for now, that’s a very short way from the absolute cheapest possible human doing no more than asking.

However, all that is assuming this is purely about economics. It may well not be; for example, if art is humanity’s version of peacock tails, then the effort is the point, and if so, then even an uploaded brain of some famous and well-regarded artist will be dismissed as “not capable of real art”.

[1] Close enough to current US average: https://www.energybot.com/electricity-rates-by-state.html

Advertisement

2 responses to “An end to copyright?”

  1. Would it make a difference while there is still a spark of value in a human work? For now, AI is reliant on existing art to mashup and modify to create it’s output… meaning the original human sparks (style, technique, original idea) are still there. Thinking of the C’thulhu dalek by Delaney King (https://www.delaneyking.com/drawings/), I think we’re still a way off AI being a replacement for truly original works.

    The other idea I’ve seen floated is that a pure AI art industry could eventually turn into a reinforcement problem (as seen in other early AI fields). Now, the source data for AI art is practically 100% human art. What about when a significant portion of the source data is AI art? 20%, 40%? It’s running the same (similar) algorithm on its own algorithmic output, will that lead us to a less diverse body of work from AI artists? Will human inputs, being mostly first-generation art, be held at some sort of gold standard for input?

    Maybe like AI in other fields (the AI ‘apologists’?) the point is not to replace human art but to aid the human artist. “Give me Batwoman fighting a Transformer” then a human goes in to add the missing leg, fix the scale issues and algorithm artifacts, remove the duplicate moon, maybe adjust composition for beauty’s sake and refine the (very detailed) sketch that the AI has done.

    Liked by 1 person

    1. Certainly, for the moment.

      For now, these current versions of these specific AI models are useful for making mood boards, concept sketches, that kind of thing, but using the output of what exists today as a final product will *usually* (not always) end in tears.

      However, chess AI went through similar phases: first where it would beat only casual amateurs like me; then experts; then a brief period where it could beat all players but a human could collaborate with one to be better than either alone; but now… now a chess AI can’t be improved by any human help.

      I think something similar will happen with these AI: last year’s version of the first DALL•E model could just about manage to make shapes not quite entirely unlike the thing being asked for (but still better than a casual amateur like me); this year’s DALL•E 2 and Stable Diffusion are usually pretty good, despite having a lot of trouble with hands and faces and occasionally messing up anatomy in a fundamental way, and have won some minor artistic awards despite their flaws; a future version will likely beat any human artist but still produce things that human artists can improve upon; but I do not know what happens then, whether we will have an AI artist whose work would only be worsened by any possible human effort, or if we regard it as so necessarily subjective that such a claim is nonsensical.

      Of one thing I am certain however, we will, you and I, live long enough to find out if beauty really is in the eye of the beholder, or if instead it can exist in an absolute form like “truth” — even a mere decade is an eternity in AI development.

      Like

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

Blog at WordPress.com.

%d bloggers like this: