Aardvark Daily

Please visit the sponsor!

AI versus copyright

Copyright law is a protection put in place to ensure that those who own creative works are protected from the unauthorised distribution of those works.

We normally see copyright being used to protect music, movies, still images, written works and other things that have been produced by individuals or businesses. Without copyright it would be extremely difficult for any writer, publisher, performer, photographer or whatever, to make a living.

One of the key factors in claiming the protection of copyright laws is determining who exactly owns the intellectual property and thus claims that copyright.

Most of the time this is a fairly simple task. If you or I create an image, prose or artwork, we own the copyright -- unless we explicity assign that copyright to another person or entity. In the case of someone who works for another in a creative capacity it is normal for the copyright to be assigned to the employing entity.

However, what happens if "the work" is created by something that is not a person?

A few years ago, a court in the USA ruled that only works created by a human can be copyrighted. This issue was highlighted in the infamous Monkey selfie copyright dispute of 2014.

So where does that leave AI?

For example, if Dall-E-2 creates a stunning image in response to a carefully crafted query, can that image be protected by copyright?

It would appear not.

On this basis, it appears that AI may not have the rosey future predicted for it in the creative world.

Businesses and individuals rely on the protection of copyright to ensure that they can earn revenues from their efforts when creating intellectual property but if that IP has been produced by AI, no such protection can apply. It's not even as if the AI can assign copyright to the business because the AI isn't entitled to hold a copyright in the first instance.

I'm pretty sure this will mean that a lot of AI-produced works will not be declared as such and that becomes a very real danger.

Why?

Well there's the AI feedback loop to consider.

Right now, AI systems such as Dall-e-2 and ChatGPT have been trained by content scraped from the internet. It's only by using this human-generated content that these systems have been able to generate the rulesets that create their AI capabilities.

Consider however, what will happen if an increasing amount of the stuff we post to the internet is itself created by AI.

There is a huge potential for a positive feedback loop to develop that could have very negative consequences.

In effect, any trends, biases or prejudices that begin to appear in AI works that are published online will be amplified as the AI then scrapes those to improve its training.

There have been calls for all AI-generated creative works to be clearly flagged as such when they're posted online -- as a method of breaking this potential feedback loop. The idea is that AI would not include its own work when "learning" from what it scrapes.

Clearly however, if copyright can not be appled to that AI-generated work then those who use AI for business purposes will be very unlikely to set that flag or acknowledge the source of those works. They will instead assert copyright over them by claiming they were produced by human resources.

Then there's the issue of the copyrights that AI may be violating when it scrapes protected works and uses them to "learn" from.

In effect, all AI works are "derived" from other works so should the copyright holders of those works be entitled to compensation for that use?

It's clear that AI was never even on the radar when the current generation of copyright laws were drafted and it may be an incredibly difficult task to come up with a set of laws that address all the issues that are now becoming apparent.

Ah,we live in interesting times eh?

Carpe Diem folks!

Please visit the sponsor!

Have your say in the Aardvark Forums.

PERMALINK to this column

Rank This Aardvark Page

Change Font