* Posts by Teesside John

1 publicly visible post • joined 9 Jan 2024

OpenAI: 'Impossible to train today’s leading AI models without using copyrighted materials'

Teesside John

Re: Sounds like...

Perhaps a more fundamental question is why have copyright at all? I guess to reward and incentive people to create stuff. Why would somebody spend months writing a novel and publisher pay for it to be edited and formatted if the first person to buy it can put it online for free or print and sell their own copies? There will always be some people who produce stuff for free, e.g fan fiction and open source software, but that should be the creator's choice and a lot won't get produced if there's no way to make money.

And not every creator is a rich rockstar - it'll be the masses that do copywriting, voice overs for corporate videos, graphic design for company logos etc that lose out first.

We therefore need a copyright system that works - one that makes it worthwhile to create stuff but doesn't leave people unable to innovate out of fear of being taken to court.

With AI, we need to make sure the copyright system still works. Imagine I'm a photographer who takes pictures of everyday things and local landmarks etc to sell on stock photo sites. I make a living - I'm happy. People use my photos - they're happy. Now imagine my photos get scraped for next to nothing and used by an AI to generate similar images for free. I now have no money (why pay when an AI can something similar - even customise it - for far less?) If I stop taking photos the AI has less data and eventually the images it produces start to look dated. Now everyone loses.

Regardless of any philosophical comparison, it's wrong to take lots of hard work, use it to make money and leave the creators with nothing. We therefore need to ask how we adjust the rules to work with AI.