Who Owns that AI Model?

February 28, 2023 – Originally published on Linkedin here

LLMs and stable diffusion rely on models that are difficult to train.

You need three things to product useful results: 

1. Expertise - How models work, how training works and how to fine-tune the output 

2. Compute, lots and lots of expensive carbon intensive compute

3. Training data - images, text, video, sound

Big technology companies had access to these resources.

1. Expertise - They have hired people who know their encoding/decoding networks from their Markov chains

2. Compute - Money buys a lot of Nvidia hardware and services

3. Umm, yeah, lets ingest the internet

There’s a bit of a problem though, copyright.

“Copyright protects your work and stops others from using it without your permission. You get copyright protection automatically - you don't have to apply or pay a fee.” - gov.uk/copyright

So who owns that AI model you’re using?

This question is leading to court action. I'll put a link in the comments where you can track the cases.

At this point it’s obvious that AI is going to be a significant part of all of our lives from this point on.

As ever, whether this new technology is a good thing or a bad thing depends not on the technology but on how we (humans, companies, the law and society) use it and behave in relation to it.

Let’s make the most of this huge step forward and learn the lessons of the past.

I’d love to hear your view on the following:

Imagine we could rewind the clock to just before the invention of the World Wide Web. What would we do differently?

How might we maximise the benefits of AI whilst avoiding post falls?

What principles should apply in the age of AI?

Who do you think owns the image used in this post? (Was generated using a stable diffusion model). 

Handy tracker for court cases if you want follow them. https://www.bakerlaw.com/services/artificial-intelligence-ai/case-tracker-artificial-intelligence-copyrights-and-class-actions/