The following podcast contains advertising.

To access an ad-free version of the LawFair podcast,

become a material supporter of LawFair at patreon.com slash law fair.

That's patreon.com slash law fair.

Also, check out LawFair's other podcast offerings,

rational security, chatter, law fair no bull, and the aftermath.

What copyright law protects is the exploitation of expression.

From the standpoint of the technologist, this looks like what we're doing is very much like what Google did.

Google was doing it from research library collections.

We're doing it from the open internet.

And web scraping is just something that people do all the time.

And therefore, it must be fair use because it's been allowed for years and years.

I'm Alan Rosenstein, associate professor of law at the University of Minnesota and senior editor at LawFair.

And this is the LawFair podcast for July 17, 2023.

The only thing more impressive than the performance of generative AI systems like GPT-4,

and stable diffusion is just the sheer volume of training data that goes into these systems.

GPT was trained on reportedly what is essentially the entire internet,

while stable diffusion and other image generation models rely on hundreds of millions,

if not billions, of existing pieces of artwork.

Of course, much of this content is copyrighted,

and the authors and artists whose work is being used to train these models,

and potentially to threaten their own livelihoods, their paying attention.

...