mastodon.social is one of the many independent Mastodon servers you can use to participate in the fediverse.
The original server operated by the Mastodon gGmbH non-profit

Administered by:

Server stats:

366K
active users

Mickey Mouse (the Steamboat Willie version) is now public domain on the east coast of the US.

@andyb a friend sent me this as well:

@mmasnick @andyb

Ironic that they're using copyrighted text and images without a license for their training data.

@alsoknownas @mmasnick @andyb

Artists don't create a transform of the training data and perform a search on the result.

@resuna Sure they do. They just do it with biological neurons instead of silicon and copper pathways.

@resuna Sure we do. Pretty much all art is a remix of previously explored techniques mixed with the artist's experiences. It's not that much different than AI mixing trained techniques with human prompts.

@mmasnick @andyb

Fair use doesn't mean "I can ignore copyright by saying magic words".

@resuna @andyb fair use does say that scanning to build a useful tool is allowed and not infringing. even for big corporate entities. no one is saying magic words. i can quote you cases if you'd like. but I feel like you don't care about facts, just emotions.

@resuna @andyb Both HathiTrust and the Google Books cases involved scanning of tons of copyright-covered books, and both were found to be fair use, even for Google, a large giant tech company, which used it to build a book search tool.

@mmasnick @andyb

Parody text and bad image generation is not the same kind of social good as Google Book search.

@resuna @andyb i mean, now you're moving the goalposts.

I pointed out that scanning large corpuses of copyright-covered data for commercial reasons has been found to be fair use, and you demanded citations.

I gave you them.

Now you're arguing something different.

It is difficult for me to believe you are engaging in this conversation honestly when you do that.

Furthermore, reducing all generative AI to "parody text and bad image generation" more or less confirms your bad faith.

@mmasnick @andyb

Don't be silly... in neither of those cases did they just go ahead and say yeah you can do that it's okay. There needed to be a significant social good to allow the scanning as fair use. You are super simplifying these cases to try to argue that basically anything that you want to be allowed is fair use.

It's not clear that spicy autocomplete is even a good at all.

@resuna @andyb I simply pointed out that scanning, even for corporate, for profit reasons, can be fair use.

If you want to go further into the details of why this particular scanning is likely fair use, that's a different discussion. But suffice it to say, it is very likely fair use as well.

The scanning is quite obviously transformative. And if it's not, you'd destroy all sorts of useful data mining initiatives & research, not to mention things like screen readers for the blind.

@mmasnick @andyb

You stated flat out that it was fair use, and that's not true. It might be. Japan thinks so. But it will take more than your flat assertion to establish that. And in the meantime for ChatGPT to be restricted from parodying Mickey Mouse, the poster child for excessive copyright, is ironic AF.

@mmasnick I clicked on the link, but it was VERY slow to render. Tomorrow (later today)...

Happy New Year, Mike! ✌️

@mmasnick Here’s wishing everyone a happy H013!