Data Poisoning could be a tool we use to identify AI that has used copyritten material

5 months ago by ekZepp to c/fuck_ai

Auf YouTube findest du die angesagtesten Videos und Tracks. Außerdem kannst du eigene Inhalte hochladen und mit Freunden oder gleich der ganzen Welt teilen.

Data Poisoning could be a tool we use to identify AI that has used copyritten material, or we use it to mess with AI.

https://www.vice.com/...

https://djmag.com/...

https://mosis.eecs.utk.edu/harmonycloak.html

https://mosis.eecs.utk.edu/...

https://www.eff.org/cyberspace-independence

cypherpunks 30 points 5 months ago

identify AI that has used copyrighted material

but, that is basically all modern "AI".

(the only LLM i've heard of which actually claims that its training corpus is freely licensed is Apertus...)

save

path: 0 23767467, hotness: undefined, score: 30, children: 8

youcantreadthis 8 points 5 months ago

We callin it Plagarized Information Stochastic Stupidity now the only PISS you've heard of

save

path: 0 23767467 23767731, hotness: undefined, score: 8, children: 0

Hackworth 4 points 5 months ago

Adobe claims to only train their image generator, Firefly, on images from their stock library.

save

path: 0 23767467 23767985, hotness: undefined, score: 4, children: 1

very_well_lost 7 points 5 months ago

lol

save

path: 0 23767467 23767985 23769214, hotness: undefined, score: 7, children: 0

cloudskater 3 points 5 months ago

Even if it didn't use copyrighted stuff, the concept of "generative" AI is fascist to its very core.

save

path: 0 23767467 23801611, hotness: undefined, score: 3, children: 1

HaraldvonBlauzahn 1 point 5 months ago

the concept of “generative” AI is fascist to its very core.

Can you explain? I might miss the connection.

save

path: 0 23767467 23801611 23805613, hotness: undefined, score: 1, children: 0

YourMomsTrashman 2 points 5 months ago

Traditionally, with machine learning, it is standard practice to mention what datasets and/or pretrains were used, so that the results are transparent and can be replicated. With GPT-2, it was "the common crawl and our own crawled 8 million web pages", and since then I feel it's mostly left out, falling back on (easily manipulated) benchmarks instead 😬

save

path: 0 23767467 23789524, hotness: undefined, score: 2, children: 2

cypherpunks 2 points 5 months ago

Yep. But just providing a list of millions of URLs and saying "we trained on this" as some models in the past have done also didn't make it possible to replicate; by the time anyone re-fetches them all, many of the URLs will inevitably have changed or disappeared.

save

path: 0 23767467 23789524 23789615, hotness: undefined, score: 2, children: 1

YourMomsTrashman 1 point 5 months ago

That's exactly why projects like the common crawl exist though !

save

path: 0 23767467 23789524 23789615 23792989, hotness: undefined, score: 1, children: 0

very_well_lost 14 points 5 months ago

People have actually been doing this to catch plagiarism for centuries, long before LLMs were a thing.

See trap streets for one of the better known examples.

save

path: 0 23767874, hotness: undefined, score: 14, children: 1

LodeMike 3 points 5 months ago

I learned about that from Doctor Who!

save

path: 0 23767874 23775130, hotness: undefined, score: 3, children: 0

sidefaceturdtalker 4 points 5 months ago

One way to push back forsure but we need to refresh the tree of liberty asap

save

path: 0 23773394, hotness: undefined, score: 4, children: 1

ZDL 1 point 5 months ago

Interestingly, literally zero of the people I've seen who word things this way ever seem to volunteer to be the ones doing the watering. Are you going to break the losing streak or are you going to continue confirming my belief that it's only chicken hawks who say this?

save

path: 0 23773394 23780212, hotness: undefined, score: 1, children: 0

ExtremeDullard 2 points 5 months ago

This idea is as old as books.

save

path: 0 23774890, hotness: undefined, score: 2, children: 0