3 ms·
You don’t find the stealing argument logically sound because you immediately frame the theft as “analyzing” to suit your own narrative and then demand people en
by bugglebeetle 2y ago
You don’t find the stealing argument logically sound because you immediately frame the theft as “analyzing” to suit your own narrative and then demand people engage with it, while proceeding to make further spurious claims like…
> I would be more convinced if AI used a fixed database during generation
Wow, I didn’t know that model weights, an elaborately compressed form of their training data, rewrote themselves every time they were invoked. Or that it’s only theft if I stole data from a fixed database to build my own service.
- ronsor 2y agoAI training is literally analyzing. That is how it works. Properly trained models (i.e., ones that aren't overparameterized or overfit) do not just "elaborately compress" training data as this is not possible. For example, you cannot compress 1 billion images into 1 billion parameters, and expect to retrieve them later. If objective facts are "my own narrative", then no rational discussion can occur.
- bugglebeetle 2y agoOh well, you should tell the folks at DeepMind and Meta about these objective facts then so they don’t waste any more time doing research: https://arxiv.org/html/2309.10668v2 https://arxiv.org/html/2309.10668v2 Maybe apply for a job there too, since you’re obviously so far ahead of everyone in understanding this problem space.
- YurgenJurgensen 2y agoYou absolutely can compress a subset of a billion images into a billion parameters if you throw out all but a thousand. Is it no longer copyright infringement if you also run enough irrelevant data through your algorithm alongside the images you’re stealing?