Xelynega

20 ms·

331.
by Xelynega 4y ago
If by "computationally heavy" you're talking about the financial overhead of hiring another layer of human filters to filter duplicates out of the dataset.
332.
by Xelynega 4y ago
If a bunch of random people on the internet see the potential copyright infringement and need for clarification, then it's doubtful makers of art generator algorithmsbdidnt think of the same potential issues and choose to ignore them.
333.
by Xelynega 4y ago
How is it that CAD programs manage to be usable to power users with graphical interfaces and keybind then? If the problem is that UI has been dumbed down to the point of being unusable for the sake of design, then I don't think adding
334.
by Xelynega 4y ago
Just like if you give your car a few decades it might learn to fly?
335.
by Xelynega 4y ago
I don't see why the options are "word salad" or "limited general intelligence". Why can't it be the statistical compression and search of large datasets that it is?
336.
by Xelynega 4y ago
It's emergent behaviour just like adding "tutorial" on the end of your Google search somehow gives you results that are more instructional, so not much to be scared about. It just so happens that chatgpt tends to generate tex
337.
by Xelynega 4y ago
If I make it sound like a sinister conspiracy or cash grab, that's because that's what it as long as it's a private entity and not a public endeavor. I don't deny that this might be a worthy pursuit or the right directio
338.
by Xelynega 4y ago
Why don't US companies pay foreign workers US wages? The answer I come up with is that they chose to hire workers in disadvantaged situations(relative to opportunity and local labour laws) to reduce operating costs. To me, using people
339.
by Xelynega 4y ago
I think your "building" and "tearing down" analogy is actually spot on. We have a whole group of people who are so concerned with what's already built and building more and more onto it, that any attempt to point ou
340.
by Xelynega 4y ago
I don't see what there is to understand. Any 'ai' model right now requires classified data to train. It takes human work to classify data. Human work is cheaper in some places than others, so companies in expensive places exp
341.
by Xelynega 4y ago
> If we have to enslave the whole world for 1 year to build an AI that elevates all for eternity. I think this shows a fundamental misunderstanding of the issues in the article. Using your analogy, the article is asking the question: &qu
342.
by Xelynega 4y ago
Let's say we restructure global labour relations such that people arent getting paid a fraction of their worth because of where they were born or choose to live. You know, since we're dealing in hypotheticals.
343.
by Xelynega 4y ago
Children working in coltan mines and on cocoa plantations are part of the same issue, so I don't understand why you take issue with that issue being brought to the forefront using 'some cool tech bandwagon'. It's a cool
344.
by Xelynega 4y ago
Why are you discrediting the authors points based off of a fluff sentence in the introductory paragraph? Typically the arguments themselves are evaluated based on their own merit. The "popular" in pop culture and the "public&
345.
by Xelynega 4y ago
When gzip turns my 10k character ASCII text file into a a 2kb archive, has it "compressed each character down to a fifth of a byte per character"? No, thats a misunderstanding of compression. Just like gzip, training stable diffus
346.
by Xelynega 4y ago
Citing fair use is putting the cart before the horse here. The debate is around whether or not the stable diffusion training and generation processes can be considered transforming the worka to create a new one in the same way we do for hum
347.
by Xelynega 4y ago
Once you start recreating copyrighted works from hashes this analogy becomes relevant, until then how can you compare the two when the distinguishing feature is its ability to reproduce the training data.
348.
by Xelynega 4y ago
There only needs to be an answer if it's determined that some number isn't copyright infringement. The easy answer would be to say that the process is what prevents the works from being transformative(and thus copyrightable) and n
349.
by Xelynega 4y ago
You took the images, encoded them in a computer process, and the result is able to reproduce some of those images. I fail to see why the size of the training set in bytes and the size of the model in bytes matters. Especially if, as other c
350.
by Xelynega 4y ago
> No semblance of the original image even remotely exists in the model What does this mean? It doesn't mean you can't recreate the original, because that's been done. It doesn't mean that literally the bits for the im
351.
by Xelynega 4y ago
"It's very unclear" in legal cases is synonymous with "it hasn't been challenged in court yet". You say they're moderately confident because they're fairly reputable, but remember that Madoff was a &q
352.
by Xelynega 4y ago
If I host a song I don't have license to on my website I'm violating copyright by distributing it to you when you listen on my site. If my parrot recites your song after hearing it and I record that and upload to YouTube. I'v
353.
by Xelynega 4y ago
I'd assume the precedent would be about sharing encoder data, which would be covered in bittorrent cases. "Training a neural network" is an implementation detail. These companies accessed millions of copyrighted works, encode
354.
by Xelynega 4y ago
So it's fine to distribute copyrighted works, as long as they're jpeg(lossy) encoded? I don't think the law would agree with you.
355.
by Xelynega 4y ago
Not really, since one of the major characteristics is being able to recover the copyrighted work from the encoded version. Since md5 hashes don't share this property, they're not "in that vein".
356.
by Xelynega 4y ago
The real tragedy is being marketed to so heavily that we construe enforcing copyright on llm/diffusion companies with shutting down an AGI. I blame companies like openai purposefully marketing themselves poorly since nobody is going to
357.
by Xelynega 4y ago
All of those examples have the exact same limitation, they can't replace processes that create novel ideas since they're statistical recombinations of existing ones.
358.
by Xelynega 4y ago
If the file that were previously honored as consent to use the copyright material is subsequently ignored, wouldn't the content creators take the indexers to court for copyright infringement?
359.
by Xelynega 4y ago
> ChatGPT output is generally not similar enough to any _particular_ source document to establish the fact that it's derivative. Isn't this exactly what a court case would be trying to clarify? If so wouldn't assuming this
360.
by Xelynega 4y ago
The bottleneck would be the number of workers on sites like mechanical turk available to create the datasets. Might take a few more years before amazon and facebook get enough third world countries to the point they can exploit their labour

More ›