Deep learning doesn’t stop at llms. Honestly, language isn’t a great use case for them. They are—by nature—statistics machines, so if you have a fuck load of data to crunch, they can work very quickly to find patterns. The patterns might not always be correct, but if they are easy to check, then it might be faster to use them and modify the result compared to doing it all yourself.

I don’t know what this person does, though, and it will depend on the specifics of the situation for how they are used.

It looks like a research group found a security vulnerability that they then used to find a single common key in all of the cards made by this company. The second part here is a reasonable concern, but the article calls the vulnerability a backdoor in the beginning, which I think is fairly misleading.

Personally, I want FDA approval to mean it is as provably safe and effective as possible. They said they wanted more evidence before approval, and I think that’s okay. Good, even.

The way the US treats recreational drug use and self-medication is horrific, but it’s not really the domain of the FDA.

I would rather have not-yet-FDA-approved legal-for-personal-use mdma than what we have now or an unproven drug approved by the FDA.

Tracing is fine if you use it to learn how to draw. It’s not fine if it ends up in the finished product. Determining if it ends up in the finished product with AI either means finding the exact pattern in the AI’s output (which you will not), or clearly understanding how AI use their training data (which we do not)

It depends on how much you compress the jpeg. If it gets compressed down to 4 pixels, it cannot be seen as infringement. Technically, the word cloud is lossy compression too: it has all of the information of the text, but none of the structure. I think it depends largely on how well you can reconstruct the original from the data. A word cloud, for instance, cannot be used to reconstruct the original. Nor can a compressed jpeg, ofc; that’s the definition of lossy. But most of the information is still there, so a casual observer can quickly glean the gist of the image. There is a line somewhere between finding the average color of a work (compression down to one pixel) and jpeg compression levels.

Is the line where the main idea of the work becomes obscured? Surely not, since a summary hardly infringes on the copyright of a book. I don’t know where this line should be drawn (personally, I feel very Stallman-esque about copyright: IP is not a coherent concept), but if we want to put rules on these things, we need to well-define them, which requires venturing into the domain of information theory (what percentage of the entropy in the original is part of the redistributed work, for example), but I don’t know how realistic that is in the context of law.

Saying that statistical analysis is derivative work is a massive stretch. Generative AI is just a way of representing statistical data. It’s not particularly informative or useful (it may be subject to random noise to create something new, for example), but calling it a derivative work in the same way that fan-fiction is derivative is disingenuous at best.

Was the aurora once-in-a-lifetime? It seemed like typical-ish solar maximum stuff, which happens every 20 or so years. Was there something special about those specifically?