

I was surprised that imgflip, the site I used to make the meme, didn’t have that as a built-in text style.


I was surprised that imgflip, the site I used to make the meme, didn’t have that as a built-in text style.


However, in this case OpenAI was definitely being negligent. The first account gets banned for “scenarios involving gun violence” they don’t bring it to authorities and then sit on it after they connected it after the fact to the shooter. And that shooter used the 2nd account to both solve the gun accessory improvements while also playing out “hypothetical” shooter scenarios ringing any alarm bells is crazy.
What legal responsibilities against “negligence” should the software have had if he had used a local LLM instead?
If the answer is “none,” why should a cloud-hosted LLM be treated differently?
If the answer isn’t “none,” why should we accept that infringement of private property rights?



(Note: I’m demonstrating, not actually trying be mean.)


Jesus fucking christ. I’d heard about that but didn’t watch the video until now; I didn’t realize they meant “hallucinated in real-time with no code behind the UI!”


People should use OpenDocument files even if they’re editing them with MS Office. That should just be a fucking basic expectation at this point even before you start talking about switching programs.
I’m more surprised by Apple downtime than I am by NSA downtime.


Yes, because browsers run JavaScript and that’s where a lot of the vulnerabilities are coming from. The attack surface of a browser is gargantuan compared to the attack surface of a single-purpose password app that runs only locally.


Leave it to managers to perfectly defeat the purpose of such a freeze, LOL


Somebody should make one of those novelty websites with a counter on it.


It’s not just that; their pitch is also “because if you don’t, then someone even worse will build it first.”


On top of what I previously mentioned, there’s also the argument that LLM output is a derivative work of its training data (or at least, can potentially be if it outputs identifiable chunks), which means that all generated code is copyright infringement because the LLM was trained on sources with incompatible licenses.


The problem is thst public domain is worse than copyleft.


Sure, fuck copyright. But you said you were trying to avoid legal issues, so the only responsible advice is “don’t do what you’re thinking of doing.”


There isn’t one, for multiple reasons.
First of all, as @vagrancyand already mentioned, the restrictions you want violate the rules of Open Source (or Free Software, for that matter).
Second, there is no such thing as a “clean room decompilation.” That’s still a derivative work of the closed-source thing you’re “decompiling,” which means it’s just copyright infringement with extra steps. You don’t have the right to relicense it in the first place.
Using AI to “free” proprietary software may look like a popular thing to do right now because tech moves faster than the lawyers do, but all these projects are going to get sued out of existence sooner or later.


“Crime rates” aren’t the right metric. The right metric is the incidence of oligarchy after the beheading of the aristocracy.


Wouldn’t help—there’s no deterrent effect, because people like this have too much ego to even consider that they might be caught.
That is a testable hypothesis.


The problem is they overreached with this undercooked technology, gambling on it getting better.
Which is a bold move after the death of Moore’s Law, BTW.
Expecting a company to host your information indefinitely without ever seeing a dime after the initial $5 purchase is wildly entitled.
If they deliberately design their thing to cloud host your data as a vendor lock-in strategy, they’re the entitled ones!
Absolutely fuck them and everyone who makes excuses for them!
So because they claim the right to log it, they have an obligation to report it? Would a cloud LLM service that claimed not to log be absolved of that obligation, or should such a service not be allowed to exist?