Isn't that what the tool is proposing to do? It profiles your python code and finds opportunities (sections of code that run poorly) and makes suggestions for libraries or algorithms to improve it? In theory proposing libraries written in C, perhaps?
Long story short, the author is wrong, and will become even more wrong over time. Cloud based LLMs like ChatGPT are not the cheapest to run right now, but it's not like they aren't economical either. With each passing week, advances are made in these models to make them available for local use, provide more powerful customization features, and make them run on less and less hardware.
I mean, anyone who claimed that either doesn't understand how they work, or is dealing with pedantry.
The probability of the next token is determined by training data. Given a specific enough context, the next token becomes whatever it was in the training data for that context.
The problem with our legislative system is that everyone and their mother will do everything they can to work the system to favor only themselves.
So now you have to rigorously define what "stolen" means.
What's the solution for a non-tech person to achieve the same results though? If I've bought a wifi security camera and I want motion alerts while I'm not at home, that device has to somehow be able to talk to my cell phone from my home network without anyone sitting between them. How does the cell phone app know my home network address? How does it remain secure? What approaches can be taken that aren't cloud based but still easy enough for grandma to plug into the wall and have it work without having to type in a home IP address, etc? What approaches can be taken that aren't cloud based but still easy?
I don't have an answer myself. I do all my own set up and non-cloud based things, but I have the skills. It's not "easy" for a lay-user to do so.
Send it to me then, luddite.
There is no argument that Unity does or does not need to find a practical monetization method.
But this wasn't it.
Ok boomer
Then piracy is ethical.
It's called extrapolation, nitwit.
You seem to be under the impression that if a solution isn't 100% perfect, then it's useless. You see this same nonsense every time someone mentions anything related to security. It's incredibly foolish.
You seem to be under the impression that if it's not 100% useless then it's worth pursuing. The idea of watermarking text is just stupid. It will accomplish so little as to be a waste of time. Depending on how they try to do it, it could also just end up hurting the models in question. We already see evidence of human guidance making the models dumber.
There are malicious actors and non-malicious actors. Criminals fall into the 'malicious actors' category. There is little reason for non-malicious users to want to remove invisible watermarks. It doesn't hurt them, and they benefit from being able to easily identify some generated content. For malicious actors, such as criminals and hostile state actors, having their generated content watermarked isn't in their best interest. Consequently, most people who would want disable watermarking or remove existing watermarks are going to be criminals and hostile state actors.
If you're not doing anything wrong, you have nothing to worry about, wink wink. I don't think you're on the right website here friend. Most of us here value privacy for privacy's sake. To wit, you're wrong that people removing watermarks must be criminals or hostile state actors. They're also people who understand that knowledge is power and that giving unknown 3rd parties excessive amounts of information isn't beneficial to them.
Where is MY companion robot?!
noob
In this case, closed source means in the cloud. The security vulnerabilities they're talking about here are malicious code being embedded in python pickle formatted checkpoints.
tl;dr: Don't download stable diffusion, pay us, the trustworthy saviors of humanity, OpenAI to use our image generator instead.
What they're probably referring to is something called Instruct training. A large language model is at it's root great for text completion. Give it an incomplete document and it will try to complete it.
Instruct fine tune training shifts that focus from "Complete this sentence to" "Treat this sentence as instructions" (Or more specifically, a sentence shaped like this is completed with instructions or information shaped like that) which is what ChatGPT used and is what makes it actually useful. You'll find some advances here soon I'm willing to bet as they figure out they don't need 175b parameters for this style of training to be useful.
As long as we're going to reinvent the wheel again, we might as well try making it round this time. - Mike Dennison
