Okay holy fuck. I just updated to the latest LLaMA 2 model on my phone and asked it an open ended prompt about the Covid lockdowns. No loaded questions.
The first line is my prompt, the rest is the response:
Discussion
Wow! Pure true and awesome that you can run this in your phone 👀
For real this is 100% local on standard consumer smartphone hardware! No cloud.
Fuk will love to try it, some guide around to do that?
I installed it in 2 minutes on linux!
Check this out: https://github.com/antimatter15/alpaca.cpp#building-from-source-macoslinux
Sorry this one was he using i guess
If you're good with Linux it's easy to setup, I even compiled it entirely on my Android phone.
Nice! Ill love to try, what repo are you using?
Standard LLaMA open source implementation, it's come a loooong way now!
Seems nice, I don't know if its anything Alex Jones wasn't saying like the whole time though. It tickles our sense of novelty to see an AI do this, but the only reason this wouldn't be the baseline expectation out of chatgpt is because its censored from dealing with a fairly obvious reality. Idk, its a fun read, I'd rather use this for generating content than others, but I wouldn't expect any insights out of it.
You are correct it's only notable because ChatGPT is so heavily censored.
But I still think it's significant for exactly that reason. Closed source AI that is programmed with the bias of its creators is going to see mainstream use. It already is, really.
The fact the standard LLaMA 2 model without any special modifications offered up a real answer without a loaded prompt gives hope for the future of actual open AI that isn't just programmed to be a propaganda machine.