Someone Experimented With a 1997 Processor and Showed That Only 128 MB of Ram Were Needed to Run a Modern AI
Someone Experimented With a 1997 Processor and Showed That Only 128 MB of Ram Were Needed to Run a Modern AI
Direct source:
https://blog.exolabs.net/day-4/
But before getting excited I'd encourage you to look at the screenshot of the "sleepy Joe" story they had it write...
It technically prints out words. And most of them are in plausible order, but it's also largely gibberish.
So it "works" just not useful
It technically prints out words. And most of them are in plausible order, but it's also largely gibberish.
So just like a normal LLM
So just like a normal LLM
No, much worse...
It's worth looking at the screenshot, and I feel like their choice to show a picture of the screen and not text that could easily be copy/pasted was intentional
I have ChatterUI on my phone. Running a model locally is butter smooth, except for the first response after changing models. I haven't done anything more than brief chats with it, so I don't know if it's more useful than that.
