Genuine question, but what do people actually use LLMs for? I've installed Ollama and tried to use it to generate some alt text with some Gemini thing, but what I got out ranged from meh to mostly unusable.
post
Oh good they scored 2.2 privacy units in privacy. How informative. This isn't a graph this is just some coloured rectangles with arbitrary text printed on them.
What measuring scale is this in, is it a recognised SI unit?
What does the "transperancy" correspond to? It doesn't look like a data collection thing.
What do the scored categories actually mean? How were they scored? How was the data sourced?
Disappointed to see this image getting upvotes as-is. Without proper context, itโs barely a step above a Facebook boomer meme in terms of credibility.
Meh. It might be better than the others in terms of privacy, but that doesn't mean it's really private.
The most private solution is to self host it or not use it at all.
And Ollama then? Kobold.cpp? Llamafile? Those might arguably be better.
Those are neither companies nor models
GPโs list contains products/services. I believe they are asking how do self hosted products compare. Also ollama is transitioning to a paid product and likely have privacy questionable functionality embedded in newer versions. It would be good to compare these available options as well.
The actual answer:
Deepseek running local. You don't have to give ANY data away.
Doesn't DeepSeek still censor sensitive questions (e.g. about Taiwan and the Tiananmen protests) when ran locally and offline?
Although it's certainly got its merits (being FOSS? and much more energy-efficient than most models), censorship of state violence is still bad.
I wonder how other chat AIs do, concerning that censorship. Grok we already know to be shit, it warps the truth beyond words. Even regarding non-political stuff, its accuracy is trash compared to other models.
How about ChatGPT? Ollama? Kobold? Llamafile? Do they censor stuff?
I use Protons Lumo from time to time, havent figured out in which cases it performs better or worse compared to mistrals LeChat but i expect they would rank high if they were part of this research. Do think ChatGPT remains superior in many other ways, unfortunately.
I second that. Lumo is private. Quality is not gpt, but Mistral is kind of shit too and I don't like their ui.
My plan was to make my own private one, now with lumo I don't know. I'd still like to tinker with it, but maybe later. An option is also Venice AI that's anonymous and pretty good, it's crypto though.
I believe Lumo uses Mistral and some other models combined under the hood
I'm guessing the scale is out of 10
Why guess if there's a labelled axis that goes to 16?
all 30 comments