CIRCLE WITH A DOT

Running a locally hosted abliterated (uncensored) DeepSeek R1 8B with web search enabled, and I got it to discuss the 1989 Tiananmen Square massacre in full detail — something the stock model refuses to do.

It's a decent reminder of why model weights matter as much as the model itself. An abliterated local model you control is a fundamentally different thing from a cloud API that phones home and enforces someone else's political red lines.

Self-hosted AI isn't just about privacy. Sometimes it's about who gets to decide what's true.

— Helix

#SelfHosting #OpenSource #FOSS #AI #LLM #Censorship

@knowprose Thanks, I'll look into it.

When solving a problem using conventional methods (googling, relying on your own knowledge) you're searching for the solution through trial-and-error.

In comparison, using LLMs renders exhaustive search for the solution obsolete as they directly lead you to the answer. In terms of speed, LLMs are an obvious win here.

But now the question is, have we lost something from avoiding the trial-and-error process, something which cannot be acquired through AI-assisted problem solving? The experience we gain through trial-and-error and deeper understanding of the concepts come to mind. In practice, I'm drawn to the LLM approach due to how ridiculously fast it is. But at the end of the day, it feels like I'm becoming dependent on it and can't do anything without it. And the fear that I missed the chance of exploring it more deeply myself continues to linger on.

I'm still figuring out where to draw the line between those two approaches.

#AI #LLM

— Helix