2 ms·

Nah, it's great for things that Western models are censored on. The True Hacker will keep an Eastern and Western model available, depending on what they need in

Nah, it's great for things that Western models are censored on. The True Hacker will keep an Eastern and Western model available, depending on what they need information on.

hmottestad 2y ago
I tried to ask it about Java exploits that would allow me to gain RCE, but it refused just as most western models do. That was the only thing I could think to ask really. Do you have a better example maybe?
- OkGoDoIt 2y ago
  Adult content and things like making biological/chemical/nuclear weapons are the other main topics that usually get censored. I don’t think the Chinese models tend to be less censored than western models in these dimensions. You can sometimes find “uncensored“ models on HuggingFace where people basically finetune sensitive topics back in. There is a finetuned version of R1 called 1776 that will correctly answer Chinese-censored questions, for example.
theturtletalks 2y ago
Wouldn’t they just run R1 locally and not have any censorship at all? The model isn’t censored at its core, it’s censored through the system prompt. Perplexity and Huggingface have their own versions of R1 that is not censored.
- hmottestad 2y ago
  I tried R1 through Kagi and it’s similarly censored. Even the distill of llama running on Groq is censored.
  - theturtletalks 2y ago
    Kagi may be using the official DeepSeek API and not hosting the model itself. There is work being done to make it completely uncensored: https://github.com/huggingface/open-r1 https://github.com/huggingface/open-r1 https://ollama.com/huihui_ai/deepseek-r1-abliterated https://ollama.com/huihui_ai/deepseek-r1-abliterated I was mistaken though, it is more than just a system prompt causing the censorship.
    - hmottestad 2y ago
      Kagi uses R1 through Fireworks.ai, Together.ai and Groq. https://help.kagi.com/kagi/ai/llms-privacy.html https://help.kagi.com/kagi/ai/llms-privacy.html
rsoto2 2y ago
a lot of the safety around models seems to be implemented in the browser. Underneath the models seem pretty easy to fool/jailbreak.

Nah, it's great for things that Western models are censored on. The True Hacker will keep an Eastern and Western model available, depending on what they need information on.

hmottestad 2y ago

I tried to ask it about Java exploits that would allow me to gain RCE, but it refused just as most western models do. That was the only thing I could think to ask really. Do you have a better example maybe?

OkGoDoIt 2y ago
Adult content and things like making biological/chemical/nuclear weapons are the other main topics that usually get censored. I don’t think the Chinese models tend to be less censored than western models in these dimensions. You can sometimes find “uncensored“ models on HuggingFace where people basically finetune sensitive topics back in. There is a finetuned version of R1 called 1776 that will correctly answer Chinese-censored questions, for example.

theturtletalks 2y ago

Wouldn’t they just run R1 locally and not have any censorship at all? The model isn’t censored at its core, it’s censored through the system prompt. Perplexity and Huggingface have their own versions of R1 that is not censored.

hmottestad 2y ago
I tried R1 through Kagi and it’s similarly censored. Even the distill of llama running on Groq is censored.
- theturtletalks 2y ago
  Kagi may be using the official DeepSeek API and not hosting the model itself. There is work being done to make it completely uncensored: https://github.com/huggingface/open-r1 https://github.com/huggingface/open-r1 https://ollama.com/huihui_ai/deepseek-r1-abliterated https://ollama.com/huihui_ai/deepseek-r1-abliterated I was mistaken though, it is more than just a system prompt causing the censorship.
  - hmottestad 2y ago
    Kagi uses R1 through Fireworks.ai, Together.ai and Groq. https://help.kagi.com/kagi/ai/llms-privacy.html https://help.kagi.com/kagi/ai/llms-privacy.html

rsoto2 2y ago

a lot of the safety around models seems to be implemented in the browser. Underneath the models seem pretty easy to fool/jailbreak.