Very informative read, thank you. Lets wait for Openb-R1 to be able for download, and use that time to check the machine’s code for bugs (likely, every larger software has them!), backdoors (can never be excluded as a possibility), and ways of further optimization.
I have to admit that their idea to “milk” DeepSeek-R1 for its own reasoning data is intriguing. I wonder how early in that training process the political bias has gotten its foot into the door. Or is this a late-stage filter?
just like for american models, the political bias is irrelevant. Realistically you are using the model for its reasoning capabilities, not its answer to “what happened in Tiananmen”.
I don’t agree with the point that an artificially induced bias is in any way irrelevant. And the tales of “reasoning” capabilities are quite overblown, imho.
Roflmasterbigpimp accused the Chinese company of spying.
I pointed out that the Chinese company can’t spy becaause it’s model was open source and could be locally run, while the US company set up its operation to allow it to spy on any use of its model.
Mubelotix claimed ChatGPT made their model public, which would only be relevant to the conversation as a evidence that the US is not spying either.
Well, there is an Android client that sends keystrokes (and loads of other data) back to Chinese servers. Which very much fulfils my definitions of spying.
ChatGPT was only released as SaaS, every thing you use it for goes through OpenAI’s servers.
Deepseek was open-sourced, you can run it on a local machine where it is physically impossible for China to spy on you.
They also released how they trained Deepseek, so you could even make your own Deepseek, as these guys are doing.
Very informative read, thank you. Lets wait for Openb-R1 to be able for download, and use that time to check the machine’s code for bugs (likely, every larger software has them!), backdoors (can never be excluded as a possibility), and ways of further optimization.
I have to admit that their idea to “milk” DeepSeek-R1 for its own reasoning data is intriguing. I wonder how early in that training process the political bias has gotten its foot into the door. Or is this a late-stage filter?
just like for american models, the political bias is irrelevant. Realistically you are using the model for its reasoning capabilities, not its answer to “what happened in Tiananmen”.
I don’t agree with the point that an artificially induced bias is in any way irrelevant. And the tales of “reasoning” capabilities are quite overblown, imho.
Of course chatgpt is saas, because it’s a service built on top of a gpt model, which they made public
The ChatGPT4 model is not public.
That’s the point.
Roflmasterbigpimp accused the Chinese company of spying.
I pointed out that the Chinese company can’t spy becaause it’s model was open source and could be locally run, while the US company set up its operation to allow it to spy on any use of its model.
Mubelotix claimed ChatGPT made their model public, which would only be relevant to the conversation as a evidence that the US is not spying either.
I said the model isn’t public.
Well, there is an Android client that sends keystrokes (and loads of other data) back to Chinese servers. Which very much fulfils my definitions of spying.
The app uses Deepseek’s servers. It physically could not function if it didn’t send your input data to their servers.
What other data does it send?
It is not about sending just the prompt. It’s about sending keystrokes, which goes beyond input to the current app.
Keystrokes you put into the app is useful for the devs to understand user intentions. But you saaid “loads of other data”.
Answer by question first, what “loads of other data”? Did you just make that up?
Post link to GPT3 or GPT4 model download