if you are china how would you beat claude code?
DistillationBased chinks going to kill AI with their IP thieving ways and there's nothing sam shitman can do about it
no need. being second best is good enough, just like their military. in fact they are number one in terms of efficiency (what you get per dollar), both with their AI models and military. now they are playing the long game, or as /pol/ would say, do nothing.
>>108491918I have been following, and qwen has it's own coder agent thing. So, we will see.
>>108491918> just train an llm on the output of claude. > release it as open source.>..>NOOO You stole the stuff we stole first, dirty thieves!
>>108492010>now they are playing the long game, or as /pol/ would say, do nothing.Chinese strategy of non-action. As a neet I feel it aligns with my own thinking very much.
>>108492832I don't know much about neural networks (though seemingly fewer and fewer people in the field actually do), but surely that method wouldn't produce a good copy of a model unless you do try every possible input (impossible) several times over.
>>108491918they dont need to now since it was fully leaked lmao
>>108491918Soon enough it won't matter because the difference between number 1 and number 2 will be impossible to notice. That's already the case for everyone who's just using the bots for casual chats.
LLMs are already outdated, you guys are always one step behind.
>>108491918Seems like Claude is beating itself right now, eating up all tokens in just a few prompts even for paying users
>>108491918Get people hired at Anthropic, use them to find supply chain attack vulnerabilities, exploit and exfiltrate. My people on the inside are never implicated so they can just keep doing it.
>>108491918>if you are china how would you beat claude code?by not playing in the AI race watching the competition hang their own rope. all while we're focusing production of longterm meaningful endeavors
>>108491918Develop actual self improving AGIs, which doesn't need thousands of experts with high 6 figure salaries to judge chain of thought traces.That's the main problem, early models just needed a ton of third worlders for RLHF. Now you need experts.
>>108491918Ask Deepseek ig.
