When are we gonna get a completelly new paper that totally changes the LLM game and creates a 500000X improvement?
Never.
two more weeks
>>109566612Soon but we will need another 2 trillion in funding from you goy.. i mean taxpayers. You wouldnt want CHINA to win would you goy?
>>109566630Yes you will never write a paper that will do this, utterly impossible for you (you have no say in the matter and never will btw)
>>109566612Just take the original transformer model paper and put it into Claude or ChatGPT and ask it to improve it and make no mistakes
>>109566612whenever researchers are allowed to cover guided training instead of llm money pits.almost every form of ai is based on a neural net, which is based on actual biological nerve structures. yet, despite the very obvious issue of required scale, they don't bother to properly guide the training and instead let it loose on hundreds of terrabytes of mostly uncurated data, often not even labeled.once researchers receive funding for research into guided training or neural net alternatives we'll see it in no time.this is ignoring the fact that neural nets are incredibly inefficient on computer hardware. we should be using other more non-uniformly structured algorithms that take advantage of binary computer architecture rather than something based on an electrochemical system.
>>109566612i'm working on it saar
