[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / s / t / u / v / vg / vm / vmg / vr / vrpg / vst / w / wg] [i / ic] [r9k / s4s / vip] [cm / hm / lgbt / y] [3 / aco / adv / an / bant / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / pw / qst / sci / soc / sp / tg / toy / trv / tv / vp / vt / wsg / wsr / x / xs] [Settings] [Search] [Mobile] [Home]
Board
Settings Mobile Home
/pol/ - Politically Incorrect


Thread archived.
You cannot reply anymore.


[Advertise on 4chan]


File: HNffGCKaUAABp6k.jpg (85 KB, 1266x772)
85 KB JPG
Chinese AI is only 6 weeks behind, and soon it will be far ahead. Open source is the best timeline.
>>
>>539191844
china is always behind because they can't create, can't innovate, can't ponder anything novel, it's also why all you're capable of is spamming the catalog with the same tiresome threads over and over

uncreative, insufferable, lesser
>>
>>539191844
All that weighting made western AI retarded. Jews really did a number on em.
>>
File: tenor.gif (736 KB, 498x280)
736 KB GIF
>>539192038
>china is always behind because they can't create, can't innovate, can't ponder anything novel
>>
>>539192038
Their games are weirdly creative
>>
>2.8 trillion parameters
Wouldn't that require around 5,600 gigabytes of VRAM to load at BF16, nevermind the amount needed to run inference on?
>>
>>539191844
>yesterday it was ahead
>now it's 6 weeks behind
>last year it was deepseek that was ahead
memeflag niggers get the rope
>>
>>539192722
It uses 4bit floating point so more like 1.4TB for loading the weights. Also it uses a mixture of experts model where main AI delegates tasks to around 900 different expert AIs. Only 10-20 experts are picked for each query and the activations typically touch only 50b parameters per query.
>>
Behind what, to do what? We have invented a carburetor but no engine.
>>
>>539192038
There are sophisticated yet easy to run analysis tools that can tell with >95% confidence whether a model was frontier (i.e. trained on raw dataset) or was distilled. If you have any evidence whatsoever that major Chinese models are distilled from American competition then feel free to post that. The weights are open source and maths does not lie. Post your evidence..
>>
>>539192467
I have a chink game. It's full of loli. It's completely self aware and un-apologetic about it.
>>
>>539194131
Amazing..
>>
File: qwen2.png (33 KB, 890x813)
33 KB PNG
it is beyond parody that BOTH leading US labs are led by jews
if mutts could compute shame, they would collectively shart themselves to death just to wash it away
>>
>>539195153
Bunny
>>
File: 1784060630116224.jpg (368 KB, 1024x1024)
368 KB JPG
>>539191844
>behind on muh arbitary amerimutt index
kek subhuman amerimutt shills on their VPNs will parrot mutt regime propaganda lies and hollow cope buzzwords about muh coppy, muh stolen weights and muh distills, while amerimutt slop AI is literally censored and wont even return real answers for medical and chemical and nuclear research that KIMI K3 can answer because its obviously not depending on muttslop vapor ware models

>>539192722
not all parameters are and need to be activated at the same time, just like your brain doesnt need to activate your neurons about mapple syrup when someone asks if its true that if you lose you win
>>
File: if i dont steal it.jpg (11 KB, 276x182)
11 KB JPG
>>539192038
>lazy and unoriginal muttoid cope pasta
the irony of these projections

>>539194100
muttoids just lie, its their second nature, its all these thieving stealing and scummy beasts have left besides their lack of morals and souls



[Advertise on 4chan]

Delete Post: [File Only] Style:
[Disable Mobile View / Use Desktop Site]

[Enable Mobile View / Use Mobile Site]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.