They have the best AI (Deepseek) and now the best chips.>China’s analogue AI chip could work 1,000 times faster than Nvidia GPU>"Chinese scientists have created a superfast analogue chip that can solve complex maths problems for advanced scientific tasks and artificial intelligence while using less power than conventional computing, according to a paper published this month.">"The analogue device designed by researchers from Peking University uses memory chips made of resistive materials. With future improvements, it could perform calculations at a processing rate 1,000 times faster than top digital processors, such as the Nvidia H100 graphics processing unit (GPU), according to the team.">"Benchmarking shows that our analogue computing approach could offer a 1,000 times higher throughput and 100 times better energy efficiency than state of the art digital processors for the same precision," say researchers.https://x.com/Dr_Singularity/status/1981098681299263705
>>106983805no amount of ai neural network transformer gpt prompting transistor slop will make ai conscious, the only way forward is picrel but that's not ethical
>>106983805>CHINA NUMBA WUAN>twatter post>no source>AI generated imagethis is getting old
>For Nth time, journalistards don't understand the wording of papers
>>106983840Wanna know what is not getting old?OPHe's scheduled for organ harvesting next week.
>>106983840>being wrong
>>106983805memristors?>>106983828human brain is literally analog, there's few neuromorphic designs out there, they didn't manage to apply them for LLMs yet. but they are running various AI algorithms, quite well, with low power draw.
>>106983875I read that as>being wong
>>106983910thats because the way for more efficient and quantum relational compute is not electrical its chemical. organic chemistry solved this on its own we are just slowly catching up, if ever.captcha: ASSM2
>>106984029>chemicalthat's nonsense. chemicals mediate electric activity what the hell are you on about lolhttps://en.wikipedia.org/wiki/Ion_channelstay in school kids
>>106983805I think its https://www.nature.com/articles/s41928-025-01477-0 . And if thats true then thats a old idea it works quite well but there was the problem of programmable weights. There is a very neat application of the same idea in magnonics but sadly no one funds that field.
>could>could>couldThe next shit I take COULD revolutionize AI. Could.
>>106983805>twatter>dr singularityOpinion discarded
>>106984585Based, the chinks will want to buy your shit!
>>106984561if I'm reading that correctly, this thing is a replacement for asics for certain tasks (that can also be done using a gpu) and ai is only tangentially affected. of course all the bots here took the twitter engagement bait at face value
>>106983805Awesome! Now it can hallucinate the wrong answer one thousand times faster!
>>106983875https://www.youtube.com/shorts/4cDy0YsaYWU
>>106983805>With future improvements, it couldyeah, couldn't we all, but. But.
>>106983805>meme post>meme image>meme idea>total ignorance
>>106984751Intel already launched picrel 4 years agohttps://open-neuromorphic.org/neuromorphic-computing/hardware/I think these are spiking neural network hardware designs.
>>106989828Cool. I’ll buy one when I order my new 3D crosspoint drive.
>>106989842it's not an Intel thing, there's more players in the market. these chips already run AI algorithms for detection and stuff. they're ultra low power. the intel one is 1W power draw and has 1 million neurons and 120 million synapsesand there's other chips with lower neuron count, like 500 neurons and draw under 1mW. it's crazy low power
>>106989828That isn't analog, retard.
>>106983840really tiresome having to hide these stupid threads.
>>106989842>3D crosspointRIP, just a few years too early for AI availability. It would have dominated the current fad with further improvement to pcie5 speeds
>>106983805>with future improvements, it could perform calculations at a processing rate bajillions times faster than top digital processors>We really don't know it made be trillions of bajillions of times faster in the future.
>>106990147https://johal.in/the-hidden-challenge-neuromorphic-computing-hardware-comparison-2025-2/>Digital Spiking Processors (Intel Loihi 2, IBM TrueNorth)my bad, they're simulating them. but the neuromorphic space is going towards analog that's pretty much clear>Analog Memristive Arrays (Knowm AHaH, BrainChip Akida)>Pros: Native synaptic plasticity, sub-milliwatt operation, nanosecond response
>>106990203>https://johal.in/the-hidden-challenge-neuromorphic-computing-hardware-comparison-2025-2/>Future Outlook: 2026 and Beyond>The neuromorphic landscape will converge around hybrid architectures by 2027, combining digital programmability with analog efficiency. Stanford's recent breakthrough in memristor stability (Nature Materials, 2024) enables 8-bit precision in analog cores – previously thought impossible. Meanwhile, quantum neuromorphic prototypes from MIT demonstrate superconducting loops mimicking synaptic behavior at near-zero energy costs. The European Union's Human Brain Project predicts that by 2028, neuromorphic co-processors will handle 90% of sensory processing in autonomous systems, reducing AI's carbon footprint by an estimated 50 megatons annually.
>>106983805deepseek isn't even the only good chinese AI model vendorsure, they won, and I can't wait for the bubble to burst so I can score server hardware for cheap and run a cluster of agents at homeAGI is a meme that betting on is pure degenerate gambling, it requires more breakthroughs and transformer nets will not get there.
>>106983805>from china>could
>>106983805>China claims, it could, in theory, possibly, in the future...
The Chinese Chip That Could...
>>106983805What's so analog about it and what makes it better?
>>106992139>bubble to burstwhy is everyone saying this shit? everyone's primed to say it these days
>>106993323the memeristors probably
>>106993334AI is unironiclly propping up the US economy and the realization is starting to hit that investors have pumped billions and billions for what?Shitty 20 second slop videos?neurotic chat bots?A way for companies to maybe kill off some entry level job positions which isn't really a good thing?The level of investment only matches the most pie in the sky, post scarcity bullshit spewed by AI evangelists which is also the least likely thing to happen.
>>106993547The even funnier part about that is that companies that could actually benefit from ai in supply chain management, like Amazon, are crabbing hard or even dipping because they aren't using it sexily enough
>>106993547All those bajillions from glowie money only to be undercut by cheapskates from China.It's really over for the western sloppa.
My professor proposed a thesis for my Bachelor's about neuromorphic devices and spiking neural networks because I told him I was interested in Interaction Nets. What should I know about them? They sound really interesting.
Neurons are fragile things that pop like nothing i hope
>>106993334Retards don't understand that IA the new search engines. That IA is running their insurance They don't realise that humanoid robots are starting to be good enough for small tasks in factories and they believe what is being done now is the reflection of what the AI sector will be in 20 years. They don't realise that investors aren't betting on the current thing, they are betting on the future, and you are AstroTurfed if you are a factory/warehouse worker in the near future
>>106993776Robots will be laying pipe in 5 years people are asleep
>>106993693SNNs introduce the time dimension into the whole AI thing. atm time is not a thing for AI. information is encoded in the timing between the spikes. kinda weird
>>106983805its embarrassing watching all the zogmerican llm's totally fail on nof1.aii guess they're designed not to lead any gentiles off the plantation, more than they're designed to be smart and useful tools
>analogueEnjoy your noise, retard.
