A_A
@A_A@lemmy.world
- Comment on Spain announces plans to ban social media for under-16s 1 week ago:
Better yet : ban Facebook and the likes for anyone younger than 40 … and also for anyone older than 39.
Do also Reddit, xTwit, microSlop and so on. - Comment on Shanghai scientists create computer chip in fiber thinner than a human hair, yet can withstand crushing force of 15.6 tons — fiber packs 100,000 transistors per centimeter 1 week ago:
Ooops, you are right : i should have seen this digital print giving scale of the image 😆
- Comment on Shanghai scientists create computer chip in fiber thinner than a human hair, yet can withstand crushing force of 15.6 tons — fiber packs 100,000 transistors per centimeter 1 week ago:
that image is under a microscope … it is 50 micrometer while a human hair is in the range of 100 micrometer.
- Comment on Shanghai scientists create computer chip in fiber thinner than a human hair, yet can withstand crushing force of 15.6 tons — fiber packs 100,000 transistors per centimeter 1 week ago:
Sny single hair will withstand a rubber truck tire rolling on it the way they describe their crushing test, so, that is a bullshit statement … yet better to get the abstract from Nature (science journal)
spoiler
Published : 2026 January 21 Fibre integrated circuits by a multilayered spiral architecture www.nature.com/articles/s41586-025-09974-0 Abstract Fibre electronic devices are transforming traditional fibres and garments into new-generation wearables that can actively interact with human bodies and the environment to shape future life1,2,3,4,5. Fibre electronic devices have achieved almost all of the desired functions, such as powering6,7, sensing8,9 and display10,11 functions. However, viable information-processing fibres, which lie at the heart of building intelligent interactive fibre systems similar to any electronic product, remain the missing piece of the puzzle12,13,14,15. Here we fill this gap by creating a fibre integrated circuit (FIC) with unprecedented microdevice density and multimodal processing capacity. The integration density reaches 100,000 transistors per centimetre, which effectively satisfies the requirements for interactive fibre systems. The FICs can not only process digital and analogue signals similar to typical commercial arithmetic chips but also achieve high-recognition-accuracy neural computing similar to that of the state-of-the-art in-memory image processors. The FICs are stable under harsh service conditions that bulky and planar counterparts have difficulty withstanding, such as repeated bending and abrasion for 10,000 cycles, stretching to 30%, twisting at an angle of 180° cm−1 and even crushing by a container truck weighing 15.6 tons. The realization of FICs enables closed-loop systems in a single fibre, without the need for any external rigid and bulky information processors. We demonstrate that this fully flexible fibre system paves the way for the interaction pattern desired in many cutting-edge applications, for example, brain–computer interfaces, smart textiles and virtual-reality wearables. This work presents new insights that can promote the development of fibre devices towards intelligent systems.
- Comment on Distinct AI Models Seem To Converge On How They Encode Reality 4 weeks ago:
i see your point and admire your resolute : there is a lot to be done without AI.
Unfortunately technology is a one-way trap in my (par time intellectual’s) opinion. Also, on the one hand, we have multi billionaire_CEOs who do more to destroy the environment with help from technologies … and on the other hand, you could find scientists (with small to medium resources) discovering new medications and impressive stuff nowadays by & through, in part, the use of AI.
We live in a complex world in which a diversity of approaches is quite valuable. So thanks for your standing and your contributions in society and at Lemmy.
- Comment on Distinct AI Models Seem To Converge On How They Encode Reality 4 weeks ago:
from the article :
“In a 2024 paper, four AI researchers at the Massachusetts Institute of Technology argued that these hints of convergence are no fluke.”
Now, publish your own findings. No ? i didn’t think so either.
- Comment on The Kimwolf Botnet is Stalking Your Local Network 5 weeks ago:
12 days ago :
lemmy.world/post/40529739
Massive Android botnet Kimwolf infects millions, strikes with DDoS17 days ago :
lemmy.world/post/40316169
史上最疯：独家揭秘感染全球180万Android设备的巨型僵尸网络Kimwolf | The craziest thing in history: Exclusive reveal of the giant botnet Kimwolf that infected 1.8 million Android devices worldwidehummm …
- Comment on New Jolla phone and Sailfish 5 offer a break from iOS-Android monotony 1 month ago:
expensive and useless, unfortunately.
- Comment on How Transformers Think: The Information Flow That Makes Language Models Work 1 month ago:
[feed-forward sublayers] … these layers are the mechanism used to gradually learn a general, increasingly abstract understanding of the entire text being processed.
in my opinion, this is the part that people who hates LLMs (large language models) chooses to ignore.
- Comment on EU's Top Court Just Made It Impossible to Run a User-Generated Platform Legally 2 months ago:
The Court said the host has to :
1- pre-check posts (i.e. do general monitoring)
2- know who the posting user is (i.e. no anonymous speech)
3- try to make sure the posts don’t get copied by third parties (um, like web search engines??)
Basically, all three of those are effectively impossible.in my opinion : #3 effectively seems impossible, #2 is contrary to Lemmy’s philosophy and #1 would require a lot of community supervision … that would require a different Lemmy software.
- Comment on DeepSeek-V3.2 Release 2 months ago:
- Comment on AI music creates unease as it tops the charts 2 months ago:
i like this “piano” piece :
the #11 on this album : …bandcamp.com/…/suno-ai-ai-generated-music
titled :
Echoes of Silence - [Piano (Solo)] 01:47i found it after listening to a few artificially generated bad ones.
- Comment on China begins assembling its supercomputer in space 8 months ago:
translation from the original :
On 2025 May 14, at 12:12 Beijing time, Guoxing Aerospace successfully launched 12 satellites for the Space Computing Constellation 021 mission using the Long March 2D carrier rocket at the Jiuquan Satellite Launch Center. The satellites entered their predetermined orbit, marking the successful launch of the world’s first space computing constellation.The successful completion of the initial constellation launch mission will usher in a new era of “space computing” globally.
- Comment on OS-busting bug so bad that Microsoft blocks Windows Insider release 8 months ago:
Not for everyday m$off user
… The Canary channel is where Microsoft tries out its latest builds of Windows, and there’s no guarantee that anything in the Windows Insider program will ever see the light of day. …
- Comment on Hybrid AI model crafts smooth, high-quality videos in seconds 9 months ago:
Arguably about 100 fold improvement over "diffusion models” like OpenAl’s SORA and Google’s VEO 2.
Such major improvements are to be expected in my opinion in the future of artificial intelligence. - Comment on What Will Remain for People to Do? The future of labor in a world with increasingly productive AI. 10 months ago:
TLDR : The author concludes he doesn’t know after a long detour arguing with himself that we are irreplaceable.
(…) labor can be immiserated and wages are driven to zero (…)
- Comment on The Llama 4 herd: The beginning of a new era of natively multimodal AI innovation. 10 months ago:
When students progress in school, they are no longer interested in simple problems and need harder problems as challenges to stay motivated. They got something similar in the work :
quote :
To address this, we removed more than 50% of our data tagged as easy by using Llama models as a judge and did lightweight SFT on the remaining harder set. In the subsequent multimodal online RL stage, by carefully selecting harder prompts, we were able to achieve a step change in performance.
Eventually, we will ask a question one of those systems and they will consider us with disdain.
Also, for what i read in their work, elaborating such systems becomes more and more convoluted. Eventually, people working on these things, will lose the trail of what they did before. So, they will forget why they came to do something one specific way.
- Comment on OpenAI to build open AI model amid competition from Meta and DeepSeek 10 months ago:
OpenAI, which until now has been a fierce defender of closed, proprietary models (…)
it’s like when Google said : “don’t be evil” but what they meant was : “don’t you be fucking evil, leave that to us”
(…) competition in the open-source space from Chinese rival DeepSeek and
Meta**Alibaba.
- Comment on How Software Engineers Actually Use AI 10 months ago:
Conclusion : under supervision AI (LLMs) is useful to most programmers (including me !)
- Comment on Researchers unveil Aardvark, an AI weather prediction system that they say uses thousands of times less computing power and is much faster than current methods. 10 months ago:
i agree there are many works in AI which are far different from large language models but I saw some bad comments negating any potential for artificial intelligence so I’m happy to see we agree on this.
As for “causation”, maybe we can agree there are many levels of causation … like what’s the cause of the cause … i will not be surprised when we see artificial systems good at this game. i will not be surprised either when robots, driven by artificial intelligence, begin to read the world by themselves, then, built new models of advance robots and evolve by themselves.
- Comment on Researchers unveil Aardvark, an AI weather prediction system that they say uses thousands of times less computing power and is much faster than current methods. 10 months ago:
2025 Mar 20
(…) Aardvark has been developed by researchers from the University of Cambridge (UK) supported by the Alan Turing Institute (UK), Microsoft Research (USA ?) and the European Centre for Medium Range Weather Forecasting (EU) (…)
So it is safe from USA’s collapse in science.
(…) But with Aardvark, researchers have replaced the entire weather prediction pipeline with a single, simple machine learning model. The new model takes in observations from satellites, weather stations and other sensors and outputs both global and local forecasts. This fully AI driven approach means that predictions are now achievable in minutes on a desktop computer. (…)
a.i. is so far from simple “word completion”.
- Comment on China’s In-House EUV Machines Reportedly Entering Trial Production In Q3 2025, Utilizing An Approach That Offers A Simpler, Efficient Design; SMIC & Huawei To Benefit Greatly 11 months ago:
!technology@lemmy.world
First posted there ~4 hours before
lemmy.world/post/26668055 - Comment on Eerily realistic AI voice demo sparks amazement and discomfort online 11 months ago:
Children growing up using such systems will develop dependencies and abilities, further and beyond whatever “addictions” people now have for their devices and games.
- Comment on Traumatic stuff gives ChatGPT ‘anxiety,’ but therapy helps. 11 months ago:
Does ChatGPT Dream of Electric Sheeps ? Or is it ChadGPT ?
- Comment on In China, AI-driven robots are ‘evolving at an incredibly fast pace’ 11 months ago:
Nothing, no specifications, no price … in this article unfortunately … except maybe for that :
full image
.
- Comment on Chinese scientists claim neural network tech unlocks 10,000X speedup in optical fiber bandwidth 11 months ago:
this guy from Tom’s hardware doesn’t have a clue what he’s talking about, so, go to the original :
interestingengineering.com/…/optical-fiber-for-br…
Single-mode fibers, used for long-distance communication, send only one light signal at a time, usually from a laser. Multimode fibers have a wider core and allow multiple light signals from LEDs to travel through them. However, these signals bounce off the fiber’s edges, which can scramble the data.Scientists have ways to fix this problem, like using artificial neural networks or spatial light modulators, but these methods take time and use a lot of energy. They also require changing the light signals into electrical signals before processing them, which slows things down.
To solve this issue, researchers added tiny diffractive neural networks, about the size of a grain of salt, to the ends of ultra-thin multimode fibers. These networks can read and process light signals in real-time without needing complex computing power.
- Comment on WD's new HDMR tech to enable record-breaking 100TB+ drives 11 months ago:
Conventional HDD are less expensive because they do not require lithography on the platters. But with this new technology that would change. So there wouldn’t be a price advantage, this is to say 100 terabytes would cost as much as if it was SSDs, i.m.o.
- Comment on WD's new HDMR tech to enable record-breaking 100TB+ drives 11 months ago:
This technology is projected to be quite expensive as bit-patterned disks must be physically patterned using lithography or etching equipment in cleanrooms.
… at this point, just go for SSD
- Comment on Cuneiform-like digital storage tech quadruples data storage 1 year ago:
There are still many challenges that face this new data storage technique, like how dust would affect them and how long they would last in storage.
… and also : how impractical and extremely expensive it all is today … it’s only at the research stage.
