291–300 of 544 posts
With deluge of Chinese models popping up recently, I believe there's a few issues one needs to evaluate before deciding to use these models: - Ethics. As known, ou American frontier AI companies are incredibly ethical. And I have yet to see any interviews or blog posts by Chinese companies where they talk about how they are ethical, or at least credible HN comments about it. - Safety. Do they covertly sabotage or at…
Earlier quoted context omitted.
> This is not a local model for any reasonable definition of local That's true for now. I am hopeful that once the hardware markets have recovered from OpenAI's sabotage, we will see more hardware dedicated to local inference that can handle these big models. Also, I'm thinking about the unique MoE routing that Apple is using with their new Apple Foundation Model. The model is trained and architected so that experts…
Reading weights out of memory is the definition of a large linear read. I'm a bit mystified someone hasn't put an embarrassingly parallel flash storage controller next to some tensor processors on a PCIe card. It could have 4Tb of flash hanging off enough channels to saturate SRAM skipping DRAM entirely, and could even offload prompt processing to a GPU in the same workstation so long as it got reasonable tokens/s in…
HBF was initially announced by SanDisk, early in 2025, then early this year Hynix has announced that they have joined SanDisk in producing HBF, and that the common specification will be standardized under the Open Compute Project.
With HBF, it would be easy to make a GPU card with 4 TB of HBF, which could run the biggest existing open weights LLMs in their native unquantized form.
Earlier quoted context omitted.
I don't consider mainland aspects to be "examples of their ideological system working for the world", it works for urban areas in China and I don't really see their foreign investment to be doing that, I think it complements what the West has done and has high impact in areas that the West ignores or hasn't taken seriously for investment, only a history of pillaging and subsequently aid their ideological system - usu…
If you can't appreciate or understand what a substantial effort it was to reduce poverty in China, then you aren't a serious person worth paying attention to. It's literally the economic question of the century and something we should seriously study because we have the potential to lift the entire world out of poverty too.
We are talking about open source ai models working really well for the people of the world
Earlier quoted context omitted.
Chinese models are the closest shining example of their ideological system working for the world than anything else they've ever done From my perspective
Couldn't agree more. Maybe it's because it's a shining example of their ideological system that actually alters /my/ life, in ways that are tangible and which i can grasp, and makes it better.
Earlier quoted context omitted.
Reading weights out of memory is the definition of a large linear read. I'm a bit mystified someone hasn't put an embarrassingly parallel flash storage controller next to some tensor processors on a PCIe card. It could have 4Tb of flash hanging off enough channels to saturate SRAM skipping DRAM entirely, and could even offload prompt processing to a GPU in the same workstation so long as it got reasonable tokens/s in…
For the last year, there has been development work at several companies for products including HBF (high-bandwidth flash memory) as a supplement to HBM, in order to enable running inference for big LLMs at a reasonable cost, e.g. on one GPU-like card. HBF was initially announced by SanDisk, early in 2025, then early this year Hynix has announced that they have joined SanDisk in producing HBF, and that the common spec…
I don’t think this stands for General Linear Model.
Why would a mathematical concept have versions.
Earlier quoted context omitted.
Have any major open weight models been "open data"? Wouldn't that entail distributing vast amounts of copyrighted data?
Olmo from AllenAI has been releasing their full pipelines including data [1]. A lot of it is just repackaged and resampled dumps from copyrighted data that has long been publicly available as dumps: Common Crawl, arxiv, Wikipedia, StackExchange, reddit --- all of which are presumably copyrighted with different licenses. Go in Huggingface and you can find massive multi TB data dumps used for pre training. It is just a…
Presumably they already know. The issue is that IP law firms are tiny compared to the trillions of capital pouring into "AI". And if you believe the USA is a capitalist country where the side with deeper pockets win, you know you're not going to win against the trillionaires.
Announcement from the founder of Z.ai: “ GLM-5.2 is Fully Open, Frontier Intelligence Belongs to Everyone Today, the sudden restriction of certain frontier models is deeply regrettable. At a time when access to frontier models is abruptly cut off for non-technical reasons, we are even more convinced of one thing: science should be global. The path to AGI (Artificial General Intelligence) must never be enclosed by hig…
What is nice about GLM is that they allow other providers that I can use on OpenRouter to filter providers that are US based and with zero data retention, unlike other open-weight Chinese models like Qwen.
Seems like there's no official blog post with benchmark results yet. But I'm once again thankful for the Chinese AI labs for being open with their work and contributing it to the world under permissive licenses like this. The Fable 5 fiasco is just another reminder of how valuable these things are to have.
Based on my first impressions it's about 6 months behind the frontier labs. So very similar to Opus in January. That is, pretty damn impressive and very useable. When it comes to architecture or complex problems it does noticeable worse but I don't think anyone expected anything else. One particular interesting strong point seems to be design and user interfaces. It does seem to punch above it's weight there but that…
According to this one benchmark, I find it amusing that Qwen3.6 27B beats ALL "frontier lab" models on coding Kotlin: https://archive.vn/RYBCL / https://gertlabs.com/rankings?mode=agentic_coding&language=k...
will simon do the pelican thing for this as well