lostmsu

29 ms·

721.
by lostmsu 9mo ago
You can make it as slow as you want. At half TDP it is silent.
722.
by lostmsu 9mo ago
It works in the sense that there are lots of professional (as in they earn money from software engineering) developers out there who do the work of exactly same quality. I would even bet they are the majority (or at least were prior to late
723.
by lostmsu 9mo ago
ROG Flow Z13 with maxxed out RAM.
724.
by lostmsu 9mo ago
> Journalistic accounts describe individuals who have developed strong beliefs that chatbots are sentient But chatbots are sentient within a single context session!
725.
by lostmsu 9mo ago
My laptop runs gpt-oss 120B with none of that. Don't know how long though. I suspect a couple of hours continuous.
726.
by lostmsu 9mo ago
It actually seems worse. gpt-20b is only 11 GB because it is prequantized in mxfp4. GLM-4.7-Flash is 62 GB. In that sense GLM is closer to and actually is slightly larger than gpt-120b which is 59 GB. Also, according to the gpt-oss model ca
727.
by lostmsu 9mo ago
No, that was the claim made. They justified it with the paper that states what you say, but that's exactly the problem. The statement of paper is significantly weaker than the claim that there's no progress without exponential inc
728.
by lostmsu 9mo ago
The point is, that contradicts the claim that lately the progress is only made by throwing more compute.
729.
by lostmsu 9mo ago
It is until it IPOs or is profitable.
730.
by lostmsu 9mo ago
No, models significantly improved at the same cost. Last year's Claude 3.7 has since been beaten by GPT-OSS 120B that you can run locally and is much cheaper to train.
731.
by lostmsu 9mo ago
Move to a swing state and vote.
732.
by lostmsu 9mo ago
No, the comment is about "will", not "is". Of course there's no definitive proof of what will happen. But the writing is on the wall and the letters are so large now, that denying AI would take over coding if not al
733.
by lostmsu 9mo ago
Yes, it does sound like an AGPL (hopefully) version of Stripe + maybe the software stores like F-Droid. At least that's what I'd want this to be.
734.
by lostmsu 9mo ago
So no comparison?
735.
by lostmsu 9mo ago
Comparison with vanilla of the same size/flops budget?
736.
by lostmsu 9mo ago
Bro is clueless. The opposite of what he believes is true: process barely changes outcomes comparatively to the models. No matter your process, you can't make Claude 3.5 do what 4.5 is capable of. There are tasks that you'd have t
737.
by lostmsu 9mo ago
Small Bluetooth receiver for an older car.
738.
by lostmsu 10mo ago
> perfectly willing dreaming
739.
by lostmsu 10mo ago
Any links?
740.
by lostmsu 10mo ago
"I used AI to write a GPU-only MoE forward and backward pass to supplement the manual implementation in PyTorch that only supported a few specific GPUs" -> https://github.com/lostmsu/grouped_mm_bf16 100% v
741.
by lostmsu 10mo ago
Surprised to see Georgia in the list
742.
by lostmsu 10mo ago
I imagine Claude would be able to at least fix this one.
743.
by lostmsu 10mo ago
No, I doesn't beat Gemini in any benchmarks. It beats Gemma, which isn't a SoTA even among open models of that size. That would be Nemotron 3 or GPT-OSS 20B.
744.
by lostmsu 10mo ago
Whisper.cpp and llama.cpp also work with Vulkan.
745.
by lostmsu 10mo ago
What else has significant impact on ability to study? General downsides are irrelevant, so I don't see a point in your comment.
746.
by lostmsu 10mo ago
You don't have to install the whole CUDA. They have a redistributable.
747.
by lostmsu 10mo ago
IMO they'll roll back. It was a bad decision, and a few unis already rolled back.
748.
by lostmsu 10mo ago
Ear plugs are under $5. If you could not think of that over years, you're not gonna be able to score well in SAT for different reasons.
749.
by lostmsu 10mo ago
It is hard to hurt someone when their score is a standardized test.
750.
by lostmsu 10mo ago
Maybe the laws should explicitly encode babies' freedom of religion of their parents.

More ›