llm_nerd

26 ms·

511.
by llm_nerd 10mo ago
Do not read this comment. Wait...you're still reading, defying my T&Cs! The tag to not display on external websites is up to Bluesky to enforce. I mean, you understand those Bluesky chirps or whatever are literally being served by
512.
by llm_nerd 10mo ago
"the junior engineer, empowered by some class of LLM tool, who deposits giant, untested PRs on their coworkers—or open source maintainers—and expects the “code review” process to handle the rest." Kind of depressing how it has bec
513.
by llm_nerd 10mo ago
If such a trivial test really did[1] filter out many candidates (beyond the technical limitation of the test that some client devices will render the =, as mentioned by users leveraging reader mode), I wonder if there is a greater observati
514.
by llm_nerd 10mo ago
Which is literally the point of the post. They have the = in >= at an opacity of 0 and a font-size of 1, which means it doesn't appear if styles are used properly. And their point is that candidates that copy/paste such trivial
515.
by llm_nerd 10mo ago
"I don't know why you continue to be a fucking asshole" Your very first comment was an obnoxious "Not true at all" to the absolutely, incontestably true statement that Gravitons offer better $/perf. So maybe yo
516.
by llm_nerd 10mo ago
So confident. And exactly the whack-a-mole nonsense I predicted. See the comment by electroly. They actually know what they're talking about. See, the FPS score is for the whole machine. The c8g gives you 8 real cores. The C8i gives yo
517.
by llm_nerd 10mo ago
There is a sudden groundswell of reports about China using nvidia chips, always by unnamed sources, and I suspect if you could trace it back you'll find nvidia pulling the levers. nvidia is facing a lot of competitive threats and their
518.
by llm_nerd 10mo ago
Groan. Yes, absolutely true. While I know this thread will turn into some noisy whack-a-mole bit of nonsense, an easy comparison is the c8g.2xlarge vs the c8i.2xlarge. The former is Graviton 4 vs Granite Rapids in the latter. Otherwise both
519.
by llm_nerd 10mo ago
While the 5 variant isn't yet available outside of the preview, you can of course spin a 4 up and run geekbench yourself. Plenty of people have and you can find them in the GB DB. And of course most people spin up their specific worklo
520.
by llm_nerd 10mo ago
In Amazon's Graviton 5 PR they note that over half of all new compute capacity added to AWS over the past three years has been Graviton-based. That's an amazing stat. It really is incredible how ARM basically commoditized processo
521.
by llm_nerd 11mo ago
>the comments that sound "AI" generated but aren't have about as much value as the ones that really are When "sounds AI generated" is in the eye of the beholder, this is an utterly worthless differentiation. I me
522.
by llm_nerd 11mo ago
I think people are just presuming that others are regurgitating AI pablum regardless. People are seeing AI / LLMs everywhere — swinging at ghosts — and declaring that everyone are bots that are recycling LLM output. While the "thi
523.
by llm_nerd 11mo ago
What a wild series of events. It's a demonstration of someone having "no cards", as Trump likes to say. Of having delusions about the hand they hold and thinking they're the dealmaker. Earlier this year, during the begin
524.
by llm_nerd 11mo ago
Microsoft's entire business model has been tying. Countless millions are forced to use Copilot because their IT department has contracts with Microsoft, and those same contracts are why they use Office, Teams, and so on. Their develope
525.
by llm_nerd 11mo ago
>Fabricated citations are not errors. Interesting that you hallucinated the word "fabricated" here where I broadly talked about errors. Humans, right? Can't trust them. Firstly, just about every paper ever written in the h
526.
by llm_nerd 11mo ago
>It's not easy to hallucinate papers out of whole cloth, but LLMs can easily and confidently do it, quote paragraphs that don't exist, and do it tirelessly and at a pace unmatched by humans. But no one is claiming these papers
527.
by llm_nerd 11mo ago
People will commonly hold LLMs as unusable because they make mistakes. So do people. Books have errors. Papers have errors. People have flawed knowledge, often degraded through a conceptual game of telephone. Exactly as you said, do precise
528.
by llm_nerd 11mo ago
>It could be a deadly blow for Google, NVIDIA, and the rest. How would this be a deadly blow to Google? Google makes TPUs for their own services and products, avoiding paying the expensive nvidia tax. If other people make similar product
529.
by llm_nerd 11mo ago
>"most favored nation" There is a weird thing Americans often do when confronted with the incredibly high price of medicine and medical care in the US of imagining that every other country is actually responsible for this (henc
530.
by llm_nerd 11mo ago
What does your comment have to do with the submission? What a weird non-sequitur. I even went looking at the linked article to see if it somehow compares with Gemini. It doesn't, and only relates to open models. In prior posts you oddl
531.
by llm_nerd 11mo ago
Two devices that sold absolutely minuscule numbers are not remotely relevant to this discussion. For non-Apple smartphone users, the overwhelming majority were rocking the atrocious trash that is micro USB half a decade after the iPhone mov
532.
by llm_nerd 11mo ago
"Over time, it has become obvious just how many posts are being generated by an LLM. The tell is the voice. Every post sounds like it was posted by the same social media manager." I'd love to see an actual study of people who
533.
by llm_nerd 11mo ago
It's just more common as a legacy artifact from when nvidia was basically the only option available. Many shops are designing models and functions, and then training and iterating on nvidia hardware, but once you have a trained model i
534.
by llm_nerd 11mo ago
For users buying H200s for AI workloads, the "ASIC" tensor cores deliver the overwhelming bulk of performance. So they already do this, and have been since Volta in 2017. To put it into perspective, the tensor cores deliver about
535.
by llm_nerd 11mo ago
Well sure, they're iterating between models. But in many cases they're quite literally copy/pasting designs. Any imagined separation between the hardware teams is fantasy based. The comment I replied to is nonsensical. "
536.
by llm_nerd 11mo ago
>iPhone on Lightning meant cable fees and control. Strange, then, that Apple already moved the iPad Pro and iPad Air to USB-C, right? Didn't they get the memo about "cable fees and control"? It's almost like they we
537.
by llm_nerd 11mo ago
Samsung released the first USB-C Galaxy S device five years after the iPhone moved to lightning (2012 vs 2017). They had Galaxy A devices on micro USB a year later in 2018. A couple of devices like the Pixel ( 4 years after lightning - 201
538.
by llm_nerd 11mo ago
Why in the world did you post this in reply to my comment? Obviously you have some weird hangup, and the fact that I mentioned inflammation quite literally in passing triggered you. Yes, inflammation is a thing in the medical world. There a
539.
by llm_nerd 11mo ago
Apple's resistance was presumably user inertia. Users had billions of cables and accessories for lightning, and Apple saw during a prior transition that people get really pissed off about this sort of change. And let's be real abo
540.
by llm_nerd 11mo ago
"I know Apple engineers working on the phone" Groan. Come on. Cite one. A single "Apple engineer" to support this ridiculous claim of insider knowledge. What year do you think it is? You understand that the SoC and I&#x2

More ›