bigyabai

4 ms·

1.
by bigyabai 5d ago
Tragic: "No Way To Prevent This" Claims Only World Superpower Losing 10,000+ Personnel Per Month
2.
by bigyabai 5d ago
It's likely that a stack of Macs will draw more power for slower prefill/decode than equivalently priced Nvidia GPUs. If power efficient inference is the goal, Macs are a non-starter.
3.
by bigyabai 5d ago
> for me at least a GPU is completely useless for anything but being a token generator. No thanks to the "macos value add" that forces you to use Metal while Valve customers frolick in Protonland.
4.
by bigyabai 5d ago
Foundation and SwiftUI definitely are, which makes porting Swift applications very difficult.
5.
by bigyabai 5d ago
Mea culpa, but it's still a decent while.
6.
by bigyabai 5d ago
The evidence seems to be on his side, at least. RSI, ASI, AGI are all no-shows.
7.
by bigyabai 5d ago
That's going from 150w 720p gaming to ~15w 720p gaming in ~10 years. Let's say an inference cluster draws 1500w to deliver a small-ish 500b model at reasonable speeds/quantization. Extrapolating from your gaming example, it w
8.
by bigyabai 5d ago
> You seriously believe he's still just happily living in Russia? Today? I mean, Bashar al-Assad is. If he can escape thousands of highly-motivated revenge campaigns to see him strung up like Gadaffi, I can believe in Snowden sipp
9.
by bigyabai 5d ago
Edward Teller was known to his peers as an alarmist crank who would lie to advance his own career. Despite his notoriety, his honesty was being questioned by Los Alamos even in the 1940s. I'm not sure how Sundial and Gnomon are relevan
10.
by bigyabai 5d ago
That can happen regardless of which storefront you use to buy a free app. It's more of a developer-specific nitpick than a chronic problem with FOSS and freeware. If you're happy to sponsor the app's development, then owning
11.
by bigyabai 5d ago
Why don't you think that AI doom is marketing? Nvidia has no trouble turning profitable quarters even without crypto/AI, OpenAI and Anthropic are trapped in negative ROI. Jensen has experience in ML, avoids hysterics, and predic
12.
by bigyabai 5d ago
In the early days of emulation, there were tons of low-quality and imprecise emulator backends. If you wanted to play a dozen SNES games, you usually needed a half-dozen emulators to get them all to boot to the main menu. They were dispar
13.
by bigyabai 6d ago
Caveat emptor , the saying goes. If people don't have an appetite for privacy, don't expect them to take a principled stance based on outrageous headlines. It's not a coincidence that federally backdoored spyware like Window
14.
by bigyabai 6d ago
You know, Facebook, built by the guy who scraped college databases for personal information without the student's permission. There was never any expectation of privacy if you knew Zuck's history. Facemash almost got him expelled
15.
by bigyabai 6d ago
That's a problem with most unsolicited online advertising. Are YouTube ad-watchers any more likely to splurge on your SaaS?
16.
by bigyabai 6d ago
Hypocrisy
17.
by bigyabai 6d ago
It won't. Laya is a finetuned version of Google's BeRT model, which is almost 10 years old right now. If BeRT had any potential to disrupt the datacenter buildout, it already would have.
18.
by bigyabai 6d ago
You should try Aurora 4X , if you haven't. Not quite a full reimagining, but still oodles of fun.
19.
by bigyabai 6d ago
Here's a prospect for them; war. Do you like that? It's where they're headed, living here. Fact is, your kid could start working 40 hours a week at age 18 for the next 50 years, and still not receive a proper pension. The g
20.
by bigyabai 6d ago
Out of taxpayer's pocketbooks? The Fed isn't running a charity, there's 40 trillion dollars of debt that we all share.
21.
by bigyabai 6d ago
The entire field of mechanistic interpretability has gotten nowhere. The formalized, "causal" understanding of latent space conceptualization isn't any better than an LLM cargo cult. It's wholly possible that you could s
22.
by bigyabai 6d ago
Not so much that, but doctors, ATC, and other skilled labor the US needs and can't replace with AI. Is it better to scale back hospitals and airports instead of filling those jobs?
23.
by bigyabai 6d ago
There's no efficient way to run inference through homomorphic encryption. If the inference server is vulnerable, it seems feasible to MITM an unencrypted version.
24.
by bigyabai 7d ago
HN had a good thread discussing the original scoop: https://news.ycombinator.com/item?id=49704132
25.
by bigyabai 7d ago
I genuinely think that GNOME is more user-friendly than macOS by a wide margin. The majority of computer users today are familiar with either Windows or iOS/Android primarily. They're not coming from UNIX CDE machines or Macs on-a
26.
by bigyabai 7d ago
> When did you experience problems with Docker on Mac?? On my 16gb Macbook Pro, compiling a monorepo for my job. Between the APFS slowdowns and the virtualization overhead, it was faster to build everything in a VPS and download the arti
27.
by bigyabai 7d ago
We do understand "thinking" though. That's literally the whole point of Attention is All You Need, the attention mechanism is what separates the transformer architecture from other neural networks. It's well worth a re
28.
by bigyabai 7d ago
Again - you are right, but it still doesn't refute the grandparent's claim that today's AI is lazy, wasteful and marketing-driven. There is room to improve, and if US labs don't take the initiative then Chinese ones will
29.
by bigyabai 7d ago
I don't think either of you are wrong. The parent's assertion is that we've known this for almost a decade. BERT was highly usable for classification and sentiment analysis a whopping 9 years ago, despite being less than 0.5B
30.
by bigyabai 7d ago
> I really can’t understand what Linux gives you that you would feel a need to change your OS on Mac. - Native Docker support in-kernel - High-performance Vulkan drivers - CUDA support (macOS doesn't sign CUDA drivers for some reaso

More ›