haolez

28 ms·

451.
by haolez 4y ago
Babylon 5 was amazing. They event hinted at what could be an usage for a LLM-based AI in the far future (light spoilers ahead). There is an episode near the end of the series where we are seeing the equivalent of a TV talk show hundreds of
452.
by haolez 4y ago
A lot of these tech layoffs seem to target product people: product managers, product owners, designers and the like. Is there a trust crisis on the value generated by this kind of roles?
453.
by haolez 4y ago
My point was about a desktop machine. You could access a cluster of Linux machines with NVIDIA cards from your AMD-based desktop.
454.
by haolez 4y ago
I've been using Tumbleweed for the last few months and it's really good as a stable rolling release distribution. However, I made the mistake of buying a laptop with an NVIDIA discrete card and I'm tormented with small bugs,
455.
by haolez 4y ago
I've been using C4 as well and it's been working just fine. Sometimes I get confused about what should be a container and what should be a component (on large systems), but that's about it.
456.
by haolez 4y ago
From a VC perspective, this sounds like easy money, since it'll probably be acquired by OpenAI or Microsoft if it succeeds just a little.
457.
by haolez 4y ago
The difference is probably related to RLHF. It seems to matter a lot more than the size of the model and the hardware's capabilities. And OpenAI probably excels at that.
458.
by haolez 4y ago
What happened to NixOps? It sounded like a great idea when I first read about it.
459.
by haolez 4y ago
Just imagine if these models fall in love with languages like APL/K/J. Even with the context windows, they will be able to do tons of work.
460.
by haolez 4y ago
Fantastic article. Just a heads up for anyone reading this reply.
461.
by haolez 4y ago
Hey! What's the difference between Nitro and Wave? Wave seems like a great fit for my usage.
462.
by haolez 4y ago
Do we have some studies on the impact of extreme quantization on these models? In my understanding, the efforts to make them run in commodity hardware involves the quantization of the weights to 8 bits or even 4 bits. Is the model still rel
463.
by haolez 4y ago
I've tested Pynecone recently and it's pretty good. I'm wondering how do you intend to create a business around this library.
464.
by haolez 4y ago
Wait. Is temperature=0 "pretty deterministic", or is it deterministic?
465.
by haolez 4y ago
The problem with USENET is spam. Moderation can be a burden. I suppose Google has some systems in place to mitigate this.
466.
by haolez 4y ago
Why? Is it a limitation of the model or just something with the configuration that you couldn't figure out for this test?
467.
by haolez 4y ago
Yeah, I find this area fascinating. Like, it's very cool to run a 7B params model locally, but it must feel like a toy when compared to ChatGPT, for example. However, the 65B parameter, according to the benchmarks, is such a beast that
468.
by haolez 4y ago
Would it be possible to run the 65B one like this as well? Is the bottleneck just the RAM, or would I need an absurd number of CPUs as well? It's not that hard to create a consumer-grade desktop with 256GB in 2023.
469.
by haolez 4y ago
I've created a big system with only SQL in the past like you just said. I wouldn't do it again because of these two pain points that I had to deal with: 1. it's really hard to debug SQL queries and stored procedures (at least
470.
by haolez 4y ago
I wonder what would be the side effects of increasing the model's temperature in this kind of situation.
471.
by haolez 4y ago
I'm using it for summarization and classification of big texts and it's showing a hit-ratio bigger than human analysts. I'm satisfied with it as is :)
472.
by haolez 4y ago
Maybe they could add a new transpiled language as well, like Civet[0] or ReScript[1]. Not every project needs to be a C#/.NET clone in TypeScript :) [0] https://civet.dev/ [1] https://rescript-lang.org/
473.
by haolez 4y ago
Well, these models were trained on data produced by artists. How far can they go into the future on their own?
474.
by haolez 4y ago
Could this be used to emulate Kubernetes on a serverless environment? Pretty cool if this is possible.
475.
by haolez 4y ago
But was this move made by Sam? I understand that he became CEO after a few years in. Genuinely curious.
476.
by haolez 4y ago
The maintainer - Erik Demaine - seems very talented[0]. [0] https://github.com/edemaine
477.
by haolez 4y ago
I think it would be interesting from a C++ perspective. That was my point. Qt is awesome as is.
478.
by haolez 4y ago
Is this the Qt version that drops the custom metaprogramming stuff for QObject and friends and adopts modern C++ in its place? If so, it's a very cool development! :)
479.
by haolez 4y ago
I've tried it out of curiosity and it's extremely unstable so far. Barely runs on my Linux box and couldn't get it to run without crashing on Windows.
480.
by haolez 4y ago
I've had an unexpected epiphany when I dug deeply into Makefiles. For some reason, its spreadsheet-like logic clicked with me an influenced my programs elsewhere.

More ›