HN user

ekropotin

362 karma
Posts3
Comments196
View on HN

I personally don’t see its being a case, based on my observation.

There is this town nearby where I live - super white, gives old money type of vibe, very expensive real estate. It’s full of free range kids running around on the streets.

It was shock to me to see it, after our diverse suburb, where kids pretty much either locked at home doing homework or at classes all the times.

So in my opinion there is definitely a cultural aspect of it.

From what I understand, macOS uses weird kernel implementation, which is almost open source, but not 100%

So we do at least agree on the fact that quality of human language <-> embeddings transition depends on how good target language is represented in the training dataset?

Even if it happens at the edge, on every conversation turn, I may assume non captured small subtleties of meaning over time can accumulate into significant error.

it’s clearly not the case that these things only make use of Spanish training data when you prompt them in Spanish.

It’s not! And I’ve never said that.

Anyways, I’m not even sure what we are arguing about, as it’s 100% fact that SOTA models perform better in English, the only interesting question here how much better, is it negligible or actually makes a difference in real world use-cases.

It’s just a subjective observation.

It just can’t be a case simply because how ML works. In short, the more diverse and high quality texts with reasoning reach examples were in the training set, the better model performs on a given language.

So unless Spanish subset had much more quality-dense examples, to make up for volume, there is no way the quality of reasoning in Spanish is on par with English.

I apologise for the rambling explanation, I sure someone with ML expertise here can it explain it better.

Fair enough! Indeed, there are ways to build performant microservices on Java, e.g. Qwarkus + GraalVM, but I've never seen anyone in enterprise world is doing it - 99% of services are still being build on Spring Boot, so in my mind Java in SB are inseparable.

I’ve Been stuck with Spring for many years too and I think it’s just awful choice for micro services that need to scale, simply due to startup time.

Not to defend Nodejs ecosystem - your points are 100% valid.

With that in mind, I think Go is much better option.

Because of my ignorance, I thought it’s the same thing.

What I actually meant to say is “not staggered”.

I use Colemak DH for many years and Dvorak before that and I am of an opinion that alternative layouts are way overrated. I even somewhat regret inventing so much time in learning them. QWERTY is just fine!

The matter I want to preach about tho are split ortolinear keyboards. I believe absolutely every typist should use them. Conventional keyboards are just bad from ergonomics perspective and eventually it’d have a toll on your wrists health. And many of these keyboards stores key mappings directly on the chip so no need to mess with weird mapping software.

Having said that, my split keyboard is one of the best investments I did in my life.

I didn’t really realize the degree of their slowness, until I migrated one of the projects on a self-hosted gitea and runners. This setup is just breezing! It’s an order of magnitude faster we’re talking about.

Granted, self-hosting git is not feasible for everyone, but GitHub + self hosted runners seems like a very good option.