tomp

I only feel sore if I haven’t been working out regularly. After a week or two, the soreness is gone, for as long as I keep regular trainings.

(I do strength training in the gym.)

I don’t get it. Sounds like all the things the state would offer anyways - education and healthcare for poor people…

But why? I mean, just put each prisoner in a separate cell, why would you need more than 1 employee per 20-50 prisoners? Ok, maybe 3, for 24 hour rotation... Make sure you never unlock more than a single cell, and keep guns, lots of guns.

> Locking 1,000 people up for a decade costs ~1 billion dollars.

This is a purely political decision, not an inherent cost of jailing.

Your number comes down to $100k per person per year. That’s just insane. Many families earn less than that (post-tax)!

And obviously jail is supposed to be cheaper than non-jail life in the first place, because you’re not paying for luxury, just food, (cheap) rent and security.

> Dismantling OpenAI is actually a business strategy for many other players now.

Which ones exactly?

NVIDIA is drinking sweet money from OpenAI.

Microsoft & Apple are in cahoots with it.

Meta/Facebook seems happy to compete with OpenAI on a fair playing field.

Anthropic lacks the resources.

Amazon doesn't seem to care.

Google is asleep.

exactly.

If that's how it works, why wouldn't you start every startup as a non-profit?

Investment is tax deductible, no tax on profits...

Then turn it into a for-profit if/when it becomes successful!

For me as well… with constant human supervision. But if you try to build a business service, you need autonomy and exact rule following. We’re not there yet.

The problem with all these speech-to-speech multi-modal models is that, if you wanna do anything other than just talk, you need transcription.

So you're back at square one.

Current AI (even GPT-4o) simply isn't capable enough to do useful stuff. You need to augment it somehow - either modularize it, or add RAG, or similar - and for all of those, you need the transcript.

What are you gonna buy? Chinese tech? Iranian tech? Russian tech?

Who do you want to be able to spy on you and compromise your hardware?

Unless you can spin up your own fab (hint: you can't) you're dependent on a hegemon. US/EU/Israel isn't perfect, but pretty much as good as it gets.

Do you think Hezbollah was buying stuff from Israel, or otherwise using Israeli supply chains?

I think it's far more likely that Mossad has infiltrated whatever foreign (non-Israeli) supply chain they were using.

So this can happen regardless of whether you're using Israeli supply chains or not.

How frequent was that?

Read it again.

> behaviors perpetuated by men – even those men

Blaming all men, twice!

If you’re still confused, compare:

> violence perpetuated by blacks - even those blacks

> terrorism perpetuated by Muslims - even those Muslims

> world controlled by Jews - even those Jews

Society is blind to sexism against men, but quite sensitive to bigotry against blacks and Jews, so it’s easier to see if you replace a few words…

If you're picking communication protocols for your API endpoints, you're doing it wrong.

The ideal API supports `/api/call.json` and `/api/call.csv` and `/api/call.arrow` as well as `/api/call.grpc`, and the only thing that differs between these is the serializer (which is standardized and very well tested).

If the App is buggy, I want to (1) check what API calls it's making (2) be able to run those API calls myself, which is why I (ideally) need a text-based (human-readable) format.

It’s not a copyright violation.

Maybe not myself, but many averagely-talented artists can draw Mickey Mouse.

They might even draw one for me if I ask! Or I can just find it on Google… (technically my computer is producing it on the screen…)

That in itself is not a copyright violation. But if I use it, in a commercial manner, then it becomes a copyright violation.

Producing copyrighted things isn’t illegal. It’s on the user to not use copyrighted things, in a way that’s illegal (not fair use or licenced).

Tokens are mapped to keys, values and queries.

Keys and values for past tokens are cached in modern systems, but the essence of the Transformer architecture is that each token can attend to every past token, so more tokens in a system prompt still consumes resources.

They titled the article: The Elon Musk / Donald Trump interview on X started with an immediate tech disaster

If they were actually neutral, they'd phrase it more like: with technical difficulties.

No, what you're describing is NeRF, the predecessor technology.

The output of Gaussian Splat "training" is a set of 3d gaussians, which can be rendered very quickly. No ML involved at all (only optimisation)!

They usually require running COLMAP first (to get the relative location of camera between different images), but NVIDIA's InstantSplat doesn't (it however does use a ML model instead!)

Not really.

First if all, most GS take posed images as input, so you need to run a traditional photogrammetry pipeline (COLMAP) anyways.

The purpose of GS is that the result is far beyond anything that traditional photogrammetry (dense mesh reconstruction) can manage, especially when it comes to “weird” stuff (semi-transparent objects).

Once the Gaussian Splats are computed (whether via ML or classical optimisation), they’re very efficient to render (similar to 3D meshes used in games). High fps isn’t incredible.

Having said that (I have yet to read the paper), "efficiency" probably refers to the first part (calculating the gaussians in the first place) not rendering.

Indeed, I'm certain they will do well regradless (even in public schools!) both because I am curious and hence knowledgeable, and because they will likely be curious as well (my genetics!).

But there's one thing that curiosity doesn't help with (in fact, it's probably detrimental) - it can leave gaping holes in knowledge (you don't know what you don't know). Hence why I'm also interested in at least somewhat (or more likely) complete treatises from experts in their respective fields...

Because they don’t distract.

Nothing I hate more than my Audi loudly beeping for non-critical infos/warnings. The only appropriate time for a car to loudly beep unprompted is if I need to stop immediately.

In many jurisdictions, promoted posts and ads must be clearly marked.

There's a difference between speed & latency when it comes to performance.

Faster-whisper can process audio faster than real-time, but AFAIK vanilla Whisper needs a few seconds long audio "frame" to do inference/STT.

Whisper Live fixes that, and reduces latency to a few tens/hundreds of ms.

Depends what you mean by “fast”.

I’ve tested WhisperLive, it’s basically real-time (i.e. low latency).

My favourite part of my math education, solutions were always nice.

Compute the eigenvalues of a random-looking (but still integers) 4x4 matrix? Oh, it's sqrt(2), I probably didn't make an error in the calculation.

Then came the advanced physics / mechanics exam. It threw a wrench into our beautiful system. The results were just about anything, incredibly ugly, like the real world :yuck: :vomit:

I think this is just your anti-capitalist bias showing. “Market” is meant in purely game-theoretic terms.

> The vast majority of people in companies are workers. Let's stop optimizing for owner wealth and start optimizing for worker happiness instead.

All these workers could set up cooperatives and work for / own them instead. The fact that this hasn’t happened suggests that in reality, the for-profit companies are better at optimising workers happiness than cooperatives (maybe they pay better? Maybe they enable more job hopping?)

you're about right.

https://smicro.eu/nvidia-hgx-h100-640gb-935-24287-0001-000-1

8x H100 HGX cluster for €250k + VAT

It’s simply unrealistic to move people.

A local good developer would cost ~200k in SF, but to move me and my family would take at least 2x that (to compensate for my partner’s lost income), possibly more (so that I can catch up to peers who’ve been saving US salaries for 10+ years, so that I can buy a house in a reasonable time).

Yes, but the problem is that the spec is such that preemptive blocking doesn't need to be implemented.

That means that Java programmers have to be very careful when writing code, lest they block the entire underlying (OS) thread!

Again, Go already went through that experience. It was painful. Java should have learned and implemented it from the start

They're not equivalent to Go's goroutines.

Go's goroutines are preemptive (and Go's development team went through a lot of pain to make them such).

Java's lightweight threads aren't.

Java's repeating the same mistakes that Go made (and learned from) 10 years ago.