lagrange77

> It's been a long time since "quality" and "success" have been decoupled.

I think so, too.

I guess quality was a property of interest in the old days, because the path e.g. for commercial music was: Maximize profit -> Maximize sales -> Maximize what the target audience likes -> Maximize quality.

For TikTok etc. they bypass the market sales stuff and replace it by an 'algorithm', that optimizes for retention, which is tightly coupled to ad revenue. I imagine the algorithm as a function of many arguments.

Just relying on quality is an inefficient approximation in contrast to that.

> In what concrete problems do you (or Veritasium) think analog computing could beat a GPU?

Solving (systems of) ODEs without the issues that can arise from numerical solving like numerical instability. Oh and it does that instantaneously.

Thank you very much for the hints! This rabbit hole gets deeper every day. :D

Hey, i'm brainstorming for a 3d vector renderer in WebGPU on JS/TS and stumbled on your project [0] yesterday.

(Thick) line drawing is especially interesting to me, since it's hard [1].

I also stumbled upon this [2] recently and then wondered if i could use that technique for every shape, by converting it to quadratic bezier curve segments.

Do you think that's a path to follow?

[0] https://github.com/Lichtso/contrast_renderer

[1] https://mattdesl.svbtle.com/drawing-lines-is-hard

[2] https://scribe.rip/@evanwallace/easy-scalable-text-rendering...

That's a really interesting application of LLMs and manim.

My problem with it, as with LLMs in general, is that i can't fully trust it. This isn't a problem with topics you know well. But those 3b1b style tutorials are meant to build an understanding or mental model of a concept for people who don't know it well yet. I'm afraid that, if the output is (partially) wrong, watching such a video could be even harmful in understanding the concept, because your 'greenfield' brain may pick up pieces of it that made sense to it.

I've ordered one last holidays and haven't had the time to use it yet. Unfortunately it doesn't fit in the famous dev board drawer.

Cool, thanks for the reference!

Maybe WebGPU would be a good porting target.

Really cool app btw!

I have once seen a video of Kip Thorne, explaining that the black hole visual effects of Interstellar were an actual physical simulation. I wouldn't have thought, that it was feasible to run on an iPhone.

Yes, but that's essentially my point. Where to draw the system boundary? The brain is also composed of multiple components and does IO with external components, that are definitely not considered part of it.

Good point!

I'm wondering wether it would count, if one would extend it with an external program, that gives it feedback during inference (by another prompt) about the correctness of it's output.

I guess it wouldn't, because these RAG tools kind of do that and i heard no one calling those self aware.

Cool, thanks and all the best with the project.

Very cool!

Does it support ArrayBuffers?

Exactly. ML models such as autoencoders can also be used for reduced order modeling / dimensionality reduction e.g. for MPC of fluid systems.

There is a big overlap between Optimal Control and Reinforcement Learning, in case you didn't know.

Also Steve Brunton does a lot on the interface between control theory and ML on his channel: https://www.youtube.com/channel/UCm5mt-A4w61lknZ9lCsZtBw/pla...

Not a command, but a little known feature of the Terminal app:

(shift+command+K) or Menu 'Shell' -> 'New Remote Connection...'

opens a SSH, S(FTP), TELNET connection manager window!

16 points and it returns a 429 (Too many requests). I wonder what the avg. factor between HN points and site visits is approximately.

"Oh, thank you for pointing that out, you're totally right! My previous answer would have killed humankind as a side effect. Here's an updated version of the solution: ..."

Interesting point of view. I have to think about it.

I'm not sure i understand your answer, to be honest, especially the cat fable.

> as needed by the economy

Even if that is the purpose of education, a mathematician, engineer or physicist who has a deep understanding is worth more to the economy, than one who can remember and perform algorithms and computations, isn't he?

So you propose to teach them different methods of solving.

I think, while your mapping from solution methods to professions makes sense, the act of solving should be taught as an afterthought.

The goal should be to build a robust mental model. Personally i find playing around with computer visualisations helps a lot with that.

The same is even more important for differential equations, imo. I wish i would have spent less time studying algorithmic DE solving techniques in my school days. The beauty and power of expression of these has nothing to do with the way you solve them, if you ask me.

The jailbreak i used is working up to FW v5.16.2, which was obsolete at the time i bought the device (approx. 2 months ago). However i was lucky, since my device came with that very FW version.

I have recently bought a Scribe 1st Gen, when it was on sale.

Rooting it was a hassle, but it was worth it.

The excellent hardware (display, enclosure) combined with Koreader makes it really joyful to read technical books and documents, especially when combined with Calibre.

As one of those books was about Lisp, i also installed Kterm, tmux, the joe editor and tinysheme, which served as a decent Scheme Playground/IDE while i was on vacation!

Yes, he doesn't exactly come across as an Airpods-wearing Wall Street yuppie.

> Even if you were, is that really a bad thing?

I noticed, that it sounded like that after i submitted the comment. Was not intended.

> Are there really bad ideas in socialism ? yes of course, there are bad ideas in late stage capitalism or in any free market system or any system.

I totally agree.

Am i a communist when i prefer a government to operate and provide basic services like the public transport system, the postal system, the telecommunication/internet service or prisons either paid by taxes or by fees, that don't generate profit beyond paying the costs, instead of using taxes to rescue bankrupt banks and companies?

EDIT: Oh and hospitals!

> WASM could succeed as well.

I would guess WASM is a big building block of the future of apps you imagine. Figma is a good example.

I'm annoyed that he was awarded just now, obviously as a reaction to ChatGPT and the breakthrough of LLMs. If his work is worthy, it has been worthy many years ago.

This reinforces the reduction of ML to LLMs, just like the use of the term AI.

2024? I haven't even finished the printed concorde manual from FS2000.

Thank you!

That's awesome. Looking forward to the code!