NegatioN

I feel like they did a somewhat recent update to the downloader which fixed things. I had issues before as well but not anymore.

The streaming issue is another matter though :/

I also really treasure that quote. Your visualization really made it hit home again though.

It does make me reflect on this piece I wrote 9(!!) years ago though, which hasn't completely materialized. I think I'm due for a re-alignment of priorities.

https://www.jrishaug.com/Who-do-you-want-to-be/

When buying a new washing machine and dryer, I actually spent hours extra to find models /without/ app requirements last summer. There were so few of them that did what I wanted, and also didn't require internet access that I'm worried the next time around there will be no more options where I can elect to keep them off the net. :/

I guess I'll take the bait. If it's "pretty clear" that all of these are wins to you, then I think you might be part of the other extreme pole to what you're arguing happens here. Leaving all nuance behind doesn't really help the discussion.

For example: I'm excited to see how trade wars with the world turns out for the US, but I'm not sure it's a guaranteed win.

Although that in itself is a true statement, there is sooo much performance left on the table everywhere.

And Lynch's Dune looks like it does, probably in no small part due to Jodorowsky, Giger and Mobius.

I recommend everyone interested in Dune to take a look at Jodorowsky's Dune [0], the documentary which explains how they did the preproduction of Dune, and pitched it to the studios. All the deals eventually fell through, but the pitch decks circulated all the studios for years afterwards, and inspired countless other movie makers.

The best things about the new Dune installments imo is the art, but all, or most of it feels like copy pasted ideas from what came before them. CGI is a huge asset to actually get it working though.

0: https://m.imdb.com/title/tt1935156

I wonder if that's at all related to different sugars configuration at different temperatures as described by AppliedScience [0].

Maybe there are other substitute sugars out there which work well at low temperatures?

0: https://youtu.be/kiuXasyB3L0?si=b_-6iNw1TD35alG4

You get to launch Windows games through Steam Proton. It's a no-hassle way to get up and running with gaming on Linux, although many are probably using Wine or Lutrix.

Not a perfect solution, but you could just use Steam to load games from GOG on Linux though. Thereby getting "the best of both worlds". I have yet to stumble upon any major issue doing this.

I think books are the best medium for learning some things, and probably in some aspects for writing.

However I'm worried some countries seem to be throwing the baby out with the bathwater.

There are many things that are easier to learn with computers/screens than without as well, they just need to fit the medium. [0]

Intended as a reply, but the comment got deleted, so I might as well include it here:

The article [0] is focused on homeschooling, so the exact points listed there doesn't necessarily have a leg up on traditional media (implying you're in the right environment to facilitate learning these skills well without computers, which I don't think most kids are).

One off-hand example [where screens can be better than a book], would probably be using simulations to assist in learning physics, instead of just solving the equation on a page. Things where interactivity sets the learning in better context than a book probably would.

I'm also very excited to try teaching our child math using apps like DragonBox, which seems to allow for much easier visualization of how to solve equations than I got at school. [1]

0: https://www.fast.ai/posts/2024-10-29-screen-time/

1: https://dragonbox.com/products/algebra-5

I haven't compared image models in a long while, so I don't know the relevant performance metrics. But even a few years ago, you would usually use a pretrained model, and then finetune on your own dataset though. So those models would also have "seen millions of images", and not just your 100k.

This change of not needing ML engineers is not so much about the models, as it is about easy API access for how to finetune a model, it seems to me?

Of course it's great that the models have advanced and become better, and more robust though.

I feel like the way Romero and Carmack wanted to make the player "connect" with the game was different than what Tom had in mind (from reading the book).

Tom wanted elaborate lore and story-telling, while the rest wanted to make the game experience what the player connected to. The instant reaction to your input, seeing your bad-ass character (and by extension yourself) inflict awesome damage on the world.

This to me is more of a conflict of _what_ the player should connect to, as opposed to not wanting the player to connect at all.

"Right now there's this thing where ethics aren't what they used to be. This idea that people are trying to replace the ideas of good and bad, with better or worse." -Dave Chappelle

What you're writing should naturally lead to the conclusion that working for Google, Meta, Verizon, AT&T etc are all in the category of companies one shouldn't strive to use their hard earned talents for. For some reason I cannot fathom, you seem to land on the idea that Palantir is okay, because all these others somehow have snuck under the radar of many people?

What gets most foreigners is usually the darkness in the winter, and not the temperature fwiw.

It's hard to describe, but many people end up quite depressed.

Just in case you're in the know:

I was really really close to trying to spin up something at work using the Max Engine. However, I see that it collects telemetry [1], and we can't really allow that. Do you know if there are plans to be able to turn this entirely off at some point?

It's nice to know if I should keep it on my radar, or if I can't consider it without changing jobs :)

[1] https://docs.modular.com/engine/faq#does-the-max-sdk-collect...

I feel like teenagers are not exactly a demographic that never sees targeted ads

That is definitely true, but my opinion would be that they need "less ads", and not "more". That's a phase of life when most people are very suseptible to societal messaging, and form their self-image (partly?) based on it.

Not all ads are created 100% equal of course, so one can only hope these ads are nice to the users.

There seems to be a focus on something about the brain "calcifying" or people getting stuck in their own pattern over time here.

That may be a component, but I think anothet thing that correlates heavily with the graphs presented in the article is simply: When you have enough spare time to prioritize music. (although you could explicitly prioritize it like some commenters mention)

I still find lots of new music, but it always comes in periods of my life when I have some leeway, and those are fewer and farther between now.

I didn't know it runs an LLM when you append a "?", but for any Kagi-users out there, you can use the bang: !fgpt $QUERY if you automatically want to jump to an LLM.

The !fgpt-bang seems to be the model: "Claude 3 Haiku" going by the developer notes. Which often outperforms at least ChatGPT 3.5, easily recouping some of the money I put into Kagi every month.

This is the first AI song I've really wanted to keep with me on my phone. If someone knows someone, we need a human performance of this!

Encore, encore!

I'm not sure if these books really are what you're looking for, because each of them is a mix of engineering history, and the description of how a group of people end up doing amazing things. In each of the books, there were nuggets of wisdom that I've tried to bring along with me in my job (as best I can). Like:

- Doing things as simply as possible to start off

- Keeping iteration time to a minimum, for maximum exploration of ideas

- Being willing to think outside the box

My takeaways above, hardly do these books justice, but they are as follows:

- Skunk Works: A Personal Memoir of My Years at Lockheed [0]

- Hackers: Heroes of the Computer Revolution. [1]

- The Dream Machine: J.C.R. Licklider and the Revolution That Made Computing Personal [2]

[0] https://www.goodreads.com/book/show/101438.Skunk_Works

[1] https://www.goodreads.com/book/show/56829.Hackers

[2] https://www.goodreads.com/book/show/722412.The_Dream_Machine

In a similar vein to the "War Stories"-clips, (even though I never played Ultima,) their Ultima post-mortem from GDC is a great watch too. https://youtu.be/lnnsDi7Sxq0

I might be atypical, for sure. I've moved through Java, Go, Python, c++ and recently to Scala now, with no contacts at any of the companies I've applied to, but hopefully good references in each case. I've only been a dev for 7-ish years and had 3 separate jobs in that time. Youth might have given me some leniency in the first moves, but maybe less so in the latest one.

Im from a small country, so it might just be that the developers have a stronger hand at the negotiation table than elsewhere, since there're fewer competent devs as a result.

In any case, as a programmer I feel like I would always hire someone that demonstrates good programming skills in the "wrong language" over someone who's mediocre in the "right language", so it's sad to hear that's not the case many places. And that might be a lesson I should keep in the back of my mind as well.

I don't understand why so many people seem to disagree with this point.

If you're a specialist that's one thing, but a normal software stack should be generalizable?

I got hired at all of my jobs so far without having major prior experience with the languages they use, and definitely not the frameworks.

It's just a database of submitted works I think. You can try scrolling down on the opening page to see random prompts and outputs.

The previous models were either 1. Limited in their capacity to create something that looked very cool, or 2. Gigantic models that needed clusters of GPUs and lots of infrastructure to generate a single image.

One major thing that happened recently (2ish weeks ago) was the release of an algorithm (with weights) called stable diffusion, which runs on consumer grade hardware and requires about 8GB of GPU RAM to generate something that looks cool. This has opened up usage of these models for a lot of people.

example outputs with prompts for the curious: https://lexica.art/

I work in the area of recommendations, and this is not a solved problem at all. You can only recommend what has been shown (without doing coldstart). One major issue is that other forms of content than 30sec clips can't easily utilize TikToks way of bootstrapping engagement when the item is fresh. Not everyone will understand or appreciate a "new Shakespeare" and it may fall by the wayside.

I too hope it gets better, but it's hard to replace a panel of experts that have sifted through their subject when it comes to quality recommendations in some fields.

The quality of the upper echelon of art may be raised, but there's still a discovery problem there. Having to sift through stuff to find the gems is already an issue imo. The OP makes a decent (pessimistic) point

For the people who answer this, I'd like to know if they booted from an sd-card or ssd-device as well. I noticed a huge difference in the responsiveness when using my Pi 4 (only via ssh and with a terminal) for some machine learning jobs after I switched to boot from an external SSD. For Pi4s made in the last couple of years, booting from ssd seems to be a built-in feature, without any need for flashing anything on the device first, which held me back from doing it earlier.

One thing that struck me while reading this is the potential connection of doomscrolling to our evolution.

Humans are in large part "information foragers", in the sense that information has been vitally important to our survival as a species, and potentially a part of why we've developed larger brains than similar animals. [0] For example: A poisonous berry has a very different utility from an edible berry. Or: A monsoon season changes the climate enough to make an important difference for your tribe's survival whenever it happens. Querying your surroundings, or other humans for this information might have a large impact.

In that sense, it makes perfect sense that we "can't stop seeking novel and potentially interesting information" on these sites. Of course the way some of them are designed to be addicting doesn't help. But it illustrates why it's hard (or impossible) to quit doing this activity in it's entirety.

Maybe we shouldn't strive to quit searching for information, but make sure we have a satisfactory information scavenging activity as our go-to? I don't know what that would look like in practice, but the first thing that pops up in my mind is something like having a list of topics that seem interesting, and that you actively seek out information on, where you partially investigate some of the forks in the road.

Then there's the problem of being too exhausted to do something actively, which might need another solution entirely of course.

[0] https://www.youtube.com/watch?v=F3n5qtj89QE Sadly I don't remember the timestamp of Jordan Peterson's statement saying humans are information foragers, but intuitively it holds up.

Torch.jit shouldn't impact your performance positively or negatively in my experience. Although I've only used it on cpu. It's as far as I know just used for model exports.

The nice thing about it though, is that you can embed native python code (that's compiled to c++) into the model artifact. It's allowed us to write almost all of the serving logic of our models very closely to the model code itself, giving a better overview than having the server logic written in a separate repo.

The server we use on top of this can be pretty "dumb", and just funnel all inputs to the model, which the Python code determines what to do with.

As for model speedups, maybe you should look into quantization? I also find that there's usually lots of low hanging fruit if you go over code and rewrite to quicker ops which are mathematically equivalent, but allocate less memory, or do less ops.