hendrik @ hendrik @palaver.p3x.de

A software developer and Linux nerd, living in Germany. I'm usually a chill

---

Read more

Posts

8

Comments

1,826

Joined

4 yr. ago

This is not NodeJS, so it won't work that way. Also tracks users a different way. Also this is the Fediverse community and this doesn't federate.

Yes, that's why I say it doesn't generalize. They mention this in the article. These old power bricks from the 90s with a heavy copper transformer inside waste a lot of power on standby compared to the modern switch-mode power supplies. But times have changed. On the flipside we have a lot more electronic gadgets these days and things in fact add up. So if you have modern things like 5 smart lightbulbs in the house, then a network switch, an internet router and a wifi extender, plus a few USB chargers at the bedside, the livingroom, a TV set with a PS4 and a soundbar plus subwoofer. A few LED strips in the gaming den... Then you might own a dishwasher and washing machine with wifi, the oven has a display, the microwave above yet another one, the cable TV has some booster in the basement... You're likely paying more than a few cents for that. And the things which run unattended 24/7 for decades, buried somewhere, tend to not get replaced every few years, so you might still own a power brick from the 90s. So I'd say it's worth looking into... I mean not super important, you can as well skip it and just pay the amount... But it's a thing. And I mean if you're unalike me and buy a new stereo every 10 years or so, that's also not necessarity helping the environment, and they cost money. So it's a bit complicated and a balance. At least I can somehow relate to the article, because the multi-outlet power strips behind the TV and my desk with the computer kind of look like the pictures there...

It doesn't completely generalize that way. I have an old stereo which uses like 7W on standby. And an old pair or computer speakers which don't really care if I press the button to turn them off. I mean that's not the power brick, but the device after that, so a bit out of scope for this article. But if I weren't unplugging them... 10W standby is 26€ a year and not just a few cents.

Tl;dr: Consider unplugging them, they all consume some small amount of standby power and that adds up. Also they wear out.

Though: I've never noticed any of the 24/7 devices I own wear out, I think that might be a myth?

Nice. I'll try it.

Yeah, I also think it's more an expression which somehow emerged, and it's not really a Eureka moment. They also seem to put it that way in the paper. The say it's an Aha for the scientists, and the whole reasoning process is an Aha, but they don't really write the model is having an Aha moment due to some insight it had.

I don't know if you watch Youtube videos, but Computerphile made a video about DeepSeek and an interesting video about Forbidden AI techniques a few days ago. That's also about the reasoning process and how LLMs can be lazy and take unwanted shortcuts, and hide information in the thinking step.

You might want to explain the reference in case we didn't all learn the paper by heart...

https://github.com/deepseek-ai/DeepSeek-R1/blob/main/DeepSeek_R1.pdf

Table 3 on page 9. The "Aha" paragraph starts on the page before that.

And yes, I think I tried DeepSeek and it wrote things like that, I'm not sure if "Aha", but similar things. Also "Wait..." and that something didn't look right and it had to backtrack... It generally writes a lot of weird stuff. Sometimes it even writes wrong things and then doesn't listen to itself and it silently corrects itself in the process. Other models phrase things with a different tone. At least that's what I've seen. Some also repeat a lot of stuff. I'll write what I want, and then it goes on and on "the user wants me to ..." and repeats everything once again, just not in first person, but in third person. And then after some nonsense, it tries to dissect the problem.

Of course. These all are different issues. Encrypted messaging has nothing to do with handing out my phone number to everyone.

I can't remember why I skipped SimpleX. I tried it some time ago, maybe it sucked too much battery on my old phone... Should I have another look at it? Respectively, is it any good for someone like me who already uses a Matrix messenger? I mean not theoretically, but for every-day use.

Yes, this. And with WhatsApp or an dedicated app they're either directly on your phone. Or have your (personal) phone number. Which isn't great. With eMail you can just have another spam address. And that's more complicated with phone numbers and most people don't have a second one dedicated to spam and advertisements...

I don't really know why that is. If you want to bypass the issue and just solve it, try sudo dpkg-reconfigure tzdata on the command line.

(Edit: Mind that sudo on the command line generally doesn't show any stars or such things. After being prompted for it, you type in your password blind and hit enter.)

More often than not, internet fights have little to no direct effect on the real world. We all love drama. And fighting some rando on the internet genuinely feels like you stuck it to someone. That's the thing with them. But we all need to ask ourselves if it did something for the people, or the cause we claim to support. If yes, it might be warranted, if no it was just mire selfish than helping.

I suppose it's like with the boiling frog syndrome. Temperature increases gradually. But there is no such point when we all jump out.

Maybe the internet already was tainted before generative AI, with all the tracking, selling of userdata, big soulless services, the attention economy.... Maybe that was it and we're way past it already?

Hmm, what's the big news here? Isn't that just a new example app for the mediapipe framework? I believe that was already available a year ago in May 2024 and could do LLM inference and image tasks back then?!

And in addition to that, it seems to be quite slow. It doesn't use any acceleration on my Pixel phone, just uses the CPU to output like 6-8T/s for the 3b model...

Oh well, some people in the USA have a really "interesting" relationship with their own constitution these days... I sometimes feel like explaining it to them. Or what a constitutional republic is.

Yeah, keeping things "off record" is the usual strategy to get away with whatever you wanted.

I would have declined that advertisement article if it were my journal. Already the premise with the market share is lacking a citation, I can't find the source and other statistics show wildly different numbers.

So they're open-sourcing some minor glue to connect to an API of a big proprietary online service, or what's this about?

Thanks, nice video and seems he has some numbers. Very inhuman that they figured out exact numbers how it has an allowance to take out 15/20 bystanders as well. Or an entire elementary school if it's an high ranking "target". I mean war is a bit of a different thing than policing. But a minimum 10% false positives plus collateral murder is quite a lot. And then I'm not sure if there is any substance to those numbers. I suppose they conveniently eliminate all the evidence with the same bomb that kills the people. And they don't do research, so I wonder how they even figured out a ratio.

I'm looking more for large-scale quantitative numbers. I mean one destroyed life is really bad. But they could argue they'd have saved 30 lives in turn, and then we'd need to discuss how to do the maths on that...

I'm not very surprised. I think even old-school face recognition does things like measure distance between your eyes, nose etc, and stuff like that (your skull) doesn't change a lot during 10 years of adulthood. The real danger is that they connect all of that information. And as you said, it's everywhere these days, they have a lot of sources and -of course- it's in the wrong hands. I say "of course", because I don't think there are many applications to help with anything. That technology is mainly good for oppression. And predictive policing, social scores are content for Black Mirror episodes or old sci-fi movies. Not reality.

Would be nice to get some numbers on the accuracy and performance of such a dystopian sci-fi technology (Minority Report?). There should be some, since it's already in operation for 13 years...

Or build a death star. I suppose that skips one conversion step, sucks energy from something and blasts this directly as energy in a death ray. I guess that would do for most cases a bomb works. (And it's kind of similar to the idea here, minus the black hole and spin energy.)