hendrik

joined 4 years ago
MODERATOR OF
[–] hendrik@palaver.p3x.de 1 points 46 minutes ago* (last edited 28 minutes ago)

Good blog post.

I couldn't think of a clever response to that. I still can't.

I think it's central to the issue they're talking about. There's demand for fast, cheap stuff. There's also demand for quality stuff. But they're not the same.

I mean, I'm sometimes sad nothing lasts anymore. Or means anything. We buy clothes, appliances, software, phones... just to throw it out a year later. Same with AI. We could do intricate art. Commission someone to draw our company logo or come up with a good advertisement video. But why? Everyone has a attention span of 30s these days and pretty much anything will do for Instagram. So rubbish it is. And we're done in 5 minutes.

Country music can also be done with AI. And it's as good as the old stuff. I think it's more that society doesn't value quality and sophisticated things any more. We rather have plenty cheap and superficial things. And for a lot of applications, it'll do. Same with some software and webdesign.

And it's the right course of action. I mean why pay someone $1,000 to do the color grading of a video clip, or use real music when nobody will pay attention? They're doomscrolling anyway and have the TV computer and phone on all at the same time. Why waste resources on lasting software solutions unless that's needed?

[–] hendrik@palaver.p3x.de 1 points 2 hours ago* (last edited 2 hours ago)

There used to be tools like Nightshade. But they're all proven to be ineffective. So it's more for fun. In reality, it doesn't change anything.

Targeted attacks should be possible. So if you know some help bot or FAQ system and how it works, you could try to get wrong answers into their databases. But I don't think there's any generic poisoning method which actually works.

[–] hendrik@palaver.p3x.de 4 points 5 hours ago* (last edited 5 hours ago)

Install OBS studio and then just start. Youtube, Twitch, Peertube, Owncast... these are all valid choices for letsplay content. And you have to start first. See what you like, what people like. Experiment a bit. I think you'll mainly need to figure it out as you go. And also build an audience as you go. Maybe start with something easy, instead of some super complicated and elaborate setup.

And drop a link once you have something to show off.

[–] hendrik@palaver.p3x.de 2 points 7 hours ago* (last edited 6 hours ago)

Soweit ich weiß, hatte er darauf auch eine Antwort. Und zwar die Oberfläche soll nicht in Richtung Internet exponiert sein. Damit ist das dann per Definition keine Sicherheitslücke mehr, sondern ein fehlerhaftes Set-Up, was der Anwender verschuldet hat.

Ich möchte ihn aber auch nicht in Schutz nehmen, oder in die Richtung diskutieren... Ich meine für den Anwender, oder die Leute die dann darunter leiden ist es egal wie es zustande kam, oder wie die technische Definition lautet. Der Schaden ist dann im Zweifel so und so da...

Letztlich wird es halt schwierig. Er verwendet seine Ressourcen lieber dafür neue Features einzubauen, herumzuspielen... Und liebt das Chaos(?) Die Zeit ist dann logischerweise nicht da um die Weboberfläche abzusichern oder solche Dinge.

Ich find's letztendlich legitim solche Entscheidungen mit seinen Privat-Aktivitäten zu treffen. Es ist ja sein Ding. Und für mich sieht es eher wie ein Kunst-Projekt aus. Es wird ja niemand gezwungen das zu Nutzen. Man kann sich ja auch Claude Code installieren, oder eine der anderen "professionellen" Agenten Plattformen ...wenn man auf sowas steht. Das hätte dann wenigstens Sicherheit irgendwie in der Projektbeschreibung.

Aber letztlich hast du sicherlich recht. Sobald Menschen zu Schaden kommen, hört irgendwie der Spaß auf. Und es ist auch vollkommen richtig die Dinge beim Namen zu nennen. Und den Leuten beizubringen mit was sie es hier tatsächlich zu tun haben. Ich denke das ist für viele Menschen nicht wirklich klar erkennbar.

Und ich lese deinen Artikel auch gerne. Danke für die Auflistung, ich lerne dabei auch noch einiges dazu. Es ist ja viel zu viel um da selbst informiert zu bleiben.

Und dann bei OpenAI anzufangen oder OpenClaw von einem der Großkonzerne übernehmen zu lassen ist auch wirklich Banane. Mir fehlen etwas die Worte. Die sind halt wirklich alle bescheuert. Und der Hype-Train ist echt auf Volldampf unterwegs.

[–] hendrik@palaver.p3x.de 1 points 7 hours ago* (last edited 7 hours ago) (2 children)

Also ich hab den c't 3003 Beitrag über OpenClaw geschaut und ein paar andere Interviews mit ihm. Ich würde sagen er ist ein zertifiziert Bekloppter. Nicht unbeding in negativem Sinne... Aber seine Grundidee ist ja alle langweiligen Sicherheitsmaßnahmen und Einschränkungen wegzulassen und mal zu schauen was so passiert. Er nimmt absichtlich das am meisten "unhinged" KI-Modell und feiert die "Banger" die es bringt. Vollzugriff auf die Computer, inklusive API-Keys. Er pushed Code und schaut ihn sich nicht an...

Also für mich ist das Sicherheitslücken anmäkeln eher so wie auf ein Metal-Festival gehen, und sich beschweren, dass dort keine Schlager laufen. Und dass die Leute da alle betrunken sind... Also ich meine, Ja? Das ist korrekt? Aber war auch irgendwie Sinn der Sache?!

Nur das das hier kein Metal-Festival ist, sondern ziemlich albern.

[–] hendrik@palaver.p3x.de 1 points 8 hours ago (4 children)

Haha, danke für den Link! Es ist etwas schwierig mit OpenClaw und all den skurillen Dingen auf dem Laufenden zu bleiben. Wusste gar nicht, dass die bloggen und Schmähbriefe schreiben.

Aber die Anzahl der Sicherheitslücken zu quantifizierten, finde ich etwas dumm. 73? Da gehe ich wohl eher mit Peter Steinberger... Das Ding ist eine große Sicherheitslücke. Und was soll überhaupt als Lücke gelten, wenn es mit Absicht vollkommen freie Hand hat? Ich denke das Wort "fail" trifft es da schon sehr gut.

[–] hendrik@palaver.p3x.de 1 points 18 hours ago* (last edited 8 hours ago)

Indeed. That looks nice. It's about 2bit quantization. So not sure if it translates to the other paper. I had a quick glance at their code, and it's specific to the Llama2 and Llama3 architectures. So, it'd need to be enhanced for other models. And what might be a bummer: they load the model at full precision to calculate the activations. That means you're looking at a system with ~480GB of (V)RAM. And we don't have machines of that size show up on the AI horde. (As far as I know.)

I think we're looking more at crowdfunding research here. I mean sorry for being overly negative. I'd like to see 1bit models as well. And I always love to see community projects and independent people push the limits. I just think the hard part is coming up with the research, the math... or even the engineering to combine two papers and adapt an approach to something. So we somehow need to crowdfund that.
In these two examples, seems the compute power isn't really the issue. I mean the 1bit training was doable on a single H100. And this LoRa isn't very complex either, and they're not using that many samples.

It just wouldn't fit on any of the 38 LLM workers currently online on the AI Horde. Not even remotely. So this and the Horde is kind of a bad / impossible fit. However, I still think compute power wouldn't be the biggest issue, we can rent that by the hour. And it's not even hard to set up or that expensive. I think the main issue is coming up with the math and the code to produce something useful. So maybe we need a research community. And these things already happen. I mean the llama.cpp community has long been working on quantization and pioneered some things. There's people on Reddit discovering new things. We've had random(?) individuals contribute substantial advancements to image and vide generation. There have been communities/projects like RedPajama, who trained a model from grounds up (and assembled the dayaset)... Seems very low precision quantization is just a tough nut to crack.

Seems to me Bitnet needs a pile of money in compute, plus a team of bright researchers to improve upon. NanoQuant doesn't perform as good as any 4bit or 8bit model with a similar resource footprint, so it'd need way more research as well. And RILQ is a bit specific, it'd need further research as well. It's not entirely clear whether that happens. There's something like publishing bias. Sometimes researchers don't publish negative results. So maybe they tried to apply it to lower resolutions, failed, and didn't write a paper about how they failed. So I'm not sure where to go with this. There isn't anything we could run or just apply as is.

And the AI Horde does inference with fixed scripts. On something like gaming GPU's and Apple silicon. People who bought a few old 3090s. But that's inference only. What's needed for general research is a new project. It'd need to provide you with cloud GPU, launch Docker containers for arbitrary workloads. And expensive enterprise GPUs, or infiniband clusters of some. So the entire software needs to be scrapped and replaced, and the hardware improved as well for cutting edge research. We maybe can call this new thing AI Horde as well. But it'd be an entirely new thing.

And I guess alignment, harvesting user data and preferences from the user's interactions could be done as well. At least from the technological perspective. I don't really know if the audience likes that. Depends a bit on how it's done.

[–] hendrik@palaver.p3x.de 1 points 1 day ago* (last edited 20 hours ago)

That kind of ketchup is unacceptable anyway. I mean what would you use it for? It just tastes of sweetened, cooked tomatos. Just buy curry ketchup. And put mayo on the fries. Or Joopie sauce.

[–] hendrik@palaver.p3x.de 2 points 1 day ago* (last edited 1 day ago) (2 children)

Sure. I mean I know we can do it for 4bit or 8bit quantizations. Question is just if it can be done for 1bit. As per Microsoft's first Bitnet paper, the answer was a clear: No. Seems I've missed the NanoQuant paper from a few days ago. They did post-training quantization. But the numbers don't look impressive to me? I mean I'm not an expert and have just skimmed it. But the benchmarks in table 3 look like some very old model from a few years ago. Every tiny modern model can reach better scores. And the perplexity in table 2 doesn't look great to me either. Sure, you could run a 1-bit version of a 70b parameter model... But that's just worse than a 1b parameter model. So you can probably just skip the entire effort with the 1-bit quantization, download the smallest variant there is, run it straight out of the box, and it'd be both more "intelligent" and at least 4x faster.

I mean ultimately, you're right. I'd like to know the benchmark results for something like a 70b model. Seems they only list those for small models. And it's a shame they didn't release any weights.

But I don't see why we need a collaborative Horde effort to find out... They write a 70b model can be compressed on an H100 in 13 hours. And for example runpod.io charges $2.39 per hour for such an instance. So all we need is a bit more than $30 to find out?!

[–] hendrik@palaver.p3x.de 2 points 1 day ago* (last edited 1 day ago) (4 children)

I was under the assumption, the main issue with these models is the costly training? As they can't be quantized, but need to be trained natively at that precision, and the process is very expensive...

I mean I didn't read the newer papers. Did they solve the issues with post-training quantization and that's usable now? Otherwise we might be looking at either, very expensive training, or very high perplexity and the model isn't usable for anything.

[–] hendrik@palaver.p3x.de 4 points 1 day ago* (last edited 1 day ago) (1 children)

Depends a bit on the country. In the United States, for sure. That's just open corruption and you scratch my back and I'll scratch yours. The government funnels $500bn taxpayer money into some Project Stargate, and God knows how much into really dark stuff with Palantir. Musk even "worked" for the government for a while... And next to the corruption money, these people are buddies. And they're all working towards the same goal. Some idea of an apocalypse.

In China, I don't think they need to bribe the government. It was the CCP who came up with the idea in the first place. And the AI race between China and the USA is yet another thing.

For Europe, I'm not so sure. There's a bit more nuance here? I mean Ursula von der Leyen is an AI shill as well. She frequently likes to talk about it. I don't think there's as much open bribery, though. And I still hope they're aware of the situation with US companies, how we diverge in our goals, and partnering with Palantir or X is likely going to end us up in a lot of pain... And the EU loves to regulate. And our own AI companies aren't as big. So there's that as well.

[–] hendrik@palaver.p3x.de 3 points 1 day ago* (last edited 1 day ago) (3 children)

Yes. I don't think it's settled yet. There's still many trials going on. The industry still tries to push the limits, including really weird stuff like Elon Musk probing if it's okay to allow deep-fakes of random existing women and minors. I think lawmakers are having a difficult time to keep up with the pace. AI companies drown them with their near unlimited resources. We need to come up with new regulation. Fight all the court battles, overhaul copyright and discuss things in society... And then there's preexisting influential structures, like Disney, the copyright industry... Sometimes they're on opposing sides, sometimes they dabble in AI as well... I mean it's complicated. And a long process. And it's difficult to defend things. I mean I also defend my server. But it's more an open war than anything with rules and terms.

 

This is a bit smart-home specific... I've long wanted to build some smart speakers with Home Assistant's voice assistant and a multi-room music solution. Sadly I could never get Snapcast work to an acceptable level, but seems there is a new kid in the block now.

Sendspin (formerly Resonate) is an open synchronized music protocol. Currently being implemented within ESPHome. A few days ago they fixed some more memory issues and seems it's getting along. Maybe we'll get an open and working multi-room audio solution added to ESPHome soon.

I got it working on both an ESP32-S3 and an old ESP32 (with PSRAM). It's got nice features like the title, artist, album and cover art being transferred so we can just display that on a small TFT. And it's just about starting to work well enough. Now they need to finalize it, get some more things changed around in ESPHome and I can finally assemble some synced speakers for the kitchen and livingroom. Yay!

Seems it's early days for the server-side of it as well. We got support by Music Assistant. Other than that the (few) libraries warn they're still more or less just a tech demo. So as of now this is somewhat limited to tinkerers, but IMO looking really promising.

 

Wir können in den letzten Monaten einen deutlichen Rückgang der aktiven Nutzenden feststellen. Doch was genau sind die Gründe dafür und was können wir dagegen tun?


Interessanter Artikel. Bin nicht sicher was ich davon halten soll. Hier ist die Statistik auf die der Artikel sich bezieht. Meine Instanz sieht auch einen massiven Ausschlag an neuen Nutzern letzten Februar/März aber auch hier im Threadiverse war es eigentlich immer schwer neue Nutzer zu halten. Wenn es denn tatsächlich auch echte neue Nutzer waren. Wir haben immer wieder diese Ereignisse mit einem Zustrom von Leuten. Aber ansonsten gehen die Statistiken für Lemmy eigentlich regelmäßig bergab. Und das ist auch schon eine ganze Weile so.

Und ich denke ich beobachte auch einen gesellschaftlichen Wandel. Also vielen Leuten ist das was mir wichtig ist zunehmend unwichtiger?! Oder wir haben resigniert? Aber eigentlich wäre doch im Moment ein guter Zeitpunkt um von den Plattformen die schon lange zunehmend kommerzieller und manipulativer werden zu Alternativen zu wechseln, die von Menschen für Menschen geschaffen werden?

 

Ich wollte nun mal endlich auf einen moderneren Passwortmanager wechseln und hatte dafür Bitwarden und KeePass im Auge. Ich hatte mich gerade schon fast für Bitwarden und zugehörigen Vaultwarden Server entschieden... Und nun muss ich feststellen, dass der Bitwarden Desktop Client überhaupt nicht in meinem Debian-Repository ist. Im F-Droid Store auch nicht, allerdings bieten sie dort wenigstens ein separates Repository an... Warum?

Sollte ich meine Idee nochmal überdenken und vielleicht doch KeePass oder etwas ganz anderes verwenden? Ich hätte gerne etwas das einfach zu benutzen ist, so dass ich es auch Freunden/Verwandten andrehen kann, gerne so dass man direkt seine 2FA Codes und Passkeys drin speichern kann. Aber bitte Freie Software und keine komischen Geschäftspraktiken und Snaps oder manuelle Installation, gerne auch kein Flatpak sondern einfach die gute althergebrachte Funktionsweise von Linux-Distributionen und ein "apt install ..." mit Updates, Maintainern die ein Auge darauf haben und wofür man generell so Linux-Distributionen überhaupt erst hat. Oder gibt es sinnvolle Erklärungen warum man das nicht macht?

 

Experten zeigen, wie KI Machtverhältnisse verstärkt und soziale Ungleichheit vertieft. Regulierung könne helfen.

 

Richard Sutton is the father of reinforcement learning, winner of the 2024 Turing Award, and author of The Bitter Lesson. And he thinks LLMs are a dead end. [...] LLMs aren’t capable of learning on-the-job, so no matter how much we scale, we’ll need some new architecture to enable continual learning. And once we have it, we won’t need a special training phase — the agent will just learn on-the-fly, like all humans, and indeed, like all animals. This new paradigm will render our current approach with LLMs obsolete.

Long interview from the Dwarkesh Patel Podcast. I like the more technical/philosophical arguments. And I think it's a more nuanced perspective than what we normally hear about AI.

https://piped.video/watch?v=21EYKqUsPfg

53
submitted 4 months ago* (last edited 4 months ago) by hendrik@palaver.p3x.de to c/selfhosted@lemmy.world
 

The Wikipedia article says Cloudflare has been used to host hate speech, websites with illegal content and forums connected to all sorts of illegal activities. And I see them being used by a lot of decent webservices but shady ones as well.

So my question, can Cloudflare be used for something alike "bulletproof hosting"? Does anyone know if they collaborate with law enforcement or care once someone sends a mail to the abuse contact? Or if there's a way to find information about a Cloudflare protected server for the public?

Hypothetical question, I'm just curious and I thought maybe someone here has first-hand experience with getting their account terminated or reporting content or doing piracy via them or whatever...

3
submitted 5 months ago* (last edited 5 months ago) by hendrik@palaver.p3x.de to c/ai_@lemmy.world
 

Short TED talk of Dustin Ballard, who runs the Youtube channel "There I Ruined It". About role of AI in music and creativity.

on piped.video | ted.com

1
submitted 7 months ago* (last edited 7 months ago) by hendrik@palaver.p3x.de to c/ask@piefed.social
 

Doesn't have to be a state secret. Just any Information to which access is restricted or it'd be dangerous or undesired if it were just handed out to the public.

Edit: Don't reveal any secrets here. This is a public forum. Proceed with caution when answering this question.

 

Tl;dr: I think we have too much "empty" content and noise here and it drags down the place for 2 years now. Does PieFed include an approach to change the situation?

I'm sorry, this is going to be a bit of a rant. And about PieFed's role in the "Lemmy" community and more broadly, what I think the place should be about. Feel free to skip this, unless you have a good amount of time to waste to read my long post and you want to think about the future of the community here.

To preface this: I'm mainly here on the Threadiverse for the comments. To have meaningful conversations with people. That could be the charm of this place. Yet, that's regularly not what happens here.

The high-frequency posters use Lemmy to dump the news of the day and re-post memes. And that's okay if people want that, I myself try to cut down a bit on news shaped by social media, so again it's mainly the comment thread underneath that I deem useful, not the post itself, since we have the news at a bazillion other places and it's not what sets this place apart. (Plus I think following the outcry of the day is corrosive and usually less informative than it seems, so I went further and actively unsubscribed from many of the big communities here.)

And the now more meaningful (to me) part isn't huge by any means. I comment on things and write answers to questions, some communities work very well and it leads to a conversation or I can help someone with their Linux woes. Half the time at least I type something into the void and it feels like I've wasted my time since I don't get any replies, maybe one or two upvotes at best and not even OP engages. So I wonder why they even made the post. Clearly not because they want to talk about something.

I think the interesting part of the Threadiverse needs to grow so I can have meaningful conversations here. When I look at the user count of Lemmy, I see how it stagnates at about 45k users for 2 years now. Sometimes we get an influx of a few thousand users but we're not attractive to them, so we always lose them again. And the place just stays whatever it is. I think not really attracting people and at the same time losing that many people constantly (who actively volunteered to have a look at the place) tells us something.

I think we could do better than that and set the place apart from countless other platforms in many ways. But that seems to a minority opinion in the bigger Threadiverse. The Lemmy devs regularly say it doesn't need to grow and it'll maybe grow organically (which it doesn't). Most users here tell me we need to dump more posts in an desperate attempt to kickstart engagement. I think we've tried that for 2 years now and it clearly doesn't work. On the contrary, it's kind of empty (or fabricated) content and I'll find out once I try to engage, that these are lower quality, less engagement than some other posts. And it actively drowns the few people talking to each other in added noise. I think the idea to address the issue this way is exactly why Lemmy stagnates and why we always lose all the users that come here, sign up to have a look and then leave again, because this isn't what they've been looking for. (And this is a multi-faceted issue, we have some other drama and issues here as well, but this post is long enough, so I'll skip that here, feel free to add your perspective in the comments.)

Now this week I've complained a bit, since I saw piefed.social communities with really high-quality conversation. And then the same people come, determine we need more content, and they dump re-posts of the lemmy.ml equivalent over their heads. And then I've taken tens of minutes out of my day to reply to posts elsewhere (not a piefed community) and give a nuanced perspective, only to find out it's unmarked Reddit re-posts, and I've basically wasted my time. It wasn't a genuine question in need for my answer, I was betrayed, tricked into increasing the number of comments underneath something that wasn't even genuine. When I could have spent that time interacting with high-quality conversations instead, which definitely exist as well. It's just that those people drain that. And I can't even tell which is which.

So it actively takes away from quality content. And I end up with a feeling like with the Reddit content bots, fabricating engagement. Which I dislike and specifically avoid. And it makes the entire place feel kind of empty to me, despite the many posts we have each day.

I think first of all people really need to stop dumping posts in an ill-conveived attempt to help. It's a misconception. We need more comments here, not posts. Yet they do the opposite and their user profiles rarely have comments, just hundreds of posts. If you want to grow and foster the place, add comments.

PieFed

That's my perspective, feel free to tell me how it feels to you. I'm definitely not against posts, just against fabricating them, and focusing on an unfit approach instead of doing the right thing.

Now my question: Does PieFed want to address that issue (if it really is an issue to more people than just me)? Is PieFed just a piece of technology, connecting me to the same community, just with an arguably better approach? Or does it go further? Push towards a certain atmosphere, change the community and behaviour? Do we do higher quality communities on piefed.social or are they basically the same thing as the ones before, just on a different domain? Do we go as far as to kick the re-posters so at least the posts aren't just exactly the same?

That'd be mainly social engineering. And I'd really welcome if we had ideals and a clear vision of where to go. We kind of have that. In contrast to some other Fediverse software where I can't see a clear vision.

And then we have technology. We could devise tools to address it. And PieFed already is about providing better tools to address some things. We have an ambivalent view of concepts like Karma. And algorithms to steer attention. I could try to address this with software. Calculate scores and devalue everyone who dumps posts and doesn't contribute to the conversation. That's likely going to give some advantage to conversation itself and foster genuine engagement. Do we want to do that?

And as a bonus question: What's with the entire voting system? Seems I deem different things interesting than what's popular. And that's all the scores underneath posts and comments tell us. So it's of little use to me. A post with 5 upvotes could be as interesting as one with 250 of them, and that happens each day to me. Once I switch the sorting method from "new" to something else, what it does is make lots of interesting content disappear from my feeds.

References:


I've "flaired" this "Feature request". Mind this is an opinion piece containing my perspective (and preaching). I'd like to hear your's and request the name PieFed to encompass a clear vision, to be not just technology but a broader approach to shape the nature of the society we want to create. And put in lots of effort to actively lead us towards accomplishing more than we do today.

And I definitely need some good ideas and tools to turn my feeds into something that caters to my own needs and wants. If there's some overlap with other people, we could talk about some specifics.

14
submitted 7 months ago* (last edited 7 months ago) by hendrik@palaver.p3x.de to c/connectasong@lemmy.world
 

Connection: Another very famous masterpiece from one year earlier. Also very long and with many melodic parts. Both talk about driving fast on the highway.

 

I just found out I can buy a decent 400W solar panel in the local hardware store for around 90€ these days.

Are there people around with experience in off-grid solar? There is quite some supply in cheap MPTT charge controllers on the internet. And I can't afford a 700€ power station. But I would be able to buy a few power tool batteries or one of the lead-acid batteries people put in their caravan. Are there projects building a power station myself? Is this even worth it?

Maybe someone alredy wrote a blog post with recommendations or findings and failures along the way. Or has something similar running at home?

(Thanks to the mods for steering me towards the correct community.)

 

I'm developing a small Python webapp as some sort of finger exercise. Mostly a chatbot. I'm using the Quart framework, which is pretty much alike Flask, just async. Now I want to connect that to a LLM inference endpoint. And while I could do the HTTP requests myself, I'd prefer something that does that for me. It should support the usual OpenAI style API, in the end I'd like it to connect to things like Ollama and KoboldCPP. No harm if it supports image generation, agents, tools, vector databases, but that's optional.

I've tried Langchain, but I don't think I like it very much. Are there other Python frameworks out there? What do you like? I'd prefer something relatively lightweigt that gets out of the way. Ideally provider agnostic, but I'm mainly looking for local solutions like the ones I mentioned.

Edit: Maybe something that also connects to a Runpod endpoint, to do inference on demand (later on)? Or at least something which I can adapt to that?

view more: next ›