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 quick, 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.

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 art, same with some software and webdesign. Also works the same way without AI. The consumer will do the beta test. And any random messenger uses 150 dependencies and Electron, and two Gigabyte of memory. That’s hardly artistry either.

Crashes my browser tab once the download reaches 93%.

It wasn’t really clear to me where you want to go with this. I mean judging by conversations with my friends, there’s a 3h conversation to be had about every nuance of health. The diet, how to work out, how much to eat and drink. There’s rules of thumb. But in reality it’s a very individual thing. And also changes with the situation, for example if you go to the gym or running and want to progress, it’d be an entirely different story a few months later. Also some people have office jobs, some do 18,000 steps each day… And don’t get me started on mental health.

Diagnosing medical conditions is hard for AI. We got some news on that a few days ago. It’s good at exam questions. But doesn’t perform in reality. So I wouldn’t call it health agent considering those kinds of questions. More a shaman. Or alternative practitioner / healer. (Or it’d need to stick to specific things. Or we need a few more years of progress in AI.)

I mean, I think the available tools aren’t even half bad? There’s smart watches with all kinds of features, apps, dashboards… Training modes and advise. They can help you define goals, track your period if you have that… Water intake, activity levels. It’s not AI, but there will be summaries, achievements, reminders…
Just the privacy part is a bit tricky, as most of these ecosystems come as cloud services.

I feel you can’t just dump in the CSV values from your Xiaomi Scale and Garmin watch… And hope AI will figure out the correct math on your body… And then also come up with good recommendations.

As far as I know, there are a few local, selfhosted health trackers available. It’s a bit tricky to own the correct gadgets that connect to it… But I don’t think there’s anything with AI.

I mean to give proper recommendations, you’d need a very elaborate setup. It needs all the sensor values. Then correlate it with what you’re doing all day long… What you eat and how much you drink… The AI (or traditional algorithms) can’t see. So maybe it can calculate your BMI in a thinking step. But it’s a whole lot of math to then figure out if your too fat, or have muscle mass… And then find out what that means for your diet. AI won’t figure that out along the way. So you’re probably looking at a few thousands of lines of code, after reading a few textbooks on biology.

I mean you can try to vibe-code some agent. But I think your best bet is to look for some open-source software cloning Google Health, or something like that. (And then maybe you can write some MCP server for that. And an agent to interpret the aggregated results.)

There’s copyright infringement on one click hosters… And a loy of them offer slow, but free downloads. Some newcomers ask a friend to copy a movie from their harddisk or DVD collection… I mean piracy in general is a bit tricky for newcomers. There’s some good resources linked in the sidebar… But a lot of piracy isn’t exactly legal to do. And it’s not really ethical to advise someone to do something that might get them in trouble… And openly recommending things is illegal in some jurisdictions. But yes. Don’t do random torrents unless you know what you’re doing.

Well, I guess if they’re still online and do silly stuff, like not use a VPN, not have a Firewall installed on their computer… Or they re-use the VPS which also has their personal blog on it… There would be ways to do something. But that’s all very unlikely.

I mean the whois is a good idea. Admins will usually want to know what they’re dealing with, and where it’s coming from. But the rest of the steps really depend on how bored an admin is. The best course of action regularly is to block it and move on. There’s so much bad stuff hammering the average webserver anyway. Launching a counterattack is a bit illegal, so that might not be an option. And if some admin has a few hours to pass until it’s 5pm and time to head home, or do it as a hobby and have time to spare they might investigate. I’ve found some hacked servers that way, wrote a few emails. But in practice, 99% of the time there isn’t anything to accomplish.

A whois will likely not do much. It’ll turn out to be some large ISP, which rents out virtual servers and all kind of stuff to private people, companies and VPN providers. And that’s regularly how far you’ll get, a name if a large company. And you can then decide if it’s worth to take someone to court, somewhere abroad… (But sometimes an email to their abuse contact helps a bit. Judging by my experience they won’t ever answer. But sometimes it’ll miraculously stop. And most of the time nobody cares about a single complaint.)

As far as I know fewer active parameters means faster. There’s less arithmetic calculations to be done per pass. But all parameters need to be kept in memory, because they might become active the next pass. So it won’t save any RAM.

They have a short paragraph in the description. It has 80B total parameters, 3B active each pass. It achieves performance like a 30-60B model (10-20x, their claim). But is way more efficiant than that with only 3B active parameters.

Yes, that will be an issue. I guess not a technical one, Linux is perfectly able to fetch a token and connect to network shares etc. Not sure how that works with Email and the modern cloud office stuff. But likely, the IT department will have to enforce that policy as well. That’s why I asked if OP has to use software on Windows (11)… Otherwise, if it worked 4 years without issues… maybe there is no issue with Active Directory…

Uh. My condolences. Do they also force you to use the software installed on Windows? Otherwise you could just image Fedora and run it in a virtual machine inside of Windows 11. Technically, I guess that’d fulfill the requirement with Windows 11 on the computer… Just that you don’t use it for more than log in, start the Linux VM and expand it full-screen.

https://youtube.com/shorts/QAW9QpBG80s

The Firefox Browser has translation built in and it works fairly well. We have LibreTranslate as a self-hosted service, I think it’s okay… Not particulary good, more okay in my experience. And what I tend to do is just copy-paste text to my local LLM and tell it to translate. Most models will do it. They have to be trained on multiple languages for that, and can’t be too small. You could try one of the Ministral models at whatever size fits and doesn’t heat up your computer. But I bet the average model from Meta and Google will do as well, I think they all have multilangual capabilities these days. And for web use, I’d recommend using Firefox. I can read Japanese websites with that. It’s not perfect by any means, but low on the resources and it only takes a few seconds, even on battery power on my laptop.

Hehe. I don’t think there’s many alternatives… Every hype will flood public discourse. That’s the nature of it (and the reason why we call something a hype, in the first place). You can mitigate a bit for this with subscribing and unsubscribing to communities. But trends will reach even the remote corners of the internet. And unsubscribing from daily news does a bit of that as well. The less social-media aspects a platform has and the more focused it is, the less random hypes we’ll get… But I’m afraid this place is modeled more with social media in mind. And the userbase reflects what’s going on in society. So we’ll get these things in most mainstream communities.

I think two vowels is a bit too common of a combination of letters to filter by that. Maybe filter for all the company and product names? Anthropic, OpenAI, Claude, Gemini, ChatGPT, Palantir…
That’s probably not going to get you all the way, but that’s going to suppress some of the news articles.

Hmm, sorry. I’d guess your internet connection doesn’t work any more. So apt can’t fetch the packages. That’s kind of hard to debug, though. You’d somehow need to fix networking before you can proceed. But we don’t really know what broke. And if that’s really the only issue at play.

Maybe a Live-USB stick and a rescue mode can help here?! Other than that I’m out of options.

Clear up some space first and then try something like this:

dpkg --force-all --configure -a  
apt --fix-broken install  
apt-get -f install  

I’ve had issues like that before and oftentimes it’s recoverable. A bit unfortunately if the wrong packages got damaged. Can be quite an effort to get it going again and it depends on the exact situation.

Yes, I’ve scrolled a bit through your past activity.

I’m not sure how you intend to pull it off. I mean you’d need to regularly fetch the reactions to your comments/posts, and then manually deal with all those people individually, and put them on some list. Which I suppose is quite a laborious process. And tomorrow it’s going to be 60% new users reacting to your comments, so it’s an everlasting battle against the windmills? At least in bigger communities.

For a smaller set of people we can block them. Not sure if that removes their vote from the score. But yeah, that removes their comments as well. Or quit a community and find a different one, either with people who also like arguing, or with more similar-minded people.

-Personally-, I think the votes just don’t work well. And I don’t think retrofiting something onto that is going to have a substancial effect. They’re social-media-style votes. People will upvote simple truths. Whatever is en vogue in a particular filter bubble. People will mess with them. And what gets the most upvotes here is memes and the daily news… While my favorite posts about microcontrollers and Linux get 4 votes combined. So the entire voting system here is kinda meaningless to me. I sort my feed by “New”, and largely ignore them. And I don’t think it’s salvageable. At least not without a good idea to go down to the roots of where it goes “wrong” and someone put in quite some effort.
But PieFed’s other features are nice. We get notes, more nuanced reactions, flairs/tags… I think I’m more with that. But I guess I somewhat share your opinion on how the voting system should be different.

Social-Media style votes are broken in general. You can filter them out in PieFed. Put this in your CSS in the user settings:

.score { display: none; }  
.voting_buttons_new { display: none !important; }  

( adopted from: https://piefed.ca/post/26333 )

No clue who’s the Nazi here, did they do anything? Or is this just the average being pissed because someone else has different opinions?

PeerTube has a collaborative wishlist and community votes on new feature proposals. If you want to see this within the PeerTube project, consider adding it to the list, or upvote it if it’s already in there:

• https://ideas.joinpeertube.org/

Just learned, Grok now requires a paid subscription to generate nudes. That won’t stop the average slop, but maybe some weird stuff. Because these people need to put in some payment details. And they’d probably need to do some extra work not to get their personal credit card connected to illegal activities.

I think the only way for us to do something, is to complain to our legislators and make them pass laws to mandate watermarking for all AI services. With proper watermarks, we could do AI filters.

And automatically detecting if an AI image is an unconsentual deep-fake… I don’t think that’s possible with technology.