Russ

Your son and daughter will continue to learn new things as they grow up, a LLM cannot learn new things on its own. Sure, they can repeat things back to you that are within the context window (and even then, a context window isn’t really inherent to a LLM - its just a window of prior information being fed back to them with each request/response, or “turn” as I believe is the term) and what is in the context window can even influence their responses. But in order for a LLM to “learn” something, it needs to be retrained with that information included in the dataset.

Whereas if your kids were to say, touch a sharp object that caused them even slight discomfort, they would eventually learn to stop doing that because they’ll know what the outcome is after repetition. You could argue that this looks similar to the training process of a LLM, but the difference is that a LLM cannot do this on its own (and I would not consider wiring up a LLM via an MCP to a script that can trigger a re-train + reload to be it doing it on its own volition). At least, not in our current day. If anything, I think this is more of a “smoking gun” than the argument of “LLMs are just guessing the next best letter/word in a given sequence”.

Don’t get me wrong, I’m not someone who completely hates LLMs / “modern day AI” (though I do hate a lot of the ways it is used, and agree with a lot of the moral problems behind it), I find the tech to be intriguing but it’s a (“very fancy”) simulation. It is designed to imitate sentience and other human-like behavior. That, along with human nature’s tendency to anthropomorphize things around us (which is really the biggest part of this IMO), is why it tends to be very convincing at times.

That is my take on it, at least. I’m not a psychologist/psychiatrist or philosopher.

I mean, sure - but if you really don’t trust Apple to keep their word, then it wouldn’t matter if their Health app was FOSS or not. iOS itself is still (and probably forever will be) a closed source operating system. That gives them the power to do anything, including hijack the data from FOSS apps.

I have zero interest in Recall, but I thought it was already done on-device? IIRC it always was that way, which is why it’s only available on new computers containing dedicated “neural coprocessors” I believe was the term.

Now given that it’s closed source, you have to trust that they aren’t silently sending data back to themselves - which is where my problem lies, I don’t trust them in the slightest.

Jesus christ, that is absolutely disgusting. Texas has plenty of issues they should actually be working on.

I’m glad I managed to get out of there years ago.

I can’t say that I’ve heard of them, no. I don’t have any need (or desire) to do any sort of identity verification within any of my own personal projects (and I have not been involved with anything of the sorts at my workplace). Because of this, I don’t have any insight or thoughts I can provide on them unfortunately.

In the context of Fediverse administration (or any service that you run yourself), even with a service that “handles it for you” I still personally wouldn’t want to step into any of it.

I highly doubt it, one that required PII to sign up would be very unlikely to have many users (especially in the current climate, so to speak).

And from the admin side, that sounds like a nightmare to deal with.

No, because since it’s only a third party app implementation, tags wouldn’t follow if I go from my phone to my desktop or any other device. It also just seems kinda… Strange?

Do you keep a journal of those you meet in-person? No judgement if you do, but if your reaction to that question was “Eww, no!” but also do user tagging I would be very curious as to what the difference is for you.

Anyways, for problematic people they either get blocked or banned (the egregious ones) which by nature of it being a first-party feature is already synced.

According to another user in here, blocking on Mastodon actually works. So seems like it is possible to do in the Fediverse.

I was not aware of this, but their implementation of how they do this does bring up the limitation I mentioned. The other user cannot see your posts only if you are on the same server:

If you and the blocked user are on the same server, the blocked user will not be able to view your posts on your profile while logged in.

I actually thought blocks were public already.

They’re not, well - the operator of your instance could go into the database and view it that way in the same way that they can see your email address. But aside from someone who has database access to your instance, blocks are not public. What is public is the list of defederated (“blocked” so to speak) instances for an entire instance (this can be viewed by going to /instances of any instance), which might be what you were thinking of?

And personally I don’t see how it would be an issue if people that I haven’t blocked can see who I’ve blocked.

How exactly would you enforce that, though? If your blocks were public, all the person who you’ve blocked would need to do is open a private browsing window and look at your profile to see that they’ve been blocked.

If we’re looking at blocks as being a safety feature, I would think that having your blocks broadcasted to every single instance would be classified as harmful and a breach of your privacy. This is why although an instance that you register with has to have your email address that you signed up with, they don’t broadcast it to all other instances (same with the encrypted value of your password) - because otherwise it would effectively be public.

Perhaps I’ve just got the wrong stance, but considering that you can never block someone from viewing your content with an absolute guarantee (if the blocks were broadcasted, you still couldn’t prevent someone from just simply logging out, or standing up their own instance and collecting the data anyways) I would not consider that tradeoff to be worthwhile. Not that my stance has any weight since I’m not a maintainer for Lemmy (or any of the Fediverse software), but I wouldn’t be surprised if that has at least come up to those who are developing the various Fediverse software.

Aside from the rest of the discussion that has already occurred here, I’m not actually sure how this would work from a technical perspective.

You and I are on two completely different instances, if I were to block you, how is your instance supposed to know this in order to stop you from reading my comment?

The only way I could see that working is if the list of users you blocked were federated too, and effectively made public (like votes currently are) - which seems counterproductive to the problem at hand.

Then what happens if you post in a community where someone you’ve blocked is a moderator? Or if you block the admin of another instance? If you can “cloak” yourself from being moderated by just blocking them, that seems like an exploit waiting to happen. As far as I’m aware, on Reddit blocking a user doesn’t hide your comments from them - but they can no longer reply to them, and I assume this is why that is the case. Unless that has very recently changed.

The biggest difference between Lemmy (and all software within the Fediverse - for example, I’m pretty sure Mastodon is this way as well), is that there is not one singular authoritative server. Actions like this need to be handled on all instances, and that’s impossible to guarantee. A bad actor running an instance could just rip out the function that handles this, and then it’s moot. I mean, they wouldn’t even need to do that - they’d have the data anyways.

You could enforce it if both users are on the same instance I suppose, but this just seems like it would only land with the blocking feature being even more inconsistent.

How is that the case? I’ve got pretty much zero experience with decompiling software, but I can’t say I’ve ever heard anyone who does say that before. I genuinely can’t imagine that it’s easier to work with say, decompiling a game to make changes to it rather than just having the source available for it.

I suppose unless the context is just regarding running software then of course it’s easier to just run a binary that’s already a binary - but then I’m not sure I see where decompiling comes into relevance.

I don’t see how that’s going to work out well. That’s asking to end up with a mess that you’re just going to have to rewrite anyways.

I do not even have a complete hatred for AI like a lot of folks do, but I don’t trust it that much (nor should anyone).

You’d be better off with an actual deterministic transpiler for that (think TypeScript -> JS but the other way around I suppose), not something with a ton of random variables like an AI.

This can sometimes come at the cost of intuitiveness however. As an example that just happened to me the other day, I was using Pinta which uses libadwaita and had opened an image to make some modifications to it.

All was going well until I wanted to save a new copy of it (and not override the original). The toolbar has all of these functions on it, open, save, undo/redo, etc… but not Save As.

Apparently there’s a tiny little overflow button on the far right side, click it and you get a whole bunch of functions - one of them being the holy “Save As” option I was looking for. I almost went down the route of making a copy of the image outside Pinta and then just overwriting the original.

Apparently the idea of making a copy of an image is blasphemy. Even Microsoft Word when they had first moved to the Ribbon UI made the save button have a little dropdown right under the save option to reveal Save As.

Don’t get me wrong, I love how some libadwaita apps look. Mission Center for example? Chef’s Kiss - but it’s a very simple application that all I need to do is open it to have a quick look at the very pretty looking graphs. Although the latest update seems to have gotten rid of being able to have the sidebar open persistently (now taking an extra click to change between performance graphs)… But I still need to double check to see if that’s intended vs being a bug before I judge that too harshly.

I cannot recall why I had this open on mobile, but sure!

https://ktor.io/docs/docker-compose.html

I’ve been working with Ktor (a web server framework for the Kotlin programming language) for about a month and a half, so the docs themselves make sense - but I haven’t been working on it from my phone, so…

Patching Comic Code? It was quite a while ago unfortunately, so I don’t have the exact commands available, but I used their Font Patcher tool in order to do so.

From what I recall, the tricky thing was actually getting the dependencies it required to be installed properly, Font Forge would be up and running but then the script’s errors indicated that it couldn’t resolve all of the necessary dependencies. Not sure what OS you’re on so your mileage may vary - but for Linux they now have an AppImage that looks to contain everything it needs, and for macOS/Windows if you have Docker available there also appears to be a pre-built container for it. There’s also quite a few examples that I don’t think were there when I used it, since I also recall not being 100% sure of what flags were needed to run it

I am surprised that its actually still a product they sell and seemingly update. Looking on their product page, wow it has git support - it can be yours for $22.99/month too!

(That will also require you to give your soul to them too, via a contract signed in blood)

Oh, so that’s what those Python notebooks are that I’ve heard people talk about!

Oh hey, someone else who uses Comic Code - greetings!

I remember when I first saw it, I laughed - and then it grew on me. Then it turned into “I can’t believe I am buying a derivation of comic sans” but it is actually a really nice monospaced font.

Only thing I didn’t like was having to figure out how to use Font Patcher to make a copy of it that supports nerd fonts, but it was a one and done process.

(I also don’t really like how it looks in my IDE the few times I find myself on Windows, but I don’t really blame the font for that one - looks perfect in the same IDE on Linux…)

No human would be stupid enough to change something’s name for zero reason.

Well, unless you’re a product manager at Google apparently… Though with them you’re lucky if its just a change of name rather than it becoming an entirely “new” thing, or just getting outright axed…

Ooh! Is that swap implementation the default? I got back into LE for the launch of the newest season, and while I haven’t had any problems on my Ally or Deck yet, I just finished the campaign so I’m barely into endgame - I hear the issues start as you get deeper into monos…

Funnily enough, I use Cachy on my desktop, but I don’t recall seeing anything regarding this, but I’m definitely happy to run it on my Ally too if it helps avoid future potential crashes.

I definitely don’t doubt their utility, despite my facetious comment regarding them - however it’s not likely I’ll ever be able to actually appreciate them (in this lifetime at least) due to my struggles with understanding far “simpler” areas of math haha.