“German Nuggets”
***
***
Hallo Leute,
day 23 of the German Advent Calendar, slowly slowly (or not so slowly) the Advent season is drawing to a close :).
So a few days ago, I did a poll where I asked you for one of the days what you’d like to see in the Advent Calendar.
The majority of you voted for the Yearly Oracle.
I didn’t expect that, to be honest, and so I asked myself, what would Jesus do.
But I didn’t like the answer, so I’ll do something different.
Seriously though, this morning when I got up, I was thinking that it’s maybe cooler to do it right around New Years. And that’d also give me a bit more time to think of some int… I mean … to connect to the 5th dimension and channel the oracle.
So instead, I’ll do the second place of the poll and that was… drum roll….
A Sneak Peek of my New Project
Yes, we already had the card game, and yes, the books are in the works as well, but I had an idea not so long ago and usually it’s best to implement ideas when they’re fresh and then let them grow slowly.
So… the project is called German Nuggets and I’ll just post a link to it, without explanation.
And it’s really JUST the pure “concept”. The simplicity is on purpose. No menu, no login, no nothing.
Told you it’s really simple :)
I have some ideas how to develop that and I might actually make it into an app with a premium version, but for now that’s the concept.
I’m really curious for your feedback, so let me know in the comments (here), how you liked it :)!
Have a great day and I’ll see you tomorrow.
Audio Speed
Dark Mode
Could you give an example for when someone would say “Doppelt hält besser”?
Ich liebe die Nuggets!
Tolle Idee! But I still would really like to hear your thoughts and explanation on ChatGPT, especially as I am a computer scientist/robotics engineer who researches and uses applied AIs
I’ve talked to ChatGPT a bit in German, and even managed to get it to converse with me while suggesting corrections for my grammar mistakes along the way. It was pretty fun! I don’t know how good its corrections were, but everything *seemed* on-the-level…
So here goes my “partially informed layman’s” perspective:
These gigantic language models are impressive and an interesting case study, but they’re a dead end on our quest toward general artificial intelligence.
It’s really like a “dialogue simulator” above anything else.
And I think it’ll do well on all kinds of tasks that involve statistics. Like… I was debating with a friend the other day, whether chatGPT like systems could replace a psychologist or coach.
I was very much “No way” while he argued for it, and actually, if you talk to a fine tuned model like this (suppose we could train it on thousands of hours of psychologist-patient conversations, which we probably can’t), it could probably give you a pretty fitting description of what your issues are and what you should work on. But that’s just a matter of statistics. Somebody with a codependency problem for example will say certain things a certain way and a large model can pick up on that.
But it lacks the “intuition” of a psychologist. This human moment of “genius” where on the surface everything looks like A, but suddenly you try B, out of a hunch.
A language based AI doesn’t have hunches, it doesn’t have intent and it doesn’t “really” understand what it’s saying and why.
Now everyone is “Wow” but that novelty will wear off soon.
And then there’s the problem of “steering”, which is already starting because OpenAI won’t let ChatGPT argue for fossil fuel use.
That’s a step too far in terms of control, imo.
All in all, I think these models are really good TOOL for some tasks, one of them being language learning. It’s REALLY good to practice with (though it runs into its limitations really fast when it has to reason why something is wrong. That’s gonna be hit and miss).
And these systems will be part of general AI. But they’ll be scaled back, because they’re too expensive for mainstream and making them ever bigger doesn’t make them ever better (Google did some research about this recently).
What’s your thoughts on all of this? And what exactly are you working on/with? I guess you use AI for motion control and spatial awareness of robotics?
Oh, and the explanation… I don’t think I can tell you anything that you don’t already know, but I will recommend a channel. Here’s a video about GPT3 and it was really insightful for me as a normie.
https://www.youtube.com/watch?v=SY5PvZrJhLE
Übergut, vielen Dank für dieses Geschenk!
AUSGEZEICHNET!!!!!!!! This is a good quantity for me for one session.
Camille
Delicious nuggets yum.
These are looking great!
I like that you link to articles so we can dip into
a topic in more depth…
I think we really need to hear you reading those tongue twisters! Perhaps an Ai quiz with tongue twisters could be a thing ?
grün – grünlich (greenish)
Someone was chatting to the other day about translating “pinkish” and they were certain this wasn’t possible in german -cool this is actually a thing.
btw, I feel obliged to mention “nonce” unfortunately already has an unpleasant meaning in british english. It has a different pronunciation to how you would pronounce “none” with a “ce” at the end – so spoken and with context you wouldn’t think of this meaning, but seeing it written down I did…
I really dig German Nuggets! It’s definitely something I could see myself flipping through for 15-20 minutes every morning.
German nuggets are a lot of fun, there’s some interesting info and just pretty awesome. I learned something new on 3 out of 4 that I looked at. You won’t easily find that kind of information in books.
I think it will be excellent!
Doubly brilliant. First that I got what I voted for without needing a bot to boost the count. Second that it’s such a lovely thing.
I like the German nugguts, especially Snäppchen. We don’t run across local usage of a word in textbook German studies.
Happy to simply join the chorus of approval for the nuggets! Each seems a fun way to spend a couple of minutes while actually learning something. Like “YouTube Shorts” or “Instagram Reels” but without ending up feeling like you lose 15 IQ points every time you get consumed by a sequence of those bloody dreadful things. (Do I sound like there’s a New Year’s Resolution in the making? Oh yes, I think I do…)
This might seem a little/lot mean-spirited on the back of that and all the work that you’ve put into making the Advent Calendar full and super-fun, but can I ask Santa Emanuel ever so nicely to consider a couple of changes to the comment notification system that you unveiled recently? The first is for that infuriating bell to stop animating (or never begin animating in the first place). I suppose the idea is a sort of “inbox zero” pressure(?), but I often like to leave the notifications unread for a while, and having that thing nodding away on every page is really awful. It also manages to consume somewhere between 8 and 10% of my CPU while it’s animating –per ydg tab!– this adds up pretty quickly, and sets the poor old fans ablaze. Admittedly I’m using Chrome on a pretty old Mac, but it’s still crazily resource-intensive for what it provides! The second potential change? While selecting Mark as read from that menu “does what it says on the can” it also navigates to the article+comment that’s being marked as read. I can kind of see the thinking here: “go there *and* mark it as read in one fell swoop/click”, which is thereby distinct from the simple “Go to comment”, but it’s often nice to just mark a few comments *in a row* as read from the menu without having the mechanism pulled out from under you when you click the first one.
Anyway, neither’s a big deal at all, of course, and I can certainly look at creative personal solutions if others disagree with my gripes :-) Thanks a million for the fun year on board!
Hahahaha, I was enraged by that bell as well and I was actually thinking about removing the wiggle.
The whole system is a plugin piece of software for the comment software that I am using, so there’s a limit to how much I can meddle around in it, but that wiggle is just css and I can certainly remove it. I’ll have you know though that your house will be colder because all the CPU heat the bell created will be missing.
Not sure about the link structure they’ve set up, though. I can access that, of course, but then I’d have to redo it every time they roll out an update because that’ll override my core change.
Aah, wonderful, a thousand thank yous. for stopping that perpetually nodding bell! Now that I’ve done a “Force Reload” to update the css it just sits there ever so peacefully. Bliss restored! And decreasing my house temperature definitely counts as a bonus at this end of the world right now: it’ll be 41C on Tuesday :-|
The other request was something of an ambit claim: (“While you’re down there…”). If it’s entangled with the page loading, rather than a separate/separable SQL call, then please don’t waste any more time trying to disentangle things! The interface is actually easy to deal with once you realise what’s happening, just slightly weird.
Actually, I just figured that I could rename it to “go there and mark”. Then it’s at least clear what’s happening! i can’t have such glitches on the UX masterpiece that is my site :D
Update: bells is restrained now.
The linking is a bit more tricky. They’re attaching the database call to a page load, so I think the best I can do quickly is have the page load again that you’re currently on.
They do offer a feature where they’ll mark comments as “read” once you load the window once.
Maybe that’s the better option. Most people won’t miss the individual marking, i believe.
What bell?
You can’t see it if you’re not logged in (and you’re not, because you’re anonymous ;)
But the bell with a number on it will still be available, Yes? Actually I liked the jiggle because it caught my attention, but I would eventually notice a number on the bell too.
The bell’s still there, the number is still pasted on top in a red circle, it’s just dead *still* :-) You might need to force reload (ie, hold down the shift key while reloading the page) to get your browser to update the css file for the site.
Do you by any chance happen to have a trick how to force a phone to do that?
Grr, the mechanism seems to change every 6 months. Currently (at least for Chrome) you can go to
Settings => Advanced => Site Settings => Data Stored, and then you need to select the site (or search for it) among the mammoth list you’ll see there. Click on “Clear & reset” in the resulting window and, in theory, all should be well.
Could become addictive!
That’s part of the idea :). It’s basically TikTok, but with actual value.
Really great! I like the range of categories. I especially like a good proverb.
That’s great! Fun plus a bit of useful vocab
Brilliant
Love it!
really good…thanks for this
I don’t often make a comment, but this I did like.
Neat! I clicked through a lot of them pretty quick. Which I think is a little dangerous because there’s really a lot of info in each one.
The categories are interesting (common mistakes, sounds like a native). I could see myself browsing by category if that ends up being an option. But it’s kind of nice to just click and get a surprise.
I think the lack of control is part of the core concept, otherwise it’ll just be “another blog”.
I am currently thinking about what kind of interaction I allow for premium.
I was thinking definitely a category filter, so you can pick or exclude categories that you’re interested in. But I don’t think I’ll allow direct access to the entire archive.
I know there’s a lot of people out there (me included) who would want to know just how many there are and which ones and then just binge the ones that have an interesting title, but yeah… that’s not the idea.