@Glorfindel If you would, it'd be nice if you made a script that would change old meta.stackoverflow.com/posts/[x]/[y] links with IDs less than 250000 so they're not broken (i.e., change the domain to meta.stackexchange.com).
Those pre-split links weren't edited as part of the rewrite scripts that took place in 2014 when the MSO-MSE split took place.
When the split happened, SE set the threshold at 250000 for post IDs on the new MSO, and links to posts with IDs lower than that redirect back here so those links aren't broken. However, the redirect only works for direct post links, not for history, timeline, or various other links.
The new site with the potential sponsorship, selling dumps to LLM (problem: the datasets are already on Hugging Face... they can't even do a search??), workforce reduction etc.
Someone at prosus must have realized that they were quite a fine bunch of deluded guys in thinking Stack would bring much $$$. And now they are desperate to cash on something, anything.
@M.A.R. bah, consider this: the high command has shown multiple times that they don't even know what the site is doing in the first place. So, it requires actual effort from them to come up with new ideas.
If the site was already making them enough money then they would not try to change anything because of operational inertia.
The fact that the high command had to come out of lethargy is a sign that they want more $$$
Seems reasonable to assume that the site is NOT healthy from a financial point of view.
> There's a few policies in the pipeline - here's one, and there's another one that's not been officially announced that a good many moderators are opposed to. I'll update when (or if) it goes through, but it's not great.
oh, wait, for a second I read that one as a different policy from the just announced one.
I've received email from "[email protected]" is it legit? How can i tell?
I won't quote whole thing but in short it's asking feedback for new feature
.. We’re testing a new feature and we'd love to hear what you think of it...
It includes sign up for chat "button" with link that star...
@SPArcheon they don't pick them up. There's "Research" category in email settings, so whoever dare tick it (or have it ticked by default) probably get those mails.
First, shouldn't the council come first? Second, if they want a small number of "evangelist" MVP style they will carefully pick players that are useful to the cause. Third, I assume that the research is not to check if this would work, it is to prove it will work. So... see the point above.
@JourneymanGeekOnStrike Wasn't Meta where evil users go?
@SPArcheon not sure what you mean, looks like you're over complicating it? My point is, that when there's new research (may it be for a new feature, or a new site) they blindly send those emails. In this case they also asked on MSE, but it's not related to emails as far as I can tell.
Also, what council? Mod council is no more, you mean the strike representatives?
I think Tink should get extra diamond once all of this is over. She's doing amazing job. While against my own opinions and ideas (as it pretty much nullify the strike), I can't help but admire this. Kudos!
@ShadowWizardStrikesBack Well they'd really, really have to make me then... You do realize they hire CMs only to fire them like 18 months later?... I have a good, stable job now. Why oh why would I ever be tempted to give that up for such a thing XD
Top answer on a popular (or infamous) question will always garner knee-kick votes.
I'm just glad that I divested myself away from this place - the events surrounding Monica's departure really affected my mental health. The more you care about something, the more something negative affects you.
@AndreasdetestsAIhype I would like to make some joke about suggesting you to move to some mobile gacha game, but I can't think of a decent one that would cause the same pain....
Maybe the recently launched Street Fighter one? I heard the rare characters rates are trash....
@SPArcheon I’ve had to deal with some mobile games… Or Minecraft servers as a child… Some of those can be quite toxic, especially when you can spend some money to earn yourself the right to harass other poor victims.
That reminds me of a guy that joined various servers just to do harassments, be a bully, and destroy everything others built on creative servers. Eventually, Mojang wrote an article on their website, praising this guy. I contacted Mojang over e-mail, but eventually got told by the support agent (ironically had the same name as the praised guy) to go fuck off.
@ShadowWizardStrikesBack Yeah. But they need a chat. Maybe it's a call to contribute with some code. I guess I could provide some low-quality code for a chat similar to the one on SE.
@ShadowWizardStrikesBack I don't like Discord, and a site like Codidact should have its own chat, and not rely on a third-party service like Discord, which is hosted separately.
@SPArcheon In your post, you call SE out for presenting lies to the media. That's fair, and well deserved, but for context; we started contacting the media about this particular strike before SE did; I mentioned doing so the 01.06, and we started compiling a list of media contacts/articles on the 04.06.
That particular article is from June 5, after we had started contacting news outlets and journalists.
I don't know if anybody started contacting the media about the strike before this.
For example, if I wanted to write the equivalent of
There are many automated teller machines in this city.
Would it be
There are many ATMs in this city.
or
There are many ATM's in this city.
(could get confused with possessive form or contraction).
or just
There are many ATM in this city.
...
@ShadowWizardStrikesBack if you're talking about a group of Community Managers, the right way to pluralize that would be to add an "s" at the end, I believe. The apostrophe would denote possessiveness, so "CM's" would be used to denote something that a single CM owns, I'd say (example: "the CM's prose was not that eloquent")
I personally never use apostrophes for pluralization (except maybe in the very rare case of pluralizing a single lowercase letter)
It's a pet peeve of mine
(But then I might just have a lot of pet peeves. There's a reason I have the most edits of all time on RPG.SE... and on our internal Teams instance :P)
@EvanCarroll Well it is raining and it has finally cooled down a bit (apartment is back into the low 22 degrees Celsius) but I also got a period soooo... Nah.
I watched 10 minutes of a video and read a few YouTube comment threads; it looks pretty much like what I was warning about, down to the (perhaps performative) surprise when the computer doesn't read minds, and "therefore all the rest of our philosophical standpoint is correct!". And what do you know, Robert Miles is active on LessWrong. One of the constraints on their theory is that it only applies to agenty things, of which genAI isn't (though systems that use them – such as corporations – can be). — wizzwizz427 mins ago
@wizzwizz4 any suggestions on how I can get a more "balanced diet" of info?
@wizzwizz4 For this, I'd advise learning enough about agenty AI that the stuff they're saying starts to seem obvious and boring; then you won't get as distracted by it.
The LessWrong school of thought definitely dominates online AI discourse.
A route I haven't tried (because I'm terrible at literature searches) but should work is to look at academic publications. Most of the best takes I've seen have been published in journals.
And ignore (read: take with a pinch of salt) everyone who talks about "aligning" non-agenty AI systems. They're not using the word "alignment" properly, which suggests they don't actually know what it means.
It doesn't go into the self-attention mechanism (which, unfortunately, I still don't quite understand), and you should take all the "and therefore the Wolfram Language is the greatest thing ever!" bits with a pinch of salt…
… and the signal:noise ratio trails off as the article continues; you can safely stop reading around:
> So what about trajectories? We can look at the trajectory that a prompt for ChatGPT follows in feature space—and then we can see how ChatGPT continues that:
And when he talks about how surprising its behaviour is, all that is actually really obvious to someone with a good understanding (or even a poor understanding, like me). But apart from that: solid, and should give you a decent background for wading through the hype.
More concretely: an "optimising agent" is something that receives a "worldstate" as input, and performs an action that modifies that worldstate such that, in expectance, it better satisfies some utility function.
GPT-2 isn't an agent, because it's not optimising anything. All it "cares about" is predicting the next word as well as it can: it only "cares about" what it does, and not the in-the-world consequences of that.
The GPT part of ChatGPT is, likewise, not an agent. However, because ChatGPT has a user feedback system, the overall system might technically be an agent, optimising for people pressing the thumbs-up button. If we assume this feedback is used for supervised fine-tuning of the GPT model, then after a while, it would tend to start showing adversarial behaviour.
For example, if I wanted to write the equivalent of
There are many automated teller machines in this city.
Would it be
There are many ATMs in this city.
or
There are many ATM's in this city.
(could get confused with possessive form or contraction).
or just
There are many ATM in this city.
...
