Alpha3031

Well, I wasn't very active between February and August, but from what I understand the guideline was only changed in March to disallow AI edits entirely instead of allowing them if you pinky promised you've checked your sources and they really do say what you're putting into the article because a lot of people were pinky promising and those promises turned out to be either lies or bullshit when people actually looked at it.

So the answer to "Why do we assume AI-generated means slop?" is probably "experience", and any proposition starting with "If the content is good enough" is trivially true as a conditional statement with a false antecedent (i.e., ⊨(⊥⇒Q) for all Q).

People apparently like Granite for RAG specifically. I'm not aware of any evaluations with numbers though, so I'm not sure if it's a real difference or just vibes.

It isn't possible, in the same way that keeping copyrighted material out of Wikipedia isn't possible: there's probably going to be a backlog spanning years or decades, but some of those investigations will eventually complete and result in an action being decided on.

Hmm, its showing as roughly 9.4 kwords which is a little higher than ideal. I may raise the suggestion to spin a section out. Maybe § Emergence can have its own article and be only summarised in the parent? I see the New York subsection is also quite long.

Archive.org should have some old dumps, for example https://archive.org/details/enwiki-20220501 is supposed to be May 2022 (All the later ones look like incrementals).

So, I've taken a look over the weekend, I can't say I've fully settled on what I want to say yet but I'd like to make an initial attempt at articulating what my thoughts are:

To start with, I think the "fringe" aspect played a bigger part in your interaction than the sourcing guidelines (MEDRS) and policies (OR § PRIMARY). I do find it unfortunate that there isn't a better way of informing people wanting to edit the first time, but anything in a contentious topic, whether it's officially designated as such and listed in the list of contentious topics... Well they're probably not the best places to get started because there are a lot of people who do want to pick individual primary sources and overemphasise their weight relative to the body of other primary sources (such as those as reviewed by MEDRS secondary sources).

So if you're editing in a field with a lot of interest from, uh, altmed type people, you're probably going to see much stricter application of the relevant guidelines than other medical topics without such interest unfortunately, and the editors working in such areas are probably very tired of anything that even vaguely looks like promoting altmed. Megavitamin therapy at the moment would still be that even if it's vastly more plausible to actually have a mechanism of action unlike certain other types of altmed.

Here is fine.

> every paper has a review section built into the background section,

The parts of a source that do secondary source things are considered a secondary source. I highly doubt many of those sections are entirely de novo without reference to prior evidence syntheses though, considering how much time a proper systematic review takes.

> it's also a bad policy for seemingly controversial topics. the medical establishment is composed of humans with plenty of biases and groupthink, so a perfectly good idea can be shunned or ignored for decades if it's inconvenient to the establishment.

The hierarchy of medical evidence being bad is not really something that Wikipedia (or any encyclopedia) can solve. Something that overturns the establishment of a field would not really be an encyclopedia.

> technically the wikipedia regulations don't outright ban primary literature as long as you give the necessary caveats, but there are plenty of megalomaniacal admins don't care about nuance (e.g. MrOllie)

If you have any specific examples that you believe merit review you are free to forward them to me.

If you find an article is too long to read comfortably, please consider adding {{too long|date=October 2026}} (or whatever the date at the time is) to the top so that people can figure out what to do with it.

A lot of national libraries also undertake web archiving within their scope (e.g. under their national TLD) and I've sometimes found that they have items not covered by Wayback, but it's not as easy to request additions to it. The risk of having a single point of failure is certainly real, and it would certainly be nice for a coalition of libraries to maybe fund a second alternative and mandate the use of both, or something like that.

Literature reviews summarise the state of knowledge in a given field, and given Wikipedia articles are about the knowledge on a subject rather than the contents of a specific paper, I don't see why it would be unusual to rely on them. Is there anything specific that you find doesn't make sense to you? It's fairly common in, e.g. medical research for individual studies to be not relied upon and meta-analyses are certainly generally considered stronger evidence on any topic.

Academics citing any website should consider also pointing towards (or at least creating) an archived version via something like the Wayback Machine. Wikipedia is not the only website that changes over time, and most don't have a easily accessible revision history.

> extend to all subcultures

> every voice of opposition to totalitarianism

Unless your definition of "opposition to totalitarianism" is very different from mine, I would hope it is transparent that the examples given in the comment you replied to are not such voices.

I've been told back in the day it was quite normal get a physical, dead-tree book with similar information sent out to you every year, until people decided that was a waste of paper.

I thought most implementations of C have stdint (intN_t, leastN_t and fastN_t etc).

Memory addresses being 64 bits due to needing to address more than 4 GiB memory doesn't mean most integer instructions operate most efficiently with 64 bits. Instructions for 32 bit integers are still more efficient than 64 bit, whereas 16 bit operands require a prefix byte meaning they're less compact and cache efficient (on AMD64 anyway).

The supply of land is inelastic. Property developers can't get more land by building more houses on it (and as a corollary, the eventual owner also do not get charged more in taxes).

You've got encoder and decoder reversed. BERTs and other models that primarily convert text (or other input) into latent representations are encoders. Models that convert their internal representations back into outputs are the decoders (in the case of GPT et al., autoregressive decoders because they perform this decoding based on past tokens).

Sometimes 4 is skipped due to being considered unlucky.

Flip both, Cario Vmodei.

You mean complete control for open models, because you can run it on your own choice of hardware?

Enc-decs are usually harder to train at frontier scale. Not 100% sure what DeepSeek has done differently here initial read seems to be something related to layer reuse but I just skimmed things so far.

Also mid-range, A57 uses Xclipse 550 (RDNA 3) and A37 uses previous generation Exynos with Xclipse 530 IIRC.

Very interesting project, I like it. Just wanted to clarify though the sentiment analysis is just the count of stripped words and used to tag things with the emoji? I was initially expecting it to be a part of the actual command construction process (even though I couldn't figure out how that would be relevant) given how it was listed.

I feel like RLHF has a pretty obvious ground truth, human feedback is used as an (albeit noisy) signal of average human preferences. Same thing with RLVR and "solving the problem".

Diffusion LMs denoise a canvas which I personally find more interesting.

I don't really disagree that human cognition is essentially a predictive task though, as I understand it, predictive coding and related theories based on the Bayesian brain hypothesis are fairly popular these days (though maybe not clearly dominant over alterative models? IDK I'm not a neuroscientist). I imagine most people would draft a few tokens before refining them like MTP or diffusion though, if we do decide to use LMs as an analogy to human cognition.

Probably Fox would be the server with the most players but at DDK basically any would work and I think getting exposure to different playing styles would be worthwhile.

Maybe not to the extent you mentioned earlier (I definitely agree that it's not "focused" on this) but there is definitely asymmetric playout training in the training data (see comments on lightvector/KataGo issues #39 and #162 mentioning it for example), that is presumably how they got the tweak for playoutDoublingAdvantage (i.e. actually having a few thousand of the millions upon millions of training games be games where playouts have been doubled for advantage).

AU user also, just checked AI studio since that seemed like the best bet and both 3.8 and 3.7 show up (and can be used for chat in playground, though IDK what the limits for that are). Chat in gemini.google.com is also 3.6 for me but I'm on free tier lol so I don't exactly expect it to show up any time soon. I think there's also another free API beyond the AI studio one (which is 20 RPD free according to docs so not really useful) but I forgot where it was (Google cloud maybe?) and what the limits for that were.

The search model is probably flash-lite based on what they give to users who aren't signed in.