MHLoppy

Well there’s 11 posts between them so there’s not much moderation style to look at, but yeah they could probably be consolidated (the latter seems to have an active mod vs the latter’s dormant mod, so maybe that’s the better choice)

There’s !speedoflobsters@lemmy.world and !speedoflobsters@sh.itjust.works but both are small and/or fairly inactive

the atrocious webp format

I continue to be confused by the level of widespread hate WebP still gets. It’s old enough to be widely (albeit not universally) supported in software like web browsers, but new enough to provide similar-or-better (usually better) lossless compression than PNG (21,578 bytes for the original image) and typically better lossy compression than JPEG at comparable perceived quality, especially for the types of images typically shared on the internet (rather than say, images saved directly from a DLSR camera). It’s why servers bother to re-encode JPEG images to WebP for delivery - they wouldn’t bother wasting the compute time to re-compress if it wasn’t generally worth doing.

I can understand it if we were, say, 10-15 years ago when the format was still not super widely supported yet, but that’s basically where we are with JPEG XL and AVIF support right now too. If one of these two had exactly the level of support that WebP does right now then yes, of course we should probably use one of them instead - but we’re not there yet. Until we are, WebP often has the best compromise between compatibility and compression efficiency as far as image formats go, and that’s why a lot of sites do this re-compression thing using WebP. I gave some examples using digital art (one of the things I was compressing a lot at the time) a year ago in a related discussion: https://lemmy.world/post/6665251/4462007

A news website local to me recently-ish started choosing to deliver AVIF-compressed (or probably re-compressed) images the same way a lot of sites currently do it for WebP because my browser supports AVIF, so at least we are starting to see a token amount of uptake on the next-gen formats in the wild.

Thanks for sharing!

Fair enough - glad you’ve found it helpful (Y)

You’re making assumptions about how they work based on your intuition - luckily we don’t need to do much guesswork about how the sorts are actually implemented because we can just look at the code to check:

CREATE FUNCTION r.scaled_rank (score numeric, published timestamp with time zone, interactions_month numeric)
    RETURNS double precision
    LANGUAGE sql
    IMMUTABLE PARALLEL SAFE
    -- Add 2 to avoid divide by zero errors
    -- Default for score = 1, active users = 1, and now, is (0.1728 / log(2 + 1)) = 0.3621
    -- There may need to be a scale factor multiplied to interactions_month, to make
    -- the log curve less pronounced. This can be tuned in the future.
    RETURN (
        r.hot_rank (score, published) / log(2 + interactions_month)
);

And since it relies on the hot_rank function:

CREATE FUNCTION r.hot_rank (score numeric, published timestamp with time zone)
    RETURNS double precision
    LANGUAGE sql
    IMMUTABLE PARALLEL SAFE RETURN
    -- after a week, it will default to 0.
    CASE WHEN (
now() - published) > '0 days'
        AND (
now() - published) < '7 days' THEN
        -- Use greatest(2,score), so that the hot_rank will be positive and not ignored.
        log (
            greatest (2, score + 2)) / power (((EXTRACT(EPOCH FROM (now() - published)) / 3600) + 2), 1.8)
    ELSE
        -- if the post is from the future, set hot score to 0. otherwise you can game the post to
        -- always be on top even with only 1 vote by setting it to the future
        0.0
    END;

So if there’s no further changes made elsewhere in the code (which may not be true!), it appears that hot has no negative weighting for votes <2 because it uses the max value out of 2 and score + 2 in its calculation. If correct, those posts you’re pointing out are essentially being ranked as if their voting score was 2, which I hope helps to explain things.

edit: while looking for the function someone else beat me to it and it looks like possibly the hot_rank function I posted may or may not be the current version but hopefully you get the idea regardless!

“Hot” is a mix of recency and votes. The posts in your example score low on votes but very high on recency (<1 hour ago) and extremely high on the size scaling because that community ( !hp_fanfiction@literature.cafe ) is tiny with only two subscribers.

You may consider Scaled to be a more appropriate sorting option for when you’re viewing the communities that you’ve subscribed to, rather than the firehose of /all

Memory connected via the pci bus to the CPU, would be too slow for application use like that.

https://www.intel.com/content/www/us/en/content-details/842211/optimizing-system-memory-bandwidth-with-micron-cxl-memory-expansion-modules-on-intel-xeon-6-processors.html

The experimental results presented in this paper demonstrate that Micron’s CZ122 CXL memory modules used in software level ratio based weighted interleave configuration significantly enhance memory bandwidth for HPC and AI workloads when used on systems with Intel’s 6th Generation Xeon processors.

Found via Wendell: YouTube

edit: typo

Using ram doesn’t add anything.

It would improve access latency vs flash though, despite less difference in raw bandwidth

Please don’t give the US any ideas ;_;

Presumably the member states can decide to interpret it however they’d like, but for whatever it’s worth I’m just paraphrasing what political scientist William Spaniel (…who I thought would have had a Wikipedia page by now) has said on the topic of Article 5 (though the context wasn’t the US invading Greenland lol)

Additionally, it’s helpful to know the specific language used in Article 5:

Article 5

“The Parties agree that an armed attack against one or more of them in Europe or North America shall be considered an attack against them all and consequently they agree that, if such an armed attack occurs, each of them, in exercise of the right of individual or collective self-defence recognized by Article 51 of the Charter of the United Nations, will assist the Party or Parties so attacked by taking forthwith, individually and in concert with the other Parties, such action as it deems necessary, including the use of armed force, to restore and maintain the security of the North Atlantic area.

Any such armed attack and all measures taken as a result thereof shall immediately be reported to the Security Council. Such measures shall be terminated when the Security Council has taken the measures necessary to restore and maintain international peace and security.” (emphasis added)

Article 5 doesn’t actually oblige NATO members to defend anything by force, it obliges NATO members to decide what actions are “deemed necessary” and then to undertake those actions. If a NATO member gets invaded, everyone could – in theory – write a sternly worded letter and call it a day (though I doubt that would be the actual response). As you/others have more or less said, the actual action chosen would largely be the result of political will.

I don’t know how well this works for Macs, but is a multi-boot environment a possibility? You could have a separate OS set up for a group of tasks which you boot into when you need to do that. It seems a bit clunky compared to e.g., virtual desktops or similar though.

So they literally agree not using an LLM would increase your framerate.

Well, yes, but the point is that at the time that you’re using the tool you don’t need your frame rate maxed out anyway (the alternative would probably be alt-tabbing, where again you wouldn’t need your frame rate maxed out), so that downside seems kind of moot.

Also what would the machine know that the Internet couldn‘t answer as or more quickly while using fewer resources anyway?

If you include the user’s time as a resource, it sounds like it could potentially do a pretty good job of explaining, surfacing, and modifying game and system settings, particularly to less technical users.

For how well it works in practice, we’ll have to test it ourselves / wait for independent reviews.

It sounds like it only needs to consume resources (at least significant resources, I guess) when answering a query, which will already be happening when you’re in a relatively “idle” situation in the game since you’ll have to stop to provide the query anyway. It’s also a Llama-based SLM (S = “small”), not an LLM for whatever that’s worth:

Under the hood, G-Assist now uses a Llama-based Instruct model with 8 billion parameters, packing language understanding into a tiny fraction of the size of today’s large scale AI models. This allows G-Assist to run locally on GeForce RTX hardware. And with the rapid pace of SLM research, these compact models are becoming more capable and efficient every few months.

When G-Assist is prompted for help by pressing Alt+G — say, to optimize graphics settings or check GPU temperatures— your GeForce RTX GPU briefly allocates a portion of its horsepower to AI inference. If you’re simultaneously gaming or running another GPU-heavy application, a short dip in render rate or inference completion speed may occur during those few seconds. Once G-Assist finishes its task, the GPU returns to delivering full performance to the game or app. (emphasis added)

Would be curious to read the LLM output.

It looks like it’s available in the linked study’s paper (near the end)

What a dumbass, commenting lemmy.zip but from a feddit.org account /s

I can’t believe that my own copy of this spam is how I found out that we have lemmy <-> mbin dms working now lmao

If a few servers are linked up and talk to each other using TCP/IP (?) but aren’t connected to the wider network, that’s not enough for it to be considered another internet (but it could be an intranet).

If a few instances are linked up and talk to each other using ActivityPub but aren’t connected to the wider network, I think that’s not enough for it to be considered another fediverse.

No worries, easy mistake to make just based on naming. Hope you find an answer!