If that's what you meant: Google. Boom.

Also, perhaps you're a bit new to industry, but that's how these things go. They burn a lot of capital building it out b/c they can always fire everyone and just serve at cost -- i.e. subsidizing business development is different from subsiziding inference, unless you're just sort of confused and angry at the whole situation and it all collapses into everyone's losing money and no one will admit it.

That's a really cruel response, because this is important work. I don't want my kids beholden to bigco.

I think it's real & important.

I also wanna make sure people like me, who have to keep tabs on the intersection of "how can I help liberate from BigCo" and "how can I make a livable wage doing so"

It is, quite literally, real, but also something you shouldn't waste time on if you're already busy. (c.f. https://jmap.io/software.html)

In the spirit of TFA:

This isn't true, at all. I don't know where the idea comes from.

You've been repeating this claim frequently. You were corrected on this 2 hours ago. llama.cpp had early access to it just as well.

It's bizarre for several reasons:

1. It is a fantasy that engineering involves seats at tables and bands of brothers growing from a hobby to a ???, one I find appealing and romantic. But, fantasy nonetheless. Additionally, no one mentioned or implied anything about it being a hobby or unserious.

2. Even if it wasn't a fantasy, it's definitely not what happened here. That's what TFA is about, ffs.

No heroics, they got the ultimate embarrassing thing that can happen to a project piggybacking on FOSS: ollama can't work with the materials OpenAI put out to help ollama users because llama.cpp and ollama had separate day 1 landings of code, and ollama has 0 path to forking literally the entire community to use their format. They were working so loosely with OpenAI that OpenAI assumed they were being sane and weren't attempting to use it as an excuse to force a community fork of GGUF and no one realized until after it shipped.

3. I've seen multiple comments from you this afternoon spiking out odd narratives about Ollama and llama.cpp, that don't make sense at their face from the perspective of someone who also deps on llama.cpp. AFAICT you understood the GGML fork as some halcyon moment of freedom / not-hobbiness for a project you root for. That's fine. Unfortunately, reality is intruding, hence TFA. Given you're aware, it makes your humbleness re: knowing whats going on here sound very fake, especially when it precedes another rush of false claims.

4. I think at some point you owe it to even yourself, if not the community, to take a step back and slow down on the misleading claims. I'm seeing more of a gish-gallop than an attempt to recalibrate your technical understanding.

It's been almost 2 hours since you claimed you were sure there were multiple huge breakages due to bad code quality in llama.cpp, and here, we see you reframe that claim as a much weaker one someone else made to you vaguely.

Maybe a good first step to avoiding information pollution here would be to invest time spent repeating other peoples technical claims you didn't understand, and find some of those breakages you know for sure happened, as promised previously.

In general, I sense a passionate but youthful spirit, not an astro-turfer, and this isn't a group of professionals being disrespected because people still think they're a hobby project. Again, that's what the article is about.

From the outside, it looks like all you get out of this is feeling upset, and it makes us wonder how you misread so wildly.

There are two physical quantities for luminance, relative, and perceptual, so that passed along a nugget for those not as wise as you who might not know that :) As you know and have mentioned, using 0.5 with the luminance calculation you mentioned, for relative luminance, would be in error (I hate being pedantic, but it's important for some parties, a11y is a de facto legal requirement for a lot of work, and 0.5 would be spot on for ensuring WCAG 2 text contrast as long as used with perceptual luminance, L*)

> doesn't align with human perception

It is 100% aligned with how humans perceive brightness, in fact, it's a stable work product dating back to the early 1900s.

> Ultimately one could use 0.18-0.3 as threshold

Perceptual luminance and relative luminance have precise mathematical definitions, one can be calculated in terms of the other.

If you need to hit contrast K with background color C, you won't be able to treat this as variable. What you pass along about it being variable is valuable, of course, in that, given K and C, output has a range, i.e. if contrast algo says you need +40 L* for your text to hit APCA/WCAG whatever, and your C has 50 L*, your palette is everything from 90 L* to 100 L* and 0 L* to 10 L*.

However, you know that both of these claims asked about are blatantly false, and were distracted by the idea that saying those claims is false also implies all alternatives proposed are based on lies.

To wit, the ask was "Is it really true that there a no theories that are proven or discarded with this experiment, and that the Chinese have plans to do it much faster? Her video is pretty damning."

Both of those things are clearly false.

The Chinese part is blatantly false, to the point it can be worked out by a laymen who knows years ascend.

The Standard Model itself is in question, modulo semantics about proven/disproven and the philosophy of certainty, by any reasonable definition, theory is at stake.

When we get new theory, then we go hunting new particles, presuming its physically possible (as you point out with the incorrect idea that this might be being built to look for confirmation of string theory)

I understand the idea this won't find new particles, is it worth it?, but the idea this is unclear, confusing, misguided, or hoping for an outcome are trivially verifiable as false.

Things like:

- "The scientific goals...are unclear" (they are very clear!)

- "(modulo some error)" (reducing the error in the glimpses of deviation from the standard model is the interesting part, 5 sigma or bust, because that lets the theorists know how to progress. This isn't just "oh we'd like to reduce error bars, a less-entitled discipline would just get some grad students on SPSS", this is "holy shit...looks like we found something is fucky in our fundamentals here, but all we know is its off. we need to figure out by how much to give the theorists more data")

- "string theory et al" (I worry very much about the effectiveness of my communication if this is coming up, to be clear, no one is attempting to verify string theory, and it doesn't come up at all even in Sabine's arguments, no? )

The IEEE article lays out this is not about discovering particles.

No one thinks new particles will be discovered.

The investment is not based on speculating new particles will be discovered.

The investment is not based on bad theory that new particles will be discovered.

The investment is not to find a sneaky way to hopefully accidentally find new particles.

Investments in colliders in general haven't been spectulatively looking for new particles in decades.

As both the IEEE, open source information, and my comment lay out above, they are specifically for nailing down these previously-assumed-settled values in the standard model. Because getting more data on the things theory can't explain leads to informed revisions in the theory. The next pendulum swing after that data would be theory to tell us a narrow band of energies to look at for any new particles theory needed to fix the standard model.

Sadly, it absolutely 100% is. She's an amazing case of this. I'll spend the next 2 years gently walking people down from her clickbait, and I'll end up with net-negative karma for it.

I honestly had given up entirely until I saw a subthread about a month ago where people who knew the area were exchanging info of lesser-known youtubers who come and clean up her messes after.

The sad news is, they are getting more and more attention (I saw one over 400K+ views), but a lot of that just comes from being loud, proud, and aggressive, as well as having 30 minutes of video to justify the up front "hey, she's at best a not even wrong contrarian, and honestly, lets be clear at this point, shes a liar for views!!!!"

Thankfully I'm old enough to see this stuff happens in waves, within 2 years it'll become common wisdom that she's X, Y, and Z, and even if I disagree and just think she's misguided, that'll be enough for the tide to ebb.

IIUC the data we have:

2K tokens / 12 seconds = 166 tokens/s prefill

120K tokens / (10 minutes == 600 seconds) = 200 token/s prefill

> entire point...smaller download could not justify...

Q4_K_M has layers and layers of consensus and polling and surveying and A/B testing and benchmarking to show there's ~0 quality degradation. Built over a couple years.

[^1] https://news.ycombinator.com/item?id=43595888

>> [elided] What am I missing here?

> Intent. Neither you nor I know what the person to whom you replied had.

Here's the part you elided:

"I looked up the definition [of lying]. It says "used with reference to a situation involving deception or founded on a mistaken impression."

That makes it quite clear whether or not I'm missing "intent".

This also makes it quite clear that I am not making an ad hominem.

I am using a simple, everyday, word used to describe the act of advancing false claims, whether through deception or mistaken impression.

No? By the way, what is this comment, exactly? What is it trying to communicate? What I'm understanding is, it is good to talk down to people about how "they can't communicate", but calling a lie a lie is bad, because maybe they were just kidding (lying for fun)

> That comment is making fun of their wording. Maybe extracting too much meaning from their wordplay? Maybe.

What does "maybe" mean here, in terms of symbolical logic?

Their claim "we tested it and it didn't get better" -- and the link shows, they tested it, it did get better! It's pretty cleancut.

The link shows a significant reduction.

grep hallucination, or, https://imgur.com/a/mkDxe78.