jtorsella

一昨日議論

Thank you for pointing that out. But it still doesn’t support what you said, which is that only 65% of pangram’s fully AI claims are accurate. The “65% strict accuracy” rating is followed in the same sentence by a “97.5% inclusive accuracy” rating for the same reasons I pointed out above: pangram is systematically underrating the percentage AI in content, and it is tuned against producing false positives. The 32.5% in between the inclusive and strict accuracy ratings is entirely due to it estimating that 12 fully ai generated works were 60-80% AI, not the reverse. You can see that in table 4.

jtorsella

一昨日議論

Your citation was incorrect. The data also shows that pangram reliably tells the difference between those two cases. The statistic you made up or hallucinated is not supported by your link, and you’re being disingenuous here.

jtorsella

3 日前議論

Did you read the paper you are linking to? It records a 0% false positive rate for evaluation of human-authored controls, and fewer than 5% of the hybrid and humanized papers had their AI levels overestimated by pangram. It is not completely clear from the data, but it seems that depending on whether the n=2 overestimates were “100% ai generated” assessments, the study you link says pangram’s 100% AI assessments were right anywhere from 95-100%

jtorsella

3 年前議論

Oh you’re right - sorry, I forgot that everyone knows that the key ideas of the article are correct so they don’t require any evidence.

To drop the sarcastic frame - the article either argues something that could be incorrect or it makes a meaningless argument. If it’s arguing something that could be otherwise then it needs to give evidence for the things it’s saying. It doesn’t and treats everything it’s saying as completely self-evident. But these are empirical assertions that need evidence to be proven true or false, and the article stands out in using zero evidence. Personally, I think that this inadvertently reveals the disdain for news is really part of a broader undervaluing of evidence in general, but I don’t think that’s provable really.

jtorsella

3 年前議論

It’s pretty interesting to read what’s basically a long series of substantive empirical claims without a single citation. All the evidence necessary to support this argument is furnished entirely within the series of loosely-connected ideas and free-associations - an incredible feat for any article, much less a stream-of consciousness. Wow!

If we existed in a world where “News doesn’t make you more informed; it just makes you more confident the information you have is all there is.” was something that could be false, it would probably be Not Good to present it as true and move on here. Luckily, since this is an an obvious analytic truth, the author is right to simply say it and move on. As a suggestion to the author, it would be helpful to list all the propositions which need to be justifiable a priori for his argument to be justified (in this case all of them), so that readers can know whether or not to take him seriously at first glance.

jtorsella

3 年前議論

I think this attitude is the reason that Elon is going to get away with agreeing with the statement in public. Most stories tiptoe around how extreme the statement is, and as a result he is now successfully deflecting on the issue.

The quoted tweet remains endorsed by the wealthiest and most powerful person on the planet. Hard to read? Disturbing? Yes, of course it is.

jtorsella

3 年前議論

It’s very obvious that Elon and co would rather us concentrate on a blog post by someone else than his own words, words which are the real cause of this controversy. To that end, and since I think the actual cause is too often elided in these discussions:

[in response to a challenge to those saying ‘Hitler was right’ to justify saying so, a twitter account replies] “ Okay. Jewish [communities] have been pushing the exact kind of dialectical hatred against whites that they claim to want people to stop using against them. I’m deeply disinterested in giving the tiniest shit now about western Jewish populations coming to the disturbing realization that those hordes of minorities that support flooding their country don’t exactly like them too much. You want truth said to your face, there it is.”

Musk replied: “You have said the actual truth.”

I think it’s clear why Yaccarino and Musk want everyone focused on how many impressions a small subset of the antisemitic content on X got, and not the breathtaking antisemitism (and racism, although clearly “hordes of minorities” doesn’t raise as many eyebrows because nobody seems to care about that one) that musk himself endorses (and has not apologized for or retracted since).

jtorsella

3 年前議論

This is very transparently a gross, cruel attempt to parody and belittle trans people, as is made clear by the people involved and the language:

“7 Tips on How To Come Out as Enhanced”

And the “believe the science” and “colonialist” bits are very much a conservative-doing-an-impression-of-a-liberal thing.

There’s worse if you read through their mission pages. I’m taken aback by the level of effort, honestly. It really astounds me how much money and effort there is behind ostracizing already marginalized people. It’s disgusting.

jtorsella

3 年前議論

I appreciate that you are making an argument and will respond when I get a chance later in the day and after I read a few of the links - I’m familiar with like half of them.

jtorsella

3 年前議論

If the existence of C cves in the kernel proves that it is impossible to write correct C, then by the same token any cves in any rust code prove the same thing about rust. This is such a lazy way of arguing. Say something about why the tradeoffs favor a more restrictive and less performant language or don’t, but don’t dismiss the work of many thousands of C developers that runs most enterprise systems with a knowing wave of the hand - it’s not serious.

jtorsella

3 年前議論

Oh wow, someone should alert the Linux kernel maintainers. Do you want to tell them that it’s impossible to write correct C code? And the rust compiler team, too. After all, if nobody can write safe assembly then whatever they’re doing is either unsafe or magically gets the computer to understand rust directly. Or are they relying on LLVM for their code generation? I forget what language that’s written in, but nothing “unsafe” happens there surely.

All code is machine code at bottom. Including the code that maintains abstractions convincing enough for you to think the “memory-safety” of rust or any other language is a static and guaranteed thing and not something that needs “unsafe” scaffolding to support it.

jtorsella

3 年前議論

I’m going to not engage on the Fauci/covid question (although I think you’re misremembering which subject he was less than forthright about - it was the efficacy of masking, and the dispute you’re talking about didn’t happen until later). But the real point the author is making is that neither “Hero Fauci smashes Rand Paul” nor “Did Fauci visit Wuhan in 2019?” are things that I want on my windows desktop default! Most of the trash isn’t even political, as the author mentions, but when it is political it is always extreme.

jtorsella

4 年前議論

The opposite is true:

  These 444 cases of blood clots are after an estimated 24.9 million first [Astrazeneca] doses, and 24.2 million second doses of the vaccine in the UK. Of the 444 people who developed blood clots, 80 died. Six of these deaths occurred after the second dose.

Source: https://www.bhf.org.uk/informationsupport/heart-matters-maga...

  Up to 23 November 2022, there were 554 suspected cases of myocarditis or pericarditis reported in the 18 to 29 age group following the vaccine. This is an average rate of 31 reports per million doses.

  In the 30 to 39 age group, there were 470 cases suspected cases of myocarditis or pericarditis reported in the same time period. This is an average rate of 27 reports per million doses.

  Studies looking at myocarditis and pericarditis after the vaccine have not found any increased risk of death or cardiac arrest, compared with being unvaccinated. A large study of 4 million vaccinated people in Denmark, published in the BMJ found there were no deaths or diagnoses of heart failure in people who were diagnosed with myocarditis or pericarditis after being vaccinated.

Source: https://www.bhf.org.uk/informationsupport/heart-matters-maga...

A higher rate of myocarditis in the highest risk groups compared to blood clotting, but there is a more than an order of magnitude difference in the severity of blood clotting vs Myocarditis. In addition, blood clotting from the Astrazeneca Vaccine is thought to be incidental to immune response development, while Myocarditis with all covid vaccines is related to the strength of the immune response and is also a symptom of COVID-19.

jtorsella

4 年前議論

Interesting caveat, I didn’t know that.

jtorsella

4 年前議論

Your voting history is publicly available (nothing about who you vote for, just whether or not you vote in a given election year). There have been several studies showing that messages along the lines of “Your voting history is public, make sure to vote this year.” Or “keep up your voting streak!” are extremely effective, I believe the most effective single messages we know of or at least the most well-known ones.

Both parties and countless civic orgs do this now, it’s been a major thing since 08.

The original experiments were randomized, but usually if you’re getting the message now you’re part of a target demo someone wants to turn out.

jtorsella

4 年前議論

Yes, that’s a good point. But it seems to me like that’s a matter of degree, right? Like, even stipulating that the chips themselves are not fungible across products and companies, they are fungible across time for the same or similar products. The price of your 2023 fridge, for example, could be lower because of excess supply of chips for 2022 models, no? And am I right in thinking that even if the chips themselves are not fungible (and there is a genuine possibility that excess production of specific chips goes completely to waste), excess productive capacity is good in and of itself despite some amount of specialization?

jtorsella

4 年前議論

I really don’t understand this. Chips are tangible goods, not services where, for example, a bunch of hospitality capacity went to waste during the pandemic. Any “extra” chips will go to use, and will only serve to reduce costs for firms and consumers purchasing them, expand the range of products it makes sense to put chips in and the speed and quality of chips companies can afford to build on. While market signals are the best way to determine these things in general, many of the reasons mentioned in the article - long spin-up times for production infrastructure, dependence on other countries with fraught geopolitical situations - are strong reasons to build reserve capacity, even setting aside the direct economic benefits.

jtorsella

4 年前議論

Done. Thanks for mentioning the paste service cli, had no idea and am definitely going to be using it!

jtorsella

4 年前議論

If anyone is interested, here are the results on my M1 Pro running Asahi Linux:

Min: 48.3 Max: 175.0 Mean: 133.0

I’ll try to copy the exact results once I have a browser on Asahi, but the general pattern is most pairs have >150ns and a few (0-1; 2-3,4,5; 3-4,5; 4-5; 6-7,8,9; 7-8,9; 8-9) are faster at about 50ns.

Edit: The results from c2clat (a little slower, but the format is nicer) are below.

CPU 0 1 2 3 4 5 6 7 8 9

   0    0   59  231  205  206  206  208  219  210  210

   1   59    0  205  215  207  207  209  209  210  210

   2  231  205    0   40   42   43  180  222  224  213

   3  205  215   40    0   43   43  212  222  213  213

   4  206  207   42   43    0   44  182  227  217  217

   5  206  207   43   43   44    0  215  215  217  217

   6  208  209  180  212  182  215    0   40   43   45

   7  219  209  222  222  227  215   40    0   43   43

   8  210  210  224  213  217  217   43   43    0   44

   9  210  210  213  213  217  217   45   43   44    0

jtorsella

4 年前議論

I'm surprised nobody has raised concerns about the authors. This group is notorious for producing near-fraudulent preprints in support of their preconceptions and widely publicizing them. Tracy Hoeg was responsible for an atrocious VAERS-based study making false claims about vaccine safety, and Neeraj Sood was intimately involved in the Hoover Institution's successful efforts in March/April 2020 to persuade certain media figures that Covid was less deadly than the flu. You can read more about that VAERS preprint here: https://sciencebasedmedicine.org/peer-review-of-a-vaers-dump... and some related recent work around child vaccines here: https://sciencebasedmedicine.org/an-impossible-trial/. I believe the third author is a professional covid skeptic associated with the organization "rational ground". This is not a reliable set of sources. I will have time to dig in later, but if this is anything like the rest of their work it's worthless garbage.

投稿

Show HN: HacheQL, a lightweight library to make GraphQL requests HTTP-cacheable

コメント