HNHacker News
TopNewBestAskShowJobs

criticaltinker

1,200 karma · joined July 23, 2021

submissionscomments
criticaltinker··on The potential of transformers in reinforcement learning
The original paper that introduced the Transformer architecture is quite accessible and outlines a lot of the history and rational for the design [1].

[1] https://arxiv.org/pdf/1706.03762.pdf

criticaltinker··on Implementing RSA in Python from Scratch
Relevant excerpt from stackexchange answer [1]:

Some (Undesirable) Properties of Textbook RSA:

It is malleable. I.e., if you give me a ciphertext 𝑐 which encrypts 𝑚, I can compute 𝑐′≡𝑐⋅2𝑒mod𝑛. When the owner of the private key decrypts 𝑐′, she will get 2𝑚mod𝑛. In other words, I can make predictable changes to ciphertexts.

It is deterministic, and thus not semantically secure. I.e., I can distinguish between the encryptions of 0 and 1 (simply by encrypting both values myself and comparing the ciphertexts).

Differences with Deployed RSA:

Padding

Chinese Remainder Theorem is sometimes used in deployed systems for more efficient decryption.

𝑒 is often statically set to 65537=2^16+1 for encryption speed (since there are only two set bits in that number).

Side-channel attack mitigations can be put in place for deployed systems too.

[1] https://crypto.stackexchange.com/a/1449

criticaltinker··on Implementing RSA in Python from Scratch
Things like constant-time arithmetic are essential for safe and correct cryptography [1].

Treatments like the OP are accessible and can be helpful for beginners to understand RSA and public key crypto at a high level. However, the tradeoff is that critical details like side-channels are often neglected or not mentioned at all.

[1] https://eprint.iacr.org/2021/1121.pdf

criticaltinker··on What scientists must know about hardware to write fast code (2020)
If you’re working with time series data and using a vectorized algorithm implemented with SIMD instructions you can achieve decompression speeds approaching 15 GB/s [1].

Anecdotally when working with biological signals at my day job, compression is a massive win and an absolute no brainer tradeoff when shuffling data across the network or even just storing in memory.

That said, I think your first point is still reasonable for certain types of data and compression algorithms.

[1] https://github.com/lemire/SIMDCompressionAndIntersection

criticaltinker··on How Much Does Education Improve Intelligence? A Meta-Analysis [pdf]
Yes I suspect Goodhart’s law is unavoidable to some degree.

That said, are there really any other viable ways to measure intelligence in a standardized and reproducible manner?

criticaltinker··on CPython's main branch running in the browser with WebAssembly
Yes exactly - numpy, matplotlib etc available in the browser via sensible JS interfaces could significantly impact the front-end ecosystem IMO. Exciting times and great work from the community.
criticaltinker··on Hunter S. Thompson Interview on Gonzo Journalism (1975) [video]
” That was the fatal flaw in Tim Leary’s trip. He crashed around America selling “consciousness expansion” without ever giving a thought to the grim meat-hook realities that were lying in wait for all the people who took him seriously… All those pathetically eager acid freaks who thought they could buy Peace and Understanding for three bucks a hit. But their loss and failure is ours too. What Leary took down with him was the central illusion of a whole life-style that he helped create… a generation of permanent cripples, failed seekers, who never understood the essential old-mystic fallacy of the Acid Culture: the desperate assumption that somebody… or at least some force – is tending the light at the end of the tunnel.”

Fear and Loathing in Las Vegas is one of my all time favorite movies. It’s a disorienting and psychedelic plot filled with subtle yet striking insights into human nature. I highly recommend giving it a watch if you’re looking for a taste of Hunter S Thompson’s legacy.

criticaltinker··on Jupiter’s energy crisis solved: Auroras roast upper atmosphere
There's a wide body of scientific evidence supporting the fact that electromagnetic forces have a significant impact at many scales throughout the solar system and cosmos. It has been observed, since at least the early 70s, that the Sun's electric field induces auroras, radiation belts [0], and convection in planetary magnetospheres [1][2]. The OP title might be a little sensational by calling this an energy crisis - but the photos of the heat bands were really interesting IMO.

I admire movements like the Electric Universe Theory [3] for spreading awareness of plasma cosmology, magnetohydrodynamics, etc. I think in the future there will be an ecosystem around solar space weather, [4] is an example of such.

[0] https://en.wikipedia.org/wiki/Van_Allen_radiation_belt

[1] https://agupubs.onlinelibrary.wiley.com/doi/abs/10.1029/RG00...

[2] https://ntrs.nasa.gov/api/citations/19700002717/downloads/19...

[3] https://www.electricuniverse.info/peer-reviewed-papers/

[4] https://www.swpc.noaa.gov/

criticaltinker··on Sinus flush after positive Covid test reduced risks 19 times
FYI the pre-print has been updated [1]. It now says 8 times less likely.

[1] https://www.medrxiv.org/content/10.1101/2021.08.16.21262044v...

criticaltinker··on Why doesn’t natural immunity count in the US?
A Systematic Review of the Protective Effect of Prior SARS-CoV-2 Infection on Repeat Infection https://www.medrxiv.org/content/10.1101/2021.08.27.21262741v...

Comparing SARS-CoV-2 natural immunity to vaccine-induced immunity: reinfections versus breakthrough infections https://www.medrxiv.org/content/10.1101/2021.08.24.21262415v...

Prior COVID-19 protects against reinfection, even in the absence of detectable antibodies https://www.journalofinfection.com/article/S0163-4453(21)002...

SARS-CoV-2 infection rates of antibody-positive compared with antibody-negative health-care workers in England: a large, multicentre, prospective cohort study (SIREN) https://pubmed.ncbi.nlm.nih.gov/33844963/

Risk of Reinfection After Seroconversion to Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2): A Population-based Propensity-score Matched Cohort Study https://academic.oup.com/cid/advance-article/doi/10.1093/cid...

Assessment of the Risk of Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) Reinfection in an Intense Reexposure Setting https://academic.oup.com/cid/advance-article/doi/10.1093/cid...

Assessment of SARS-CoV-2 Reinfection 1 Year After Primary Infection in a Population in Lombardy, Italy https://jamanetwork.com/journals/jamainternalmedicine/fullar...

SARS-CoV-2 antibody-positivity protects against reinfection for at least seven months with 95% efficacy https://www.sciencedirect.com/science/article/pii/S258953702...

Association of SARS-CoV-2 Seropositive Antibody Test With Risk of Future Infection https://jamanetwork.com/journals/jamainternalmedicine/fullar...

Quantifying the risk of SARS-CoV-2 reinfection over time https://pubmed.ncbi.nlm.nih.gov/34043841/

Longitudinal observation of antibody responses for 14 months after SARS-CoV-2 infection https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8325385/

Persistence of neutralizing antibodies a year after SARS-CoV-2 infection https://www.medrxiv.org/content/10.1101/2021.07.13.21260426v...

Antibody Responses 8 Months after Asymptomatic or Mild SARS-CoV-2 Infection https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7920668/

Necessity of COVID-19 vaccination in previously infected individuals https://www.medrxiv.org/content/10.1101/2021.06.01.21258176v...

criticaltinker··on Why doesn’t natural immunity count in the US?
Why is this post flagged? And why is the vouch button unavailable?
criticaltinker··on Disclosing E2EE vulnerability in multiple Matrix clients
> 100 one-time keys (OTKs) on your server

Is 100 a safe default for devices with limited memory, or are there other more salient constraints that I'm ignorant of?

criticaltinker··on Disclosing E2EE vulnerability in multiple Matrix clients
> This is not a protocol or specification bug, but an implementation bug which was then unfortunately replicated in other independent implementations.

> We will also accelerate our work on matrix-rust-sdk as a portable reference implementation of the Matrix protocol, avoiding the implicit requirement that each independent library must necessarily reimplement this logic on its own

I like the idea of providing a core Rust/C++ library that can be called in any other runtime via FFI, compiling to WebAssembly, etc. I've used the same technique to ensure consistency in usage of a custom time-series compression format across an organization. Overall this technique seems in-line with the "DRY" principle, and the tradeoffs have been favorable in my experience. I think it will be a good step forward for the Matrix ecosystem.

criticaltinker··on Unvaccinated are 5X more likely to catch Delta, 11X more likely to die
Can't be emphasized enough. Even if simplified down to three groups:

1) immunologically naive (never been exposed to SARS-CoV-2)

2) vaccinated

3) recovered from previous infection

Most discussions on HN completely miss the distinction between these groups. The truth is, so far scientific consensus indicates that groups 2 & 3 are at least equally well protected and pose equivalent risk to others.

Much of the bickering here conflates groups 1 & 3 - it's such a common mistake I've grown tired of trying to discuss it.

criticaltinker··on I Got a 'Mild' Breakthrough Case
1) That CDC study has a relatively small sample size (N=738) and uses data from a single state during a 2-month period. The confidence interval on the "2.34" odds ratio is large (95% CI = 1.58–3.47). Most importantly, a tremendous amount of literature contradicts those findings. Nearly every large scale and long term serological study has demonstrated that immunity acquired through previous infection is at least equally effective as vaccination in preventing reinfection.

- A previous history of SARS-CoV-2 infection was associated with an 84% lower risk of infection, with median protective effect observed 7 months following primary infection. This time period is the minimum probable effect because seroconversions were not included. This study shows that previous infection with SARS-CoV-2 induces effective immunity to future infections in most individuals. [1] (N=25,661)

- In conclusion, documented SARS-CoV-2 reinfections were exceedingly rare, with an incidence of 0.3 infections for every 1000 persons-week, and none were severe. Seroconversion after symptomatic or asymptomatic SARS-CoV-2 infection seems to be associated with a 10-fold reduction in risk of successive viral infection contamination, lasting at least 8 months. [2] (N=1,494)

- The study results suggest that reinfections are rare events and patients who have recovered from COVID-19 have a lower risk of reinfection. Natural immunity to SARS-CoV-2 appears to confer a protective effect for at least a year, which is similar to the protection reported in recent vaccine studies. [3] (N=15,075)

- Reinfection is rare in the young and international population of Qatar. Natural infection appears to elicit strong protection against reinfection with an efficacy ~95% for at least seven months. [4] (N=192,967)

- The degree of protection (10-fold) associated with seropositivity appears to be comparable to that observed in the initial reports of the efficacy of mRNA vaccines in large clinical trials.* [5] (N=3,257,478)

2) The preprint you cite is a "large community-based survey". I'll trade you pre-prints, this one is a retrospective observational study with data out of Israel [6]:

"This study demonstrated that natural immunity confers longer lasting and stronger protection against infection, symptomatic disease and hospitalization caused by the Delta variant of SARS-CoV-2, compared to the BNT162b2 two-dose vaccine-induced immunity. Individuals who were both previously infected with SARS-CoV-2 and given a single dose of the vaccine gained additional protection against the Delta variant."

3) Natural infection elicits antibodies that vaccination alone does not (for example nucelocapsid protein antibodies) [7]. The paper you cited does not provide evidence that vaccination on top of existing immunity confers any significant benefit to health outcomes in the face of viral mutation and variants of concern.

[1] SARS-CoV-2 infection rates of antibody-positive compared with antibody-negative health-care workers in England: a large, multicentre, prospective cohort study (SIREN) https://pubmed.ncbi.nlm.nih.gov/33844963/

[2] Risk of Reinfection After Seroconversion to Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2): A Population-based Propensity-score Matched Cohort Study https://academic.oup.com/cid/advance-article/doi/10.1093/cid...

[3] Assessment of SARS-CoV-2 Reinfection 1 Year After Primary Infection in a Population in Lombardy, Italy https://jamanetwork.com/journals/jamainternalmedicine/fullar...

[4] SARS-CoV-2 antibody-positivity protects against reinfection for at least seven months with 95% efficacy https://www.sciencedirect.com/science/article/pii/S258953702...

[5] Association of SARS-CoV-2 Seropositive Antibody Test With Risk of Future Infection https://jamanetwork.com/journals/jamainternalmedicine/fullar...

[6] Comparing SARS-CoV-2 natural immunity to vaccine-induced immunity: reinfections versus breakthrough infections https://www.medrxiv.org/content/10.1101/2021.08.24.21262415v...

[7] Immunogenicity and crossreactivity of antibodies to the nucleocapsid protein of SARS-CoV-2: utility and limitations in seroprevalence and immunity studies https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7879156/

criticaltinker··on One in Three Americans Already Had Covid-19 by the End of 2020
There is strong evidence that infections have been massively under-detected since the beginning of the pandemic, by a factor of 5x to 100x [1][2]. Recent estimates put the global average IFR closer to 0.15% [3], which is likely still over-estimated, and will likely continue to decrease as more data is collected.

[1] Evaluating the massive underreporting and undertesting of COVID-19 cases in multiple global epicenters https://www.sciencedirect.com/science/article/pii/S253104372...

[2] Undiagnosed SARS-CoV-2 seropositivity during the first 6 months of the COVID-19 pandemic in the United States https://stm.sciencemag.org/content/13/601/eabh3826

[3] Reconciling estimates of global spread and infection fatality rates of COVID- 19: An overview of systematic evaluations https://onlinelibrary.wiley.com/doi/pdfdirect/10.1111/eci.13...

criticaltinker··on Poll: Covid Vaccine Mandate in Tech
> here is no reason whatsoever to not get vaccinated except under very very very limited circumstances

Do those limited circumstances include:

1) individuals with immunity acquired through previous infection?

2) the need to prioritize other individuals who have not been vaccinated once already?

3) vaccine resistance - i.e. in populations with dominant variants that to continue to evolve substantial resistance to current formulations?

4) vulnerable sub-populations (ie children & pregnant women) that have not yet participated in any long-term or large-scale clinical trials?

criticaltinker··on [dead]
> Coronavirus was and is a real crisis. They do sometimes happen

Yes and crises precipitate change. It's in everyone's best interest to not let that change be an erosion of hard-won human rights and civil liberties. What kind of crises would it take for us to give those up? It's an important question, and the diversity of everyone's opinion is a crucial feature.

criticaltinker··on I Got a 'Mild' Breakthrough Case
> everyone needs to take the vaccine

This is simply not true, anyone who claims this is ignoring a vast body of scientific evidence supporting the fact that individuals who’ve recovered from previous infection will be at least as well protected as vaccinated individuals against reinfection.

If you were to say “most people who don’t already have natural immunity ought to be vaccinated” that would be a much more sensible recommendation, in-line with general scientific consensus (and public health policies in other countries).

criticaltinker··on I Got a 'Mild' Breakthrough Case
If you think you're getting sick, the number one action you should take is to begin treatment early. You need to take action before symptoms get worse. The author had a high grade fever for a week but never mentioned seeking any treatment.

Things you should do yourself: take vitamins and nutritional supplements, drink plenty of water, get your body moving, isolate yourself from others, wear a mask.

Things your doctor should do for you: prescribe antivirals, also corticosteriods & antithrombotics if symptoms continue to worsen. The goal is to avoid the hospital and severe outcomes - many clinics will prescribe these safe & widely available over the counter medicines with little to no consultation.

Whether you're vaccinated or not, this should be the standard of care for everyone - it's no different than most other viral infections. For some reason very few people seem to be aware of this advice, and all the scientific evidence supporting it (in the context of SARS-CoV-2 and in general).

criticaltinker··on Boys more at risk from Pfizer jab side-effect than from Covid, suggests study
> this "study" was searching for myocarditis symptoms in the VAERS database and extrapolating from there

It's fair to be critical of this study. However it's important to be aware that the CDC is conducting their own investigation of myocarditis and myopericarditis associated with vaccination. The CDC reviews VAERS reports and then actively investigates them, keeping a tally of the cases that meet the CDC's own stringent criteria.

According to the CDC (as of August 18th), for males age 16-17 the reporting rate of myopericarditis after 2 doses of Pfizer is 71.5 per 1 million doses administered (0.0071%). For males aged 18-24 the reporting rate after 2 doses of Pfizer or Moderna is ~37 per 1 million doses (0.0037%) [1]. They note that for males the observed cases exceed the expected cases by a significant amount in age groups through 49 years.

> myocarditis occurs at a rate of about 200/million in general

For accurate scientific discussion it's important to stratify by age. Following the reference chain from your citation, the important nuance is that "more than one-half of all cases [of myocarditis in the pediatric population] are seen in the first year of life" [2]. Consequently the incident rate you've cited isn't quite as representative as you suggest. This fact is in alignment with the CDC mentioning that the observed cases of vaccine associated myocarditis are greater than expected, particularly for male teenagers.

> It's frankly irresponsible journalism to have a headline like this

I disagree - the title accurately reflects available evidence and official data from government sources.

> that's going to cause people to run scared

People shouldn't be scared by this data. The vast majority of young adults recover quickly from clinical and subclinical myocarditis [1], although the long-term effects aren't well established. It's important to note that viral infection can also cause myocarditis, so this phenomenon isn't just a knock against vaccination. The takeaway is that these findings are an important factor to consider when considering vaccination strategies for children and teenagers - we should proceed with an abundance of caution, but not fear.

[1] https://www.cdc.gov/vaccines/acip/meetings/downloads/slides-...

[2] The Diagnostic and Clinical Approach to Pediatric Myocarditis: A Review of the Current Literature https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6352488/

criticaltinker··on Adults with history of recreational cannabis use have altered speech production
Yes agreed - the title doesn't seem to accurately reflect the details in the abstract.

Beyond a larger sample size, IMO it would have been also been useful to take repeat measurements of the cohort over time and also under the influence, especially given the observation of "change in line with duration of abstinence from cannabis use".

Perhaps those details are in the full paper, I haven't been able to find a copy.

criticaltinker··on FBI releases newly declassified record on Sept. 11 attacks
Given a heavily redacted government document, how can a member of the general public verify its authenticity?

I’m not necessarily questioning the document alluded to in OP, or the conclusions it suggests. Just wondering if there’s some existing crypto signature, watermark, data hash, etc that could be relied upon to independently verify claims in any such government document.

In the age of disinformation and advanced three letter agencies it seems increasingly difficult to establish provenance and trust. I haven’t found any links to the declassified document despite this story being published verbatim by every major news outlet.

criticaltinker··on L-arginine in severe Covid-19 decreases the length of hospitalization
For others who are curious, N=101 in the OP paper, which is a preliminary subset of the total subjects in the clinical trial (N=290, according to the registration).

To your second point, arginine supplementation can still be helpful for otherwise healthy individuals who are undergoing physiological stress due to illness. FWIW there are some supporting references on the Wikipedia page.

criticaltinker··on L-arginine in severe Covid-19 decreases the length of hospitalization
This is yet another piece of compelling evidence showing that nutraceuticals can significantly reduce recovery time and improve health outcomes.

Combined with early treatment using widely available medicines, it’s clear we have a variety of tools at our disposal to mitigate this pandemic.

Unfortunately many people are unaware of the evidence supporting these types of treatments, partly because they receive such little attention in mass media.

As usual I’m ready for the downvotes - for some reason people here think this information is contentious or disagreeable.

criticaltinker··on Windy.com
Also worth mentioning that Windy provides several different forecast models that you can choose between. There are high resolution models like NAM, and lower resolution models like GFS - toggling between them often gives me a better sense of what to expect.
criticaltinker··on USA: All government employees will be vaccinated against Covid-19
I agree, why is this conspicuously missing in rhetoric from the US government? Other countries recognize that immunity acquired through previous infection is at least as protective as vaccination. This is well established in the scientific literature and should not be a controversial point.
criticaltinker··on OSHA to require employers with 100 employees vaccinate or test workforce
It's pretty egregious that there is no mention of accommodation for individuals who have already acquired effective immunity through previous infection. Other countries are well ahead of the U.S in this regard.

These mandates appear to completely neglect entire aspects of scientific consensus.

criticaltinker··on New details on silicon, subatomic particles and possible ‘fifth force’
It seems like the title length constraint on HN is a bit too limiting and often results in these types of clarifying comments.

Couldn’t we expect more accurate and higher quality titles by relaxing the length constraint? I’m sure it’s been discussed before here, but I’m struggling to think of downsides from such a change.

criticaltinker··on Absorption of 5G Radiation in Brain Tissue
This is a good point that was acknowledged in the study and by other commenters here. However don’t you think comparing with the sun is a slippery analogy? Specifically when discussing the health effects of duration of exposure, beam shape, pulse pattern, etc?
← PreviousPage 3 of 15Next →