arXiv.org e-Print archive

Mesurer les déformations de l’espace-temps avec la déviation de la lumière - YouTube
▻https://www.youtube.com/watch?v=lubOV27-8OI

Comment est-il possible que l’on voit plusieurs fois d’une même étoile ? Pourquoi est-ce que les galaxies nous apparaissent parfois plus allongées qu’elles ne le sont en réalité ? Dans cette vidéo, on vous explique comment le lentillage gravitationnel fonctionne. Vous allez comprendre pourquoi la première observation du lentillage par Eddington a rendu Einstein célèbre, en lui donnant raison par rapport à Newton. Et puis surtout on va vous présenter nos tous nouveaux résultats, où l’on utilise le lentillage pour mesurer la distorsion de l’espace et du temps au cours de l’histoire de l’Univers. Et on a eu quelques surprises !

L’article scientifique qui décrit cette recherche peut être téléchargé à ▻https://arxiv.org/abs/2312.06434

Musique :
Walking In The Sky - Nico Staf.
Gravity by extenz, Creative Commons License.
All Music from RFM - NCM : www.bit.ly/2xGHypM

A Survey of Large Language Models (31/03/2023)
▻https://arxiv.org/pdf/2303.18223.pdf

https://ssphub.netlify.app/infolettre/infolettre_13/featured.jpg

Abstract […] Considering this rapid technical progress, in this survey, we review the recent advances of LLMs by introducing the background, key findings, and mainstream techniques. In particular, we focus on four major aspects of LLMs, namely pre-training, adaptation tuning, utilization, and capacity evaluation. Besides, we also summarize the available resources for developing LLMs and discuss the remaining issues for future directions. This survey provides an up-to-date review of the literature on LLMs, which can be a useful resource for both researchers and engineers.

via l’incontournable
Infolettre Avril 2023, première quinzaine (#13) | SSPHub
▻https://ssphub.netlify.app/infolettre/infolettre_13

Pourquoi détecter un texte de ChatGPT est-il si dur ? - Numerama
▻https://www.numerama.com/tech/1252886-pourquoi-detecter-un-texte-ecrit-par-chatgpt-est-il-si-dur.html

[…]
Des outils qui ne marchent pas encore complètement
Ces aides sont encore loin d’être convaincantes. « GPTZero marche bien en anglais, mais pas trop en français », nous indique Giada Pistilli, éthicienne pour l’entreprise spécialisée dans l’apprentissage automatique et l’IA Hugging Face. La même conclusion s’applique pour les textes écrits par IA, mais modifiés à postériori par un humain. « Il ne détecte pas tout, ce n’est pas un outil toujours performant. D’ailleurs, GPTZero dit que ses résultats sont probables, pas qu’ils sont sûrs ».

Même chose pour les deux autres outils : les résultats sont plutôt bons en anglais, mais manquent de précision pour les autres langues. DetectGPT n’a pas reconnu notre article écrit par ChatGPT, et AI Text Classifier a, à l’inverse, cru qu’un texte écrit par mes soins avait « probablement » été écrit par une intelligence artificielle.

Pourtant, ils utilisent tous une méthode de détection différente : GPTZero analyserait la « perplexité » d’un texte (la variable de « hasard » détectée dans un texte), DetectGPT utilise la « courbure locale de la fonction de probabilité logarithmique du modèle », et AI Text Classifier utilise une intelligence artificielle, elle-même entraînée pour déterminer si un texte était écrit par une IA. Peu importe la méthodologie, il n’existe donc pas pour l’instant de solution infaillible.

Le développement d’un watermark pour les IA
Une autre piste est actuellement étudiée pour permettre une authentification fiable dans le futur : l’implantation d’un « watermark » dans les textes. Ce filigrane fonctionnerait en « insérant des tokens de manière aléatoire dans les textes écrits par IA », souligne Giada Pistilli.

Excess death rates for Republicans and Democrats during the COVID-19 pandemic – comment by Paul Goldsmith-Pinkham @paulgp – Twitter
▻https://threadreaderapp.com/thread/1576899935147991041.html

le tweet initial
▻https://twitter.com/paulgp/status/1576899935147991041

New NBER working paper with @jwswallace and @jasonlschwartz on Covid mortality: “Excess death rates for Republicans and Democrats during the COVID-19 pandemic” nber.org/papers/w30512
Ungated on arxiv here: arxiv.org/abs/2209.10751

🧵1/
Political affiliation has emerged as a potential risk factor for COVID-19, amid evidence that Republican-leaning counties have had higher COVID-19 death rates than Democrat-leaning counties and evidence of a link between political party affiliation and vaccination views.
▻https://arxiv.org/abs/2209.10751

original paper
▻https://www.nber.org/papers/w30512

2/ A popular commentary on the Covid crisis has been how much higher the Covid death toll has been in Republican vs. Democratic counties in the U.S.

U.S. Covid Deaths Get Even Redder
The partisan gap in Covid’s death toll has grown faster over the past month than at any previous point.
▻https://www.nytimes.com/2021/11/08/briefing/covid-death-toll-red-america.html

https://pbs.twimg.com/media/FeI_V1YXgAA4P1V.jpg

3/ There are natural reasons to believe that there are strong differences by Republicans vs. Democrats: survey data suggests that there are big differences by party ID on Covid-19 vaccination:

KFF COVID-19 Vaccine Monitor Dashboard
Using a combination of surveys and focus groups, the KFF COVID-19 Vaccine Monitor will track the dynamic nature of public reactions as vaccine development unfolds, including vaccine confidence and ÷

▻https://www.kff.org/coronavirus-covid-19/dashboard/kff-covid-19-vaccine-monitor-dashboard

https://pbs.twimg.com/media/FeJBf1xXoAIScnA.jpg

4/ The challenge, of course, is whether it’s really about Republicans vs. Democrats living in these areas, or just the areas where individuals sort into are different.

5/ This statistical analysis runs into a serious challenge, driven by the fact that publicly available data on Covid deaths, and measures of political party, are typically only available at the county level.

6/ The focus on Covid deaths and counties has lead researchers to try to account for these locational differences (by controlling for features at the county level), but are still limited by the aggregated nature of the data:

▻https://www.healthaffairs.org/doi/full/10.1377/hlthaff.2022.00085
The Association Between COVID-19 Mortality And The County-Level Partisan Divide In The United States | Health Affairs Journal
Partisan differences in attitudes toward the COVID-19 pandemic and toward the appropriateness of local policies requiring masks, social distancing, and vaccines are apparent in the United States.

7/ The other issue with this approach is that it focuses on reported Covid deaths as an aggregate measure. This measure may not fully capture the “counterfactual” deaths in the absence of the pandemic. Our world in data does an excellent discussion:

►https://ourworldindata.org/excess-mortality-covid#excess-mortality-during-covid-19-background
Excess mortality during the Coronavirus pandemic (COVID-19)
Excess mortality is a term used in epidemiology and public health that refers to the number of deaths from all causes during a crisis above and beyond what we would have expected to see under ‘normal’…
►https://ourworldindata.org/excess-mortality-covid#excess-mortality-during-covid-19-background

8/ Intuitively, calculating excess death rates requires a prediction of death rates in 2020 and 2021 based on previous years for the group of interest: namely Democrats and Republicans. Fortunately, we have mortality data with party affiliation, age, and location in this paper!

9/ We construct data using individual-level voter registration in 2017, linked to death records from 2018 to 2021, for Ohio and Florida. We then construct excess death rates that control for differences in mortality rates (pre-Covid) at the age-by-party-by-county-by-month level

10/
This lets us ask and answer three questions:

11/ Q1: Does excess death in 2020 and 2021 differ by political party, how much and when does this occur?

A1: Yes, the excess death rate for Republicans was 5.4 p.p., or 76%, higher than for Democrats. The gap was exclusively in the post-vaccine period (10.4 pp or 153%).

https://pbs.twimg.com/media/FeJEXMUX0AAf1XB.jpg

12/ Q2: Is this difference explained by geographic or age differences in political party affiliation?

A tiny share of the difference is explained by differential impacts of age-by-county during Covid (recall that excess deaths already controls for pre-Covid differences):

https://pbs.twimg.com/media/FeJEgiqXoAA6Wlm.png

13/ Q3: How much can we point to vaccines?
A3: This is harder, since we don’t have individual-level data on vaccines. However, two facts emerge:

A. The association between the Rep.-Dem. gap and county-level vaccination rates grows significantly after they become available:

https://pbs.twimg.com/media/FeJE3LzXkAA7qDI.png

14/
B. Moreover, pre-vaccine, the relationship across counties between realized vax rates and excess deaths was identical for both groups.

Post-vaccine, the Democrat rate fell and Republican rate climbed; and the gap between the two was near zero in high-vax counties.

https://pbs.twimg.com/media/FeJFZ0CWYAY2TnM.jpg

16/ If this is really a story about vaccines, the continued story of low take-up of vaccines + boosters among Republicans may perpetuate some of these differences:

KFF COVID-19 Vaccine Monitor: September 2022
Our latest Vaccine Monitor survey finds that half of the public has heard either “a lot” or “some” about the newly-available bivalent COVID-19 boosters, and a third (32%) say they’ve already gotten…
▻https://www.kff.org/coronavirus-covid-19/poll-finding/kff-covid-19-vaccine-monitor-september-2022

https://pbs.twimg.com/media/FeJFfuWXwAAfQPI.png https://pbs.twimg.com/media/FeJFmh6XkAAegr3.jpg

17/ We’re working on expanding this out now to contrast our results with the existing literature a bit and highlight some more points, but would welcome any comments or suggestions.

fin/ It is important to reiterate that our results hold fixed differences in mortality by age, location, and party pre-Covid, and can account for location-by-age differences post-Covid. Hence these are within-age-and-location differences in mortality outcomes by political party.

For a decade+, grammar-of-graphics approaches have been a leading way to make visualizations – Jeffrey Heer @jeffrey_heer – Twitter
▻https://twitter.com/jeffrey_heer/status/1570084226153979905
Thread by @jeffrey_heer on Thread Reader App
▻https://threadreaderapp.com/thread/1570084226153979905.html

For a decade+, grammar-of-graphics approaches (ggplot, Tableau, #d3js, Vega/Altair) have been a leading way to make visualizations. Beyond chart templates & low-level programming, are there compelling alternatives? Or does the future lie in abstractions on top of these grammars?
There’s exciting research work on new/extended grammars, including:
– probability expressions 📊 (▻https://www.mjskay.com/papers/chi2020-pgog.pdf )
– responsive charts 📲 (▻https://twitter.com/hyeok__kim/status/1506395769615568904 )
– animation 🕺 (▻http://idl.cs.washington.edu/papers/gemini2 , gganimate, and new @vega_vis work coming to VIS’22 from @mitvis)

@mcnutt has written a valuable survey of JSON-style grammars (▻https://arxiv.org/pdf/2207.07998.pdf ), concluding there is “No Grammar to Rule Them All”. Should we expect a proliferating multiverse of visualization grammar variants? Where might we look beyond (or building on) such grammars?
I’m increasingly interested in higher-level support for analysis tasks. Grammars are a powerful way for (both people and machines!) to say “make this chart”. But most grammars are agnostic to the larger task and intent. Could we (should we?) bridge that gap?
(And for any friends wondering: Yes, I should be working on #chi2023 papers right now. And no, I’m not writing a CHI paper on this topic... #procrastination 😅)

p.s. thinking about this also reminds me how much I miss Lee Wilkinson (RIP). He was always bubbling with ideas on this topic... before he passed we had been discussing multimodal grammars, including sonification, tactile displays, etc. Important to also think beyond the visual.

C’est un gag où ils ont intriqué un tardigrade avec des qubits ?
▻https://arxiv.org/pdf/2112.07978v1

[Edit] Oh des jolies photos et un article en français sont arrivés depuis : ▻https://trustmyscience.com/tardigrade-premier-organisme-multicellulaire-subir-intrication-quant

Pour déterminer si l’état d’intrication avait été atteint entre le tardigrade et le qubit, les chercheurs ont mesuré la fréquence à laquelle la combinaison tardigrade-qubit vibrait. Résultat : les calculs (basés sur les mesures) n’avaient de sens que si les deux objets étaient considérés en état d’intrication quantique.

Après avoir effectué leurs mesures, les chercheurs ont lentement dépressurisé et réchauffé le tardigrade, le faisant sortir de son état de relaxation et le ramenant à la vie. Mais les records établis par ce super tardigrade ne s’arrêtent pas là : la température en question (qui était à peine 0,01 °C au-dessus du zéro absolu) est la plus basse à laquelle un tardigrade ait jamais survécu.

#sciencefiction #surrealisme #quantique #vivisection #recherche #ilssontfous #schroedinger #chatnoirchatblanc

How to poison the data that Big Tech uses to surveil you
▻https://www.technologyreview.com/2021/03/05/1020376/resist-big-tech-surveillance-data

Algorithms are meaningless without good data. The public can exploit that to demand change. Every day, your life leaves a trail of digital breadcrumbs that tech giants use to track you. You send an email, order some food, stream a show. They get back valuable packets of data to build up their understanding of your preferences. That data is fed into machine-learning algorithms to target you with ads and recommendations. Google cashes your data in for over $120 billion a year of ad revenue. (...)

#Google #algorithme #activisme #[fr]Règlement_Général_sur_la_Protection_des_Données_(RGPD)[en]General_Data_Protection_Regulation_(GDPR)[nl]General_Data_Protection_Regulation_(GDPR) #BigData (...)

##[fr]Règlement_Général_sur_la_Protection_des_Données__RGPD_[en]General_Data_Protection_Regulation__GDPR_[nl]General_Data_Protection_Regulation__GDPR_ ##microtargeting

https://wp.technologyreview.com/wp-content/uploads/2021/03/AP_18312600261563-e1614960773408.jpg

De la carte au GPS, comment l’image de la route oriente votre vision du monde
▻https://www.franceculture.fr/histoire/de-la-carte-au-gps-comment-limage-de-la-route-oriente-votre-vision-du-

Histoires d’images | Entrez votre adresse de départ, entrez votre adresse d’arrivée, temps estimé : 26 minutes, tournez à droite, tournez à gauche, vous êtes arrivés. Qu’avez-vous vu ? Si vous êtes incapables ou presque de décrire l’espace que vous avez traversé, ne craignez rien : vous pouvez blâmer votre GPS. Vous allez quelque part ? Si vous êtes en voiture, vous avez probablement déjà paramétré votre GPS. Si vous êtes en ville, une application qui a recours à la géolocalisation peut vous aider à (...)

#cartographie #géolocalisation

https://cdn.radiofrance.fr/s3/cruiser-production/2021/03/e484dc0a-e8a3-4ec5-a7bf-a9e203731381/838_gettyimages-118639231.jpg

Can This New Map Fix Our Distorted Views of the World?

Most of the world maps you’ve seen in your life are past their prime. The Mercator was devised by a Flemish cartographer in 1569. The Winkel Tripel, the map style favored by National Geographic, dates to 1921. And the Dymaxion map, hyped by the architect Buckminster Fuller, debuted in a 1943 issue of Life.

Enter a brash new world map vying for global domination. Like sports, the mapmaking game can sometimes grow stale when top competitors are stuck on the same old strategy, said J. Richard Gott, an astrophysicist at Princeton who had previously mapped the entire universe. But then along comes an innovator: Think Stephen Curry of the Golden State Warriors, splashing 3-pointers from areas of the court the rest of basketball hadn’t thought were worth guarding.

“We were sort of reaching the limit of what you could do,” Dr. Gott said. “If you wanted any significant breakthrough, you had to use a new idea.”

Dr. Gott’s version of Steph Curry’s wait-you-could-shoot-from-there 3? Use the back of the page, too. Make the world map a double-sided circle, like a vinyl record. You could put the Northern Hemisphere on the top side, and the Southern Hemisphere on the bottom, or vice versa. Or to put it differently: You could deflate the 3-D Earth into two dimensions. And if you did, you could blow the accuracy of previous maps out of the water.

https://i.imgur.com/suDlSJE.png

No flat map of our round world can be perfect, of course. First you need to peel off Earth’s skin, then pin it down. This mathematical taxidermy introduces distortions. If you have a Mercator projection on your classroom walls, for example, you may grow up thinking Greenland is the size of Africa (not even close) or Alaska looms larger than Mexico (also nope). This warped worldview might even bias you, subconsciously, to under-appraise most of the developing world.

Shapes also change in map projections. Distances vary. Straight lines curve. Some projections, such as Mercator, aim to excel at one of these concerns, which aggravates other errors. Other maps compromise, like the Winkel Tripel, so named because it tries to strike a balance between three kinds of distortion.

Starting in 2006, Dr. Gott and David Goldberg, a cosmologist at Drexel University in Philadelphia, developed a scoring system that could sum up these different kinds of error. The Winkel Tripel beat out other major contenders. But one big source of distortion persisted: a mathematical incision, often running from pole to pole down the Pacific. The resulting shape can never again be stretched and pulled back into the unbroken surface of a sphere. “This does violence to the globe,” Dr. Gott said.

His new kind of double-sided map, crafted with Dr. Goldberg and Robert Vanderbei, a mathematician at Princeton, skips the topological violence entirely. The map simply continues over the edge. You could stretch a string over the side; an ant could walk there. Without any cut, the map’s Goldberg-Gott distortion score blows all other maps currently in use out of the water, the team reports in a draft study.

Cartographers who regularly study world maps — perhaps fewer than 10 people — will now have time to react. “It never came up to me that it could be done in this way,” said Krisztián Kerkovits, a Hungarian cartographer working to develop his own projections.

https://i.imgur.com/lhS61sg.png

But while the new map excels at addressing distortion, Dr. Kerkovits said it also introduced a new weakness. You can see only half of the planet at once, unlike the Winkel Tripel and Mercator. That undermines the basic premise of flaying out the whole world for inspection on a single page or screen.

To Dr. Gott, this is no different than the 3-D globe itself. But Dr. Kerkovits isn’t quite sure: After all, you can always rotate a globe slightly to see the neighbors of any chosen point. But in the double-sided map, you might have to flip the entire thing.

Ultimately a map’s success depends on which applications it’s used for, and how its popularity grows over time. Dr. Gott, whose paper also presents double-sided projections of Jupiter and other worlds, envisions the new map style as a physical object to turn over in your hands.

You could cut one out of a magazine, or you could store a whole stack of them in a thin sleeve, showing different planets or different data layers. And he hopes you may be tempted to try to print out and make your own using the appendix of his paper.

“Glue it back to back with double-stick tape — I think that’s better than Elmer’s Glue, but you can use glue,” Dr. Gott said. Then cut it out. “Maybe use card stock paper,” he added.

►https://www.nytimes.com/2021/02/24/science/new-world-map.html

#projection #visualisation #monde #cartographie #distorsion
via @reka

«Algebraic Property Graphs» par Joshua Shinavier et Ryan Wisnesky, 13.10.2020
▻https://arxiv.org/abs/1909.04881

« One of the main challenges of building an enterprise knowledge graph is guaranteeing interoperability between a single graph data model and the diverse and often changeable ecosystem of non-graph data models, languages, and tools which surround and support the graph. The simple and developer-friendly property graph family of data models lends itself to this task, yet the lack of a formal specification deprives the graph of well-defined semantics. In this paper, we observe that algebraic data types are a common foundation of most of the enterprise schema languages we deal with in practice, and are also a suitable basis for a property graph formalism. We introduce this formalism in terms of type theory, algebra, and category theory, also providing algorithms for query processing and data migration with guarantees of semantic consistency across supported languages and datasets. These results have clear connections to relational database theory, programming language theory, and graph theory, providing starting points for significant future work. Open research challenges described in the paper include adding constraints, query and schema languages, and logics on top of the basic type system, interfacing with specific graph and non-graph data models, and performing operations which are typically difficult or ill-defined on property graphs, such as graph merges. »

#enterprise #knowledge #graph #addressability

School closures cut Covid-19 infections, says study

The closure of Swiss schools in the spring of 2020 was one of the most effective measures to reduce travel and so the transmission of Covid-19, says a new study.

Researchers at the Federal Institute of Technology ETH Zurich found that the closure of schools reduced mobility by around a fifth, report Sunday newspapers Le Matin Dimanche and Sonntagszeitung.

“The closure of the schools reduced mobility by 21.6%,” ETH professor of computer science and management Stefan Feuerriegel, who led the study, told French news agency AFP.

The Swiss research team analysed 1.5 billion movements between February 10 and April 26, 2020, using data from Swiss mobile operators to assess the impact of measures against Covid-19 on mobility. All changes in postcode areas were taken into account.

According to the study, the ban on gatherings of more than five people reduced mobility by 24.9%, while the closure of bars, restaurants and other non-essential shops brought a drop of 22.3%, while school closures came just behind.

“If the schools are closed, we can hope for a big change in behaviour,” says Feuerriegel. “Not only do children stay at home, but sometimes it also means a change for the parents.”

Le Matin Dimanche says the study is likely to rekindle controversy over whether schools should be closed to combat the spread of Covid-19. While governments in Switzerland and elsewhere closed schools during the first wave of the pandemic, they have since sought to keep them open if possible.

Studies have suggested that school closures are harmful to the education particularly of younger children and increase inequalities between them. And while children are unlikely to develop serious forms of Covid-19, their role in spreading it remains unclear.

▻https://www.swissinfo.ch/eng/school-closures-cut-covid-19-infections--says-study/46275244
#écoles #école #contamination #fermeture #covid-19 #coronavirus #pandémie #enfants #ETHZ

ping @simplicissimus

Towards decolonising computational sciences - Abeba Birhane, Olivia Guest
▻https://arxiv.org/abs/2009.14258v1

https://files-rghsjnolk.vercel.app/abeba.png

This article sets out our perspective on how to begin the journey of decolonising computational fields, such as data and cognitive sciences.
We see this struggle as requiring two basic steps:
a) realisation that the present-day system has inherited, and still enacts, hostile, conservative, and oppressive behaviours and principles towards women of colour (WoC); and
b) rejection of the idea that centering individual people is a solution to system-level problems.

The longer we ignore these two steps, the more “our” academic system maintains its toxic structure, excludes, and harms Black women and other minoritised groups. This also keeps the door open to discredited pseudoscience, like eugenics and physiognomy. We propose that grappling with our fields’ histories and heritage holds the key to avoiding mistakes of the past.

For example, initiatives such as “diversity boards” can still be harmful because they superficially appear reformatory but nonetheless center whiteness and maintain the status quo. Building on the shoulders of many WoC’s work, who have been paving the way, we hope to advance the dialogue required to build both a grass-roots and a top-down re-imagining of computational sciences — including but not limited to psychology, neuroscience, cognitive science, computer science, data science, statistics, machine learning, and artificial intelligence.

We aspire for these fields to progress away from their stagnant, sexist, and racist shared past into carving and maintaining an ecosystem where both a diverse demographics of researchers and scientific ideas that critically challenge the status quo are welcomed.

Let’s BUILD a COMPUTER in CONWAY’s GAME of LIFE ⠠⠵
▻https://www.youtube.com/watch?v=Kk2MH9O4pXY

Possibility of Disinfection of SARS-CoV-2 (COVID-19) in Human Respiratory Tract by Controlled Ethanol Vapor Inhalation
Tsumoru Shintake, arXiv:2003.12444 [physics.med-ph], le 15 mars 2020
▻https://arxiv.org/abs/2003.12444

The author suggests that it may be possible to use alcoholic beverages of 16~20 v/v% concentration for this disinfection process, such as Whisky (1:1 hot water dilution) or Japanese Sake, because they are readily available and safe (non-toxic). By inhaling the alcohol vapor at 50~60∘C (122~140∘F) through the nose for one or two minutes, it will condense on surfaces inside the respiratory tract; mainly in the nasal cavity.

#coronavirus #science #alcohol #whisky #saké #humour ?

« Assessing the Privacy Benefits of Domain Name Encryption » une étude de la #viePrivée avec les techniques récents de défense du #DNS et de #TLS.

L’article montre que, même avec du SNI chiffré <▻https://www.bortzmeyer.org/8744.html>, et avec #DoT/#DoH, l’adresse IP du serveur est souvent un bon moyen de retrouver le domaine car beaucoup de serveurs n’hébergent qu’un petit nombre de domaines. Si l’attaquant maintient une base noms<->adresses, il peut facilement retrouver le nom.

▻https://arxiv.org/abs/1911.00563

Bonne analyse, bien fouillée et avec plein de détails quantitatifs.

Parmi les suggestions des auteurs, demander aux hébergeurs de changer souvent les adresses IP allouées aux clients (pour casser les bases mentionnées plus haut). D’après DNSDB, seenthis.net n’a pas changé d’adresse IP (217.182.178.243) depuis plus de trois ans... Et 217.182.178.243 n’héberge que SeenThis.

The Future of Work Is Here
▻https://arxiv.org/pdf/2007.05843.pdf

Abstract This commentary traces contemporary discourses on the relationship betweenartificial intelligence and labour and explains why these principles must becomprehensive in their approach to labour and AI. First, thecommentaryasserts that ethical frameworks in AI alone are not enough toguarantee work-ers’ rights since they lack enforcement mechanisms and the representation ofdifferent stakeholders. Secondly, it argues that current discussions on AI andlabour focus on the deployment of (...)

#algorithme #technologisme #travail

The #Sci-hub Effect: Sci-hub downloads lead to more article citations
▻https://arxiv.org/abs/2006.14979v2

Citations are often used as a metric of the impact of scientific publications. Here, we examine how the number of downloads from Sci-hub as well as various characteristics of publications and their authors predicts future citations. Using data from 12 leading journals in economics, consumer research, neuroscience, and multidisciplinary research, we found that articles downloaded from Sci-hub were cited 1.72 times more than papers not downloaded from Sci-hub and that the number of downloads from Sci-hub was a robust predictor of future citations. Among other characteristics of publications, the number of figures in a manuscript consistently predicts its future citations. The results suggest that limited access to publications may limit some scientific research from achieving its full impact.

Des données chinoises trafiquées ?
Marie-Claude Bourdon, Actualités UQAM, le 19 Mai 2020
▻https://www.actualites.uqam.ca/2020/covid-19-donnees-chinoises-trafiquees

https://www.actualites.uqam.ca/sites/default/files/styles/grande/public/benford.graphique_v2_w.jpg?itok=-nwoK9W2&c=15612f7d7fed09dac2bb0a70612bbfd9

Plus intéressant qu’il n’y parait :

Prenez n’importe quelle série de nombres, par exemple le prix des produits dans votre épicerie, le nombre d’habitants des villes canadiennes, le nombre de votes obtenus par différents candidats dans différentes circonscriptions, et regardez les premiers chiffres de ces nombres (1 pour 18, 2 pour 26 000, 3 pour 3372, etc.). Il y a de fortes chances que le chiffre 1 soit le plus fréquent, suivi du 2, et ainsi de suite. Il s’agit d’un phénomène contre-intuitif puisque l’on s’attendrait à observer autant de 1 que de 9, par exemple. Ce phénomène, bien connu des statisticiens, peut être expliqué par la loi de Benford. Cette loi mathématique de probabilité permet d’examiner des séries de nombres pour déceler des anomalies et ainsi détecter de possibles fraudes et autres falsifications. Le professeur du Département de mathématiques Jean-François Cœurjolly l’a testée avec les données sur la pandémie de COVID-19 fournies par la Chine, le Canada, les États-Unis et la France.

Dans un article publié en 1972, l’économiste Hal Varian a été le premier à proposer l’idée d’utiliser la loi de Benford pour détecter une fraude fiscale. Il a montré que dans les données falsifiées, les premiers chiffres significatifs 5 et 6 prédominaient : 40% pour le 5 (au lieu de 7.9%) et plus de 20% pour le 6 (au lieu de 6,7%).

En 1993, un employé du Trésor de l’Arizona a été reconnu coupable d’une tentative de fraude de deux millions de dollars. Une série de chèques qu’il avait émis ne suivait pas la loi de Benford !

« La loi de Benford a été utilisée dans de nombreuses autres circonstances, indique Jean-François Cœurjolly : pour détecter des fraudes électorales, pour illustrer des abus sur les prix lorsque la monnaie française est passée à l’euro en 2000, pour détecter des fraudes dans les publications scientifiques, etc. »

les données chinoises ne s’écartent pas davantage du modèle que celles des trois autres pays considérés.

#coronavirus #Chine #mathématiques #fraude #falsification #Loi_de_Benford #Simon_Newcomb #Frank_Benford

Designer Snowflakes - SnowCrystals.com
▻http://www.snowcrystals.com/designer/designer.html

http://www.snowcrystals.com/designer/IMG_5128-34-A1asm.jpg

a photograph of a designer snowflake — a real sliver of ice, grown from water vapor, but created under controlled conditions in the laboratory. They might also be called synthetic crystals, but I call them designer snowflakes because I am free to design the final shape by changing the temperature and humidity as the crystal grows.

et un livre complet sur les flocons de neige :
▻https://arxiv.org/ftp/arxiv/papers/1910/1910.06389.pdf

L’urgence de santé publique, ce n’est pas la surveillance massive mais le dépistage de masse !
►https://reflets.info/articles/l-effet-placebo-du-tracage-informatique

https://reflets.info/system/images/attachments/d7c2ce69-9079-4470-a536-4323fc0f9836/big.jpg

Tracking GSM, GPS ou Bluetooth ? Aucun système ne pourra répondre à l’urgence sanitaire

…

Quand bien même les applis Bluetooth seraient assez précises, un autre contre-argument n’est jamais pointé du doigt : les ondes passent à travers les cloisons, portes et fenêtres, ou tout simplement à travers une vitre ou une plaque en plexiglass ! Ainsi, être à moins d’un mètre derrière un guichet protégé (conforme aux « gestes barrières ») pourrait être identifié, à tort, comme « contact à risque »… Comme de côtoyer tous les jours un voisin très proche... derrière un mur mitoyen. Tout autre système de géolocalisation (antennes relais comme GPS) possède le même inconvénient.

Une foule d’autres points noirs invalident l’utilité même du contact tracing. Comme la nécessité d’atteindre une certaine « masse critique » d’utilisateurs pour être efficace — comme l’a reconnu le gourou numérique du gouvernement, Cedric O. Cette masse critique, selon divers avis convergents, doit représenter au moins 60 % d’une population. Même à Singapour, État policier — et ultraconnecté — de 7 millions d’habitants, moins de 20% utilisent l’appli. Sans même parler des taux d’équipements en smartphone, très inégalitaires selon les régions ou les classes sociales, et qui excluent en premier les plus de 70 ans.