#text-as-data
LightNews LightNews
buffalocia.bsky.social
Buck fucking wild that 8 Democrats voted for this.
Screenshot of New York Times article from 3 hours ago. Text: The provision would make it a violation of the law to not notify a senator if their phone records or other metadata was taken from a service provider like a phone company. There are some exceptions, such as 60-day delays in notification if the senator is considered the target of an investigation. The language of the bill states that “any senator whose Senate data, or the Senate data of whose Senate office, has been acquired, subpoenaed, searched, accessed, or disclosed in violation of this section may bring a civil action against the United States if the violation was committed by an officer, employee, or agent of the United States or of any Federal department or agency.” Because the provision is retroactive to 2022, it would appear to make eligible the eight lawmakers whose phone records were subpoenaed by investigators for Mr. Smith as he examined efforts by Donald J. Trump to obstruct the results of the 2020 presidential election.
November 10, 2025 at 10:05 PM
themarkup.org
A new California law could change how all Americans control their data when using the internet, according to experts. Here’s what it does:

1/2
Hed: What the new law does Text: It requires companies that make web browsers to offer users an opt-out “signal” that automatically tells websites not to share or sell their personal information as they browse. Browsers like Google’s Chrome and Microsoft’s Edge will have until the beginning of 2027 to create a way for consumers to select the signal.
November 7, 2025 at 10:35 PM
clivethompson.bsky.social
Interestingly, younger Americans are significantly more likely to dislike data centers than older ones

and dislike is bipartisan

This from a Heatmap story (paywalled): heatmap.news/energy/data-...
An infographic titled “Younger People Really Don’t Like Data Centers” presents survey results showing attitudes toward data centers being built nearby, broken down by age group. A brief description explains that data centers house servers for the internet, apps, and artificial intelligence, and respondents were asked if they would support or oppose one being built near where they live. The results are displayed in horizontal bar charts with three colors: red for “Total support,” yellow for “Not sure,” and blue for “Total oppose.” Among 18- to 34-year-olds, 36% support, 16% are unsure, and 48% oppose. For ages 35 to 49, 37% support, 12% are unsure, and 51% oppose. For ages 50 to 64, 47% support, 14% are unsure, and 40% oppose. For those 65 and older, 51% support, 17% are unsure, and 32% oppose. A caption notes that the poll was of 3,741 registered American voters, conducted August 22–29, 2025, by Heatmap Pro and Embold Research, with a margin of error of ±1.7 percentage points. The Heatmap logo appears at the bottom right. A screenshot of a paragraph from a Heatmap article discussing public opposition to data centers. The text explains that Heatmap Pro conducted polling to understand why Americans are resistant to nearby data centers despite the billions spent on cloud computing and AI development. The article notes that fewer than half of Americans support having a data center near them and that dislike is widespread. A highlighted section reads: “The hostility crosses party lines, with Republicans nearly as likely to express disdain towards these projects as Democrats.” The passage concludes that frustration with data centers increases among younger generations, especially those aged 18 to 49, who are more familiar with AI.
November 10, 2025 at 10:16 PM
thecovidinfoguy.bsky.social
US Weekly COVID update: Nov 10, 2025

🔸1 in 162 Actively Infectious
🔸301,000 New Daily Infections
🔸1,990,000 Infections In The Past Week
🔸199,000,000 Infections in 2025
🔸100,000 to 400,000 Weekly Long COVID Cases
🔸600 to 900 Weekly Deaths

Source: pmc19.com/data/
A line and area chart titled “SARS-CoV-2 New Daily Infections, Wastewater-Derived Estimates (U.S.)” dated November 10, 2025. The x-axis shows years from 2020 to 2025. The y-axis shows estimated daily infections ranging from 0 up to 5 million. The chart displays repeating large waves of infection, with a very sharp peak in early 2022 (reaching near the top of the chart), followed by multiple smaller but still large waves through 2023, 2024, and 2025. A label in the center reads “Nov 10, 2025.” At the bottom, text explains that current transmission is estimated to be about 25% of the peak from two months prior, but still very high overall. It notes many COVID-conscious patients are scheduling appointments soon, despite lack of localized CDC data. The chart is credited to Michael Hoerger, Pandemic Mitigation Collaborative (PMC). A U.S. map labeled “COVID-19 Heat Map, Based on CDC Wastewater Data and Levels (U.S.)” dated November 10, 2025. Most states are shown in uniform grey labeled “No Data,” indicating a widespread loss of CDC wastewater reporting. Puerto Rico is highlighted in yellow and light orange, showing low to moderate transmission. A key at the bottom shows color categories ranging from “Very High” (dark red) to “Very Low” (yellow), as well as “No Data” (grey) and “Limited Coverage” (striped). The surrounding text explains that the CDC wastewater dashboard remains largely offline, leaving only Puerto Rico reporting separately. It advises consulting localized dashboards and WastewaterSCAN data, but cautions that non-CDC regional comparisons are often inconsistent and should be interpreted carefully. The map is credited to Michael Hoerger, Pandemic Mitigation Collaborative (PMC). A data summary table titled “National COVID-19 Estimates (U.S.)” dated November 10, 2025. The table is divided into three sections: Infections, Long COVID, and Excess Deaths. Infections: Proportion actively infectious: 1 in 162 people (0.6%) Estimated new daily infections: 301,000 Estimated infections in the past week: 1,990,000 Estimated total infections in 2025 so far: 199,000,000 Average cumulative infections per person: 4.72 Long COVID: Long COVID cases resulting from new daily infections: 15,000 to 60,000 Long COVID cases resulting from new weekly infections: 100,000 to 400,000 Excess Deaths: Excess deaths from new daily infections: 90 to 140 deaths Excess deaths from new weekly infections: 600 to 900 deaths A note below states that new daily infections are estimated at 301,000 and transmission is flat but expected to rise soon. The source listed is pmc19.com/data from the Pandemic Mitigation Collaborative. A table titled “National COVID-19 Risk Table (U.S.)” dated November 10, 2025. The table shows the probability that at least one person is infectious with SARS-CoV-2 in a group of a given size, assuming average national conditions and no masking or testing. The left column lists the number of people present (1, 2, 3, 4, 5, 10, 15, 20, 25, 30, 50, 75, 100, 200, 300). The right column lists the corresponding chance that someone is infectious, starting at 0.6% for a group of one person and increasing with group size: about 3% for a group of 5, 6% for 10, 14% for 25, 27% for 50, 46% for 100, 71% for 200, and 84% for 300. A note below explains that this is based on an estimated national infection rate of 1 in 162 people actively infectious, and highlights that even with seemingly low prevalence, exposure risk in classrooms and larger gatherings can be substantial.
November 9, 2025 at 11:46 PM
rob-sheridan.com
David Fincher's THE SOCIAL NETWORK, and Trent Reznor / Atticus Ross's Oscar-winning score, turned 15 years old this fall. Here's some of my artwork for the album, created by manually corrupting images from the film by editing their raw hex data in a text editor, damaging them from the inside. 1/
"Trent Reznor and Atticus Ross: The Social Network" album cover art by Rob Sheridan, 2010. A glitched image of Jesse Eisenberg as Mark Zuckerberg from the film. Full gallery and commentary at https://www.patreon.com/posts/41677761 Artwork by Rob Sheridan from the Trent Reznor and Atticus Ross soundtrack album "The Social Network." Digitally corrupted still from the film. Full gallery and commentary at https://www.patreon.com/posts/41677761 Artwork by Rob Sheridan from the Trent Reznor and Atticus Ross soundtrack album "The Social Network." Digitally corrupted still from the film. Full gallery and commentary at https://www.patreon.com/posts/41677761 Artwork by Rob Sheridan from the Trent Reznor and Atticus Ross soundtrack album "The Social Network." Digitally corrupted still from the film. Full gallery and commentary at https://www.patreon.com/posts/41677761
November 7, 2025 at 8:17 PM
eyeofacat.bsky.social
Bluesky developer blog on ‘group-private data’ (ie, private accounts/circles/friendslock):

docs.bsky.app/blog
screenshot: text reads “Bluesky. Private Data. We believe that group-private data is absolutely necessary for the long-term success of the protocol and the Atmosphere. Every day, we wish that we had this figured out and developed already. But as mentioned earlier, we believe that we need to land and mature the existing protocol around public broadcast data before we move on to the next big chunk of work.” screenshot: text reads “We continue to have internal discussions around private data. Paul shared some leaflets that give a sense of the approaches that we're considering and the rubric by which we're judging them. The AT Protocol Community is also coordinating a Private Data Working Group to explore some designs for how the system could work.”
November 7, 2025 at 5:54 PM
stolenpyjak.bsky.social
🚀 SynthTextEval, our open-source toolkit for generating and evaluating synthetic text data for high-stakes domains, will be featured at EMNLP 2025 as a system demonstration!

GitHub: github.com/kr-ramesh/sy...
Paper 📝: aclanthology.org/2025.emnlp-d...

#EMNLP2025 #EMNLP #SyntheticData
GitHub - kr-ramesh/synthtexteval: SynthTextEval: A Toolkit for Generating and Evaluating Synthetic Data Across Domains (EMNLP 2025 System Demonstration)
SynthTextEval: A Toolkit for Generating and Evaluating Synthetic Data Across Domains (EMNLP 2025 System Demonstration) - kr-ramesh/synthtexteval
github.com
November 7, 2025 at 12:53 AM
confluent.io
Call for Papers for Current26 in Bengaluru and London is now open!

Share your talks on Data Streaming, Kafka, or the future of AI in production etc with the world's most innovative data community.

Deadline December 22nd 2025

Bengaluru ➡️ https://cnfl.io/49L8V2J
London ➡️ https://cnfl.io/47IBj2Y
Alt Text: Promotional image for "Current: The Data Streaming Event" by Confluent, announcing that the call for papers is now open. The event dates are listed as April 22, 2026, in Bengaluru and May 19-20, 2026, in London. The background is dark with flowing neon lines.
November 10, 2025 at 8:00 AM
mariozechner.at
Can an LLM create a paper on LLM-usage by newspapers as analyzed by @schenior.bsky.social and myself here on BlueSky based on our BlueSky threads and the YouTube video transcripts?

Yes, yes it can. In 15 minutes. Let's submit it and see if it gets accepted.
Screenshot Chrome. Left, an unrolled thread by Jenia where she details some stylometric statistics based on the newspaper article data she analyzed. Right, a browser AI is reading that thread and compiling it into a paper, along with other threads and YouTube video transcripts it read before. Prompt: additional bsky threads freom my Partner in crime, I have three more tabs open now for sky view. Please look at all of these as well and work those into the markdown file as well, please. They are all related to the other three threads. You can probably figure out the order. They also have figures in them and charts with alt text. So work those in also as appropriate. And make sure everything you do is comprehensive. And once you've completed that task, ask me what data you still require to finish off the paper.
November 7, 2025 at 7:40 PM
mutualaidmamas.bsky.social
Our Fall Free Store is coming up fast! Shop for clothes, shoes, accessories, menstrual products, and diapers completely free, Sat., Nov 22, 10AM-2PM, at the playground at East Somerville Community School.

We also need volunteers! Sign up link in bio 🥰

Español/Português/Kreyòl Ayisyen 🧵⬇️
A flier with a lime green background, orange and red-toned leaves, socks, a hat, and a scarf is pictured. There is a white, oval in the center with text. "MAMAS" is on top and centered, in multi-colored text. "FREE STORE" is written in large, black text below. "FREE CLOTHES, SHOES, & ACCESSORIES FOR ALL AGES & SIZES" is written in smaller text below. Under that, "Saturday, November 22 10 AM-2PM, Rain date: 11/23 East Somerville Community School, 50 Cross St, Somerville" is written in black text. At the bottom "Email askyourmamas@gmail.com with any questions." is written. MAMAS TIENDA GRATUITA ROPA, ZAPATOS Y ACCESORIOS GRATIS PARA TODAS LAS EDADES Y TALLAS Sábado 22 de noviembre 10:00 a 14:00 Fecha alternativa por lluvia: 23 de noviembre Escuela Comunitaria East Somerville 50 Cross St, Somerville Envía un correo electrónico a askyourmamas@gmail.com si tienes alguna pregunta. MAMAS LOJA GRATUITA ROUPAS, SAPATOS E ACESSÓRIOS GRATUITOS PARA TODAS AS IDADES E TAMANHOS Sábado, 22 de novembro das 10h às 14h. Data alternativa em caso de chuva: 23/11 East Somerville Community School 50 Cross St, Somerville Envie um e-mail para askyourmamas@gmail.com com qualquer dúvida. MAMAS MAGAZEN GRATIS RAD, SOULYE, AK AKSESWA GRATIS POU TOUT LAJ AK TOUT GWOSÈ Samdi 22 novanm 10 AM - 2 PM Dat lapli: 23/11 Lekòl Kominotè East Somerville 50 Cross St, Somerville Voye yon imèl nan askyourmamas@gmail.com pou nenpòt kesyon.
November 10, 2025 at 9:16 PM
luisazintgraf.bsky.social
So what does the DataRater learn? It automatically identifies and down-weights data that aligns with human intuitions of low quality, such as incorrect text encodings, OCR errors, and irrelevant content.
November 6, 2025 at 11:29 AM
taconite.bsky.social
Adding the vision checks meant that the old "store in texture" version needed to be upgraded.

The data itself wasn't bad, but storing it on disk took a couple of versions. Best of all, it compresses into basically nothing!

#gamedev #indiegame #unity3d #solodev
Screenshot of unity editor showing the inspector viewing a Text Asset called "Humanoid Mask_OLD". The file contains a JSON serialization of some structs, with all the fieldnames baked in. How many million bytes are wasted storing "distances" every damn time... Screenshot of the unity editor showing the inspector viewing a Text Asset called "Humanoid Mask". The file contains the same structs, as custom Base64 encoded bytes. Apparently unity 6 has binary serialization, but I haven't updated yet... Screenshot of windows explorer showing the two files (and associated compressed versions) and their relative file sizes. The old version is 242MB (6MB compressed), the new is 11MB (1.6MB compressed). Screenshot of VS showing a function in C# called ToBytes: public static byte[] ToBytes(Map mp) { byte[] buf = new byte[mp.GetSizeBytes()]; uint indx = 0; // Record size MathExtras.IntToBytes(buf, ref indx, mp.size.x); MathExtras.IntToBytes(buf, ref indx, mp.size.y); MathExtras.IntToBytes(buf, ref indx, mp.size.z); // Record data for (int i = 0; i < mp.storage.Length; i++) Map.Data.ToBytes(buf, ref indx, mp.storage[i]); return buf; }
November 8, 2025 at 2:22 AM
penny.girlfag.club
and, the reason she thinks it works, is, in the end, it's just data. if that data is transmitted audibly, thats great (and hot), but, if that data is transmitted through text, as long as she's able to process it, it works just as well
don't have that problem as much with auditory hypno
November 9, 2025 at 9:36 PM
spacethey.gay
alt text of review continued
please see previous post for the first part of the alt text: In the end, Data is ruled as not property, and while the episode ends on a note of uncertainty as to how to define Data as a being, the mere fact that this uncertainty exists is precisely why they cannot treat Data like a machine. We humans cannot prove our own consciousness to exist, but we believe it is there. I fucking love this episode, and maybe a five is generous, but it’s my review so fuck you.
November 9, 2025 at 4:37 AM
surrealdb.com
Wrestling with unstructured data and need to query it in a structured form? Discover how to bring historical texts to life by using LLMs to structure the text for storing and querying in SurrealDB as knowledge graphs. 👉 sdb.li/4nuv74M
November 5, 2025 at 2:02 PM
asaakyan.bsky.social
N-gram novelty is widely used as a measure of creativity and generalization. But if LLMs produce highly n-gram novel expressions that don’t make sense or sound awkward, should they still be called creative? In a new paper, we investigate how n-gram novelty relates to creativity.
N-gram novelty is widely used to evaluate language models' ability to generate text outside of their training data. More recently, it has also been adopted as a metric for measuring textual creativity. However, theoretical work on creativity suggests that this approach may be inadequate, as it does not account for creativity's dual nature: novelty (how original the text is) and appropriateness (how sensical and pragmatic it is). We investigate the relationship between this notion of creativity and n-gram novelty through 7542 expert writer annotations (n=26) of novelty, pragmaticality, and sensicality via close reading of human and AI-generated text. We find that while n-gram novelty is positively associated with expert writer-judged creativity, ~91% of top-quartile expressions by n-gram novelty are not judged as creative, cautioning against relying on n-gram novelty alone. Furthermore, unlike human-written text, higher n-gram novelty in open-source LLMs correlates with lower pragmaticality. In an exploratory study with frontier close-source models, we additionally confirm that they are less likely to produce creative expressions than humans. Using our dataset, we test whether zero-shot, few-shot, and finetuned models are able to identify creative expressions (a positive aspect of writing) and non-pragmatic ones (a negative aspect). Overall, frontier LLMs exhibit performance much higher than random but leave room for improvement, especially struggling to identify non-pragmatic expressions. We further find that LLM-as-a-Judge novelty scores from the best-performing model were predictive of expert writer preferences.
November 4, 2025 at 3:08 PM
sorayanadiamcdonald.com
if not Ponzi, why Ponzi-shaped?

www.theatlantic.com/technology/2...
Screenshot text from The Atlantic: For now, money is still pouring into the AI industry. But there’s also something circular about these investments. To wit: OpenAI has agreed to pay $300 billion to Oracle for new computing capacity, Oracle is paying Nvidia tens of billions of dollars for chips to install in one of OpenAI’s data centers, and Nvidia has agreed to invest up to $100 billion in OpenAI as it deploys Nvidia chips. Attempts to illustrate these circular investments have produced a series of byzantine charts that one software engineer referred to on X as “the technocapital hyperobject at the end of time.” The consensus seems to be that although this is legal, it likely cannot go on forever.
November 3, 2025 at 8:14 PM
sheev.calamitousintent.com
guardian (UK case):

The judge ruled: “An AI model such as Stable Diffusion which does not store or reproduce any copyright works (and has never done so) is not an ‘infringing copy’.”

www.theguardian.com/media/2025/n...
file photo of james cameron AI firm wins high court ruling after photo agency's copyright claim Ruling in case brought by Getty Images against Stability Al is seen as a blow to copyright owners Robert Booth UK technology editor A London-based artificial intelligence firm has won a landmark high court case examining the legality of Al models using vast troves of copyrighted data without permission. The ruling is seen as a blow to copyright owners' exclusive right to reap the rewards of their work, with one senior lawyer, Rebecca Newman, a legal director at Addleshaw Goddard, warning it means "the UK's secondary copyright regime is not strong enough to protect its creators". There was evidence that Getty's images were used to train Stability's model, which allows users to generate images with text prompts. Stability was also found to have infringed Getty's trademarks in some cases. The judge, Mrs Justice Joanna Smith, said the question of where to strike the balance between the interests of the creative industries on one side and the AI industry on the other was "of very real societal importance". But she was only able to rule on relatively narrow claims after Getty had to withdraw parts of its case during the trial this summer. Getty Images sued Stability AI for infringement of its intellectual property, alleging the AI company was "completely indifferent to what they fed into the training data" and scraped and copied millions of its images. Boom or bubble? Inside the Satn AL Inside the 33tn Al datacentre The judgment comes amid a row over how the Labour government should legislate on the issue of copyright and AI, with artists and authors including Elton John, Kate Bush, Dua Lipa and Kazuo It is looking at whether to introduce a "text and data mining exception" into UK copyright law, which would allow copyright works to be used to train AI models in the UK unless the rights holder opts their works out of such training, said lawyers at Mishcon de Reya who have been following the issue. Getty had to drop its original copyright claim as there was no evidence the training took place in the UK. But it continued with its suit claiming Stability was still using within its systems copies of its visual assets, which it called the "lifeblood" of its business. It claimed Stability AI had infringed its trademarks because some Al- generated images included Getty watermarks, and that it was guilty of "passing off". In a sign of the complexity of Al copyright cases, it essentially argued that Stability's image- generation model, called Stable Diffusion, amounted to an infringing copy because its making would have constituted copyright infringement had it been carried out in the UK. The judge ruled: "An AI model such as Stable Diffusion which does not store or reproduce any copyright works (and has never done so) is not an 'infringing copy'." She declined to rule on the passing off claim and ruled in favour of some of Getty's claims about trademark infringement related to watermarks.
November 11, 2025 at 2:58 AM
evil.red
Joe Armstrong spoke very highly of tiddlywiki:

tiddlywiki.com

You can run it as a local file, or as an application.

It serves a very similar niche to OneNote, and I prefer it.

You can run it as a local file, an application, or if you really want to a cloud thing.
Text as follows: Welcome to TiddlyWiki, a unique non-linear notebook for capturing, organising and sharing complex information Use it to keep your to-do list, to plan an essay or novel, or to organise your wedding. Record every thought that crosses your brain, or build a flexible and responsive website. TiddlyWiki stores its data and code in a single HTML file, requiring no installs, no external dependencies, just a web browser. TiddlyWiki lets you choose where to keep your data, guaranteeing that in the decades to come you will still be able to use the notes you take today. TiddlyWiki is infinitely customisable and extensible with many plugins that add new features. TiddlyWiki is the product of a collective of developers, part of an extensive community of users.
November 6, 2025 at 8:24 PM
courtneymilan.com
This is what you will see first about Alli Jackson when you hit her website.

This is not someone talking about public-private partnerships and paying homage to small business. She is prioritizing immigrants, unionization, the environment.
A photo of Alli Jackson, a Black woman, holding a mic and talking while wearing a flannel shirt. Text: Alli Jackson is a foster mom and social worker who knows what struggling families face every day. She's running to bring affordable housing, dignity, and community first to Aurora city council. The Platform Alli Jackson's priorities: Keep immigrant families safe and welcome in Aurora Aurora is home to more than 83,000 immigrants-21% of the population. As the daughter of an immigrant, Alli will ensure City Hall supports immigrant families and keeps them safe. Todos son bienvenidos en Aurora. Protect city workers with fair pay and safe workplaces Aurora's city workers keep the city running. Alli will push for: • Fair wages • Safe Workplaces • The right to organize a union Strong unions mean strong families, and Alli will support pro-worker city contracts. Ban fracking and protect clean air and water Fracking puts Aurora's health at risk. Alli will: • Ban fracking in Aurora's backyard • Create a community-led Environmental Commission • Use data to guide stronger protections for air, water, and land Ensure dignity for all with public trash cans and restrooms
November 5, 2025 at 5:44 PM
stephjayb.bsky.social
@mattseybold.bsky.social's proposal to (among other things) reestablish faculty and students' control over their data & work here is...brilliant, actually?

(Tho it would strip upper admin of their power to decide WHO gets to buy that data...😢)

theamericanvandal.substack.com/p/mamdani-wi...
Text reads: If CUNY is successful in developing internally owned-and-operated substitutes for any or all of its Software As A Service (SaaS) contracts with for-profit EdTech firms (and their private equity owners), the development costs will very rapidly pay for themselves, merely by freeing the associated institutions (and the city) from ongoing Ponzi austerity extraction of public resources at the expense of investment elsewhere in the CUNY system. CUNY-owned EdTech platforms would provide an additional public benefit by reclaiming student, instructional, and employee data, the reserve currency of technofeudalism which is presently being mined by EdTech operators like OpenAI, Oracle, KKR, Vista and Silver Lake Partners. Universities are currently paying these SaaS providers and their investors twice, first with large subscription fees and again with access and aggregation power over large troves of data produced by members of their campus communities.
November 5, 2025 at 9:11 PM
libbyheeren.bsky.social
I had SO MUCH FUN making the data science trivia game for positconf Virtual Day this yr, & all I've wanted since then is to do it again! But it was a big lift to make it happen. What if I did daily trivia instead? I could do it as a poll on the data science community discord! #databs #python #rstats
Screenshot from the Posit::conf(2025) virtual event “Data Science Trivia Game.” Libby Heeren appears in the top-right corner, laughing as she hosts. The quiz question on screen reads: “This line of code will only run in one of these languages – which is it?” The code shown is {x: x**2 for x in [1,2,3,4,5]}, and the answer options are Julia, R, and Python. The Slido quiz interface shows it’s “Live quiz (1/42)” with 4 seconds left on the timer, and a colorful confetti-style emoji on the left is just below text that reads, “Join at slido.com.”
November 4, 2025 at 2:55 PM
thecovidinfoguy.bsky.social
US Weekly COVID update: Nov 3, 2025

🔸1 in 209 Actively Infectious
🔸234,000 New Daily Infections
🔸1,680,000 Infections In The Past Week
🔸196,000,000 Infections in 2025
🔸84,000 to 340,000 Weekly Long COVID Cases
🔸500 to 800 Weekly Deaths

Source: pmc19.com/data/
Graph titled “SARS-CoV-2 New Daily Infections, Wastewater-Derived Estimates (U.S.)” by the Pandemic Mitigation Collaborative (PMC), dated Nov 3, 2025. It shows estimated U.S. COVID-19 daily infections from 2020 to 2025 based on wastewater data. Peaks are visible in early 2022, late 2022, mid-2023, and early 2024, followed by smaller waves in 2025. Current estimated infections are around 234,000 per day, just below 20% of the summer 2025 peak. Text below notes that transmission remains high but has declined in relative risk, and that many COVID-conscious patients are scheduling medical appointments over the next two weeks. Map of the United States titled “COVID-19 Heat Map, Based on CDC Wastewater Data and Levels (U.S.)” by the Pandemic Mitigation Collaborative (PMC), dated Nov 3, 2025. The map shows nearly all states in gray, indicating “No Data”, as CDC wastewater data remains largely offline. Only Puerto Rico is shaded, showing similar trends to the national decline. A note explains that transmission has dropped into a relative lull but warns that non-CDC estimates (like WastewaterSCAN) vary in accuracy, especially for cross-regional comparisons. A color legend at the bottom ranges from Very High (dark red) to Very Low (yellow). Table titled “National COVID-19 Estimates (U.S.)” from the Pandemic Mitigation Collaborative (PMC), dated Nov 3, 2025. It lists national estimates based on wastewater-derived data: Infections: • 1 in 209 people (0.5%) actively infectious • 234,000 new daily infections • 1.68 million infections in the past week • 196 million infections in 2025 so far • 4.71 cumulative infections per person Long COVID: • 12,000–47,000 new daily long COVID cases • 84,000–340,000 weekly long COVID cases Excess deaths: • 70–110 daily • 500–800 weekly Text below notes that transmission is “basically flat” at present and unlikely to decline further. Table titled “National COVID-19 Risk Table (U.S.)” from the Pandemic Mitigation Collaborative (PMC), dated Nov 3, 2025. It shows the probability that at least one person in a group is infectious with COVID-19, assuming 1 in 209 (0.5%) are actively infectious nationally: 1 person → 0.5% chance 5 people → 2.4% 10 people → 4.7% 25 people → 11.3% 50 people → 21.3% 100 people → 38.1% 200 people → 61.7% 300 people → 76.3% Text below notes that even with relatively low prevalence, exposure risk remains significant in schools and large gatherings where many people interact.
November 3, 2025 at 11:52 PM
mariozechner.at
Dieser Cell Artikel behandelt auch konkret jenes Paper, auf dessen Methode Herr Guenzel aufbaut.

Ich hätte ja erwartet, ein Deutscher, wie Hern Guenzel, hätte dbzgl. ein bissl mehr Sensibilität. Das wird einem aber anscheinend in den Wirtschaftsstudien ausgetrieben.
Physiognomy resurrected “Physiognomy” is “the facility to identify, from the form and constitution of external parts of the human body, chiefly the face, exclusive of all temporary signs of emotions, the constitution of the mind and the heart.” Georg Christoph Lichtenberg, 1778 Recent years have seen an abundance of papers promulgating physiognomic methods resting on ML models.26 Work of this ilk is undertaken by academic research groups, private firms, and government agencies. A number of representative instances of each claim to have trained ML classifiers to predict personality, behavioral, or identity characteristics from image, text, voice, or other biometric data. Inferred labels have included race,30 sexuality,2021 mental illness,16 criminal propensity,7 autism,8 and neuroticism.9 These studies have predominantly relied on deep learning neural networks (DNNs), sometimes in tandem with more simplistic regression techniques. The practice of wielding the methods of ML toward the (putative) prediction of internal mental states, dispositions, or behavioral propensities based on outwardly visible morphology has been labeled “AI pseudoscience,” “digital phrenology,” “physiognomic AI,” “AI snake oil,” “bogus AI,” and “junk science.”31 These technologies, however, do not only exist in the abstract—a growing number of companies now market physiognomic capabilities, including the ability to detect academic dishonesty in students24 and future performance in prospective employees.23 Remarkably, a single tool marketed to defense contractors boasts of the ability to predict “pedophilia,” “terrorism,” and “bingo playing.”22 In this section, we review the details of several representative examples of physiognomic ML. These case studies are intended to be illustrative of the kinds of reasoning, epistemic foundations, and logic behind research and applications of automated inference from images portraying human likenesses. Personality psychology Kachur et al.9 write that “morphological and social cues in a human face provide signals of human personality and behaviour.” Their stated hypothesis is that a “photograph contains cues about personality that can be extracted using machine learning.” The authors further claim to have “circumvented the reliability limitations of human raters by developing a neural network and training it on a large dataset labelled with self-reported Big Five traits.” Here, deep learning is invoked as a means to obtain objectivity beyond human judgment; however, the training dataset was self-labeled by human raters. The predictive accuracy is interpreted as prima facie evidence for their hypothesis that structural features of human faces contain information of human personality and behavior, and the authors state that their “study presents new evidence confirming that human personality is related to individual facial appearance.”9 In this study, participants self-reported personality characteristics by completing an online questionnaire and then uploaded several photographs, which the researchers then used to construct their training and test datasets. In this example, as in Wang and Kosinski,2021 researchers used the accuracy of their ML model as confirmatory evidence of a joint causal basis for both facial morphology and self-reported personality. Kachur et al. report “several theoretical reasons to expect associations between facial images and personality” including that “genetic background contributes to both face and personality.”9 Kachur et al. described their results as being indicative of “a potential biological basis” to the discovered association between face images and self-reported personality characteristics.9
November 7, 2025 at 8:57 AM
aida-alvera.bsky.social
yes we do recover them 😅 We remain coastal so I think mobile network would be enough (another Q is position accuracy) My doubt is also if cheap luggage tags give position as text, or if they only provide a map, which would be of no use since we want the students to work with the data afterwards...
November 7, 2025 at 12:10 PM