Pinboard (jm) https://pinboard.in/u:jm/public/ recent bookmarks from jm TurboQuant: Redefining AI efficiency with extreme compression 2026-03-25T16:29:26+00:00 https://research.google/blog/turboquant-redefining-ai-efficiency-with-extreme-compression/ jm - High-quality compression (the PolarQuant method): TurboQuant starts by randomly rotating the data vectors. This clever step simplifies the data's geometry, making it easy to apply a standard, high-quality quantizer (a tool that maps a large set of continuous values, like precise decimals, to a smaller, discrete set of symbols or numbers, like integers: examples include audio quantization and jpeg compression) to each part of the vector individually. This first stage uses most of the compression power (the majority of the bits) to capture the main concept and strength of the original vector. - Eliminating hidden errors: TurboQuant uses a small, residual amount of compression power (just 1 bit) to apply the QJL algorithm to the tiny amount of error left over from the first stage. The QJL stage acts as a mathematical error-checker that eliminates bias, leading to a more accurate attention score. QJL: The zero-overhead, 1-bit trick QJL uses a mathematical technique called the Johnson-Lindenstrauss Transform to shrink complex, high-dimensional data while preserving the essential distances and relationships between data points. It reduces each resulting vector number to a single sign bit (+1 or -1). This algorithm essentially creates a high-speed shorthand that requires zero memory overhead. To maintain accuracy, QJL uses a special estimator that strategically balances a high-precision query with the low-precision, simplified data. This allows the model to accurately calculate the attention score (the process used to decide which parts of its input are important and which parts can be safely ignored). PolarQuant: A new “angle” on compression PolarQuant addresses the memory overhead problem using a completely different approach. Instead of looking at a memory vector using standard coordinates (i.e., X, Y, Z) that indicate the distance along each axis, PolarQuant converts the vector into polar coordinates using a Cartesian coordinate system. This is comparable to replacing "Go 3 blocks East, 4 blocks North" with "Go 5 blocks total at a 37-degree angle”. This results in two pieces of information: the radius, which signifies how strong the core data is, and the angle indicating the data’s direction or meaning). Because the pattern of the angles is known and highly concentrated, the model no longer needs to perform the expensive data normalization step because it maps data onto a fixed, predictable "circular" grid where the boundaries are already known, rather than a "square" grid where the boundaries change constantly. This allows PolarQuant to eliminate the memory overhead that traditional methods must carry. ]]> ai tech vectors search quantization turboquant research algorithms compression papers qjl error-detection polarquant https://pinboard.in/ https://pinboard.in/u:jm/b:1abb51ace2eb/ Measuring Agents in Production 2026-03-19T18:15:49+00:00 https://muratbuffalo.blogspot.com/2026/03/measuring-agents-in-production.html jm llm usage real-world ai agents papers via:muratbuffalo https://pinboard.in/ https://pinboard.in/u:jm/b:1cfeb59fce4f/ On the Biology of a Large Language Model 2026-03-19T09:49:31+00:00 https://transformer-circuits.pub/2025/attribution-graphs/biology.html jmThe black-box nature of [LLMs] is increasingly unsatisfactory as they advance in intelligence and are deployed in a growing number of applications. Our goal is to reverse engineer how these models work on the inside, so we may better understand them and assess their fitness for purpose. [...] In recent years, many research groups have made exciting progress on tools for probing the insides of language models. These methods have uncovered representations of interpretable concepts – “features” – embedded within models’ internal activity. Just as cells form the building blocks of biological systems, we hypothesize that features form the basic units of computation inside models. However, identifying these building blocks is not sufficient to understand the model; we need to know how they interact. In our companion paper, Circuit Tracing: Revealing Computational Graphs in Language Models, we build on recent work (e.g. ) to introduce a new set of tools for identifying features and mapping connections between them – analogous to neuroscientists producing a “wiring diagram” of the brain. We rely heavily on a tool we call attribution graphs, which allow us to partially trace the chain of intermediate steps that a model uses to transform a specific input prompt into an output response. Attribution graphs generate hypotheses about the mechanisms used by the model, which we test and refine through follow-up perturbation experiments. ]]> claude llm research llms ai anthropic papers tracing https://pinboard.in/ https://pinboard.in/u:jm/b:67244e5e3bc4/ Ditching bike helmets laws better for health 2026-02-06T15:58:27+00:00 https://theconversation.com/ditching-bike-helmets-laws-better-for-health-42 jmIn 1991 Australia introduced mandatory bicycle helmet laws requiring all adults and children to wear a helmet at all times when riding a bike, despite opposition from cycling groups. The legislation increased helmet use - from about 30 to 80% - but was coupled with a 30 to 40% decline in the number of people cycling. Rates of head injuries among cyclists, which had been dropping through the 1980s, continued to fall before levelling out in 1993. We didn’t see the kind of marked reduction in head injury rates that would be expected with the rapid increase in helmet use. In fact, any reductions in injuries may simply have been the result of having fewer cyclists on the road and therefore fewer people exposed to the risk of head injuries. One researcher noted that after mandatory helmet laws were introduced there was a bigger decrease in head injuries among pedestrians than there was among cyclists. The improvements in the general road safety environment introduced in the 1980s are likely to have contributed far more to cyclist safety than helmet legislation. And the effects when compared against the benefits of physical activity:
A recent analysis compared the risks and benefits of leaving the car at home and commuting by bike. It found the life expectancy gained from physical activity was much higher than the risks of pollution and injury from cycling. Increased physical activity added 3 to 14 months to a person’s life expectancy, while the life expectancy lost from air pollution was 0.8 to 40 days. Increased traffic accidents wiped 5-9 days off the life expectancy. It is clear that the benefits of cycling outweigh the risks, with helmet legislation actually costing society more from lost health gains than saved from injury prevention.
]]> transport bikes safety health papers science helmets cycling laws australia https://pinboard.in/ https://pinboard.in/u:jm/b:0a923278298a/ _Cheap science, real harm: the cost of replacing human participation with synthetic data_ [pdf] 2025-12-18T10:47:45+00:00 https://synthetic-data-workshop.github.io/papers/13.pdf jmDriven by the goals of augmenting diversity, increasing speed, reducing cost, the use of synthetic data as a replacement for human participants is gaining traction in AI research and product development. This talk critically examines the claim that synthetic data can “augment diversity,” arguing that this notion is empirically unsubstantiated, conceptually flawed, and epistemically harmful. While speed and cost-efficiency may be achievable, they often come at the expense of rigour, insight, and robust science. Drawing on research from dataset audits, model evaluations, Black feminist scholarship, and complexity science, I argue that replacing human participants with synthetic data risks producing both real-world and epistemic harms at worst and superficial knowledge and cheap science at best. "Synthetic data: stereotypes compressed" is absolutely spot on. This doesn't give insights into human behaviour and beliefs, just into stereotypes. It is increasingly common in social science fields, under the names of "digital twins" and "silicon samples".]]> data surveys abeba-birhane papers ai synthetic-data digital-twins simulation testing social-science silicon-samples https://pinboard.in/ https://pinboard.in/u:jm/b:e5314d304e99/ Long COVID consensus 2025-05-12T15:54:05+00:00 https://link.springer.com/content/pdf/10.1186/s12941-025-00793-9.pdf jm long-covid research health medicine covid-19 papers diseases https://pinboard.in/ https://pinboard.in/u:jm/b:0250a0bb67d0/ CaMeL 2025-04-17T08:55:25+00:00 https://arstechnica.com/information-technology/2025/04/researchers-claim-breakthrough-in-fight-against-ais-frustrating-security-hole/ jmGoogle DeepMind has unveiled CaMeL (CApabilities for MachinE Learning), a new approach to stopping prompt-injection attacks that abandons the failed strategy of having AI models police themselves. Instead, CaMeL treats language models as fundamentally untrusted components within a secure software framework, creating clear boundaries between user commands and potentially malicious content. The new paper grounds CaMeL's design in established software security principles like Control Flow Integrity (CFI), Access Control, and Information Flow Control (IFC), adapting decades of security engineering wisdom to the challenges of LLMs. Honestly, this is great. Data flow tracing/taint checking is exactly the method that needed to be applied, IMO, so good job DeepMind. Also as Jeremy Kahn suggested, the name is definitely a shout-out to Perl, the language where taint checks were first widely-used. :) Paper: https://arxiv.org/pdf/2503.18813 (Via Jeremy Kahn.)]]> llms ai security via:trochee data-flow infosec taint-checking taint camel papers https://pinboard.in/ https://pinboard.in/u:jm/b:99a7637577e0/ Practical Rateless Set Reconciliation 2025-04-08T17:05:54+00:00 https://dl.acm.org/doi/pdf/10.1145/3651890.3672219 jmSet reconciliation, where two parties hold fixed-length bit strings and run a protocol to learn the strings they are missing from each other, is a fundamental task in many distributed systems. We present Rateless Invertible Bloom Lookup Tables (Rateless IBLTs), the first set reconciliation protocol, to the best of our knowledge, that achieves low computation cost and near-optimal communication cost across a wide range of scenarios: set differences of one to millions, bit strings of a few bytes to megabytes, and workloads injected by potential adversaries. Rateless IBLT is based on a novel encoder that incrementally encodes the set difference into an infinite stream of coded symbols, resembling rateless error-correcting codes. We compare Rateless IBLT with state-of-the-art set reconciliation schemes and demonstrate significant improvements. Rateless IBLT achieves 3–4× lower communication cost than non-rateless schemes with similar computation cost, and 2–2000× lower computation cost than schemes with similar communication cost. We show the real-world benefits of Rateless IBLT by applying it to synchronize the state of the Ethereum blockchain, and demonstrate 5.6× lower end-to-end completion time and 4.4× lower communication cost compared to the system used in production. ]]> set-reconciliation algorithms papers via:xmal sets data-structures bloom-tables error-correction https://pinboard.in/ https://pinboard.in/u:jm/b:9b964645d62b/ The Unbelievable Scale of AI’s Pirated-Books Problem 2025-03-20T12:47:54+00:00 https://www.theatlantic.com/technology/archive/2025/03/libgen-meta-openai/682093/?gift=iWa_iB9lkw4UuiWbIbrWGYDRoX8kfg3ZQZL6J-W0kQE jmOne of the biggest questions of the digital age is how to manage the flow of knowledge and creative work in a way that benefits society the most. LibGen and other such pirated libraries make information more accessible, allowing people to read original work without paying for it. Yet generative-AI companies such as Meta have gone a step further: Their goal is to absorb the work into profitable technology products that compete with the originals. Will these be better for society than the human dialogue they are already starting to replace? Also, I found this quote from a Meta Director of Engineering in the legal discovery output interesting: "The problem is that people don’t realize that if we license one single book, we won’t be able to lean into fair use strategy". huh.]]> books knowledge papers meta llama llms law piracy ip libgen genai fair-use https://pinboard.in/ https://pinboard.in/u:jm/b:b26e866b7827/ HAJPAQUE 2025-03-06T10:35:16+00:00 https://www.cse.iitd.ac.in/~srsarangi/files/papers/jsonparser.pdf jm State-of-the-art analytics pipelines can now process data at a rate that exceeds 50 Gbps owing to recent advances in RDMA, NVM, and network technology (notably Infiniband). The peak throughput of the best-performing software solutions for parsing, querying, and validating JSON data is 20 Gbps, which is far lower than the current requirement. > > We propose a novel [hardware-]based accelerator that ingests 16-bytes of JSON data at a time and processes all the 16 bytes in parallel as opposed to competing approaches that process such data byte by byte. Our novel solution comprises lookup tables, parallel sliding windows, and recursive computation. Together, they ensure that our online pipeline does not encounter any stalls while performing all the operations on JSON data. We ran experiments on several widely used JSON benchmarks/datasets and demonstrated that we can parse and query JSON data at 106 Gbps (@28 nm). (Via Rob)]]> accelerators papers asics json parsing throughput performance via:rsynnott https://pinboard.in/ https://pinboard.in/u:jm/b:0360f1cd8da2/ Language Models Do Addition Using Helices 2025-02-13T15:45:36+00:00 https://arxiv.org/abs/2502.00873 jmMathematical reasoning is an increasingly important indicator of large language model (LLM) capabilities, yet we lack understanding of how LLMs process even simple mathematical tasks. To address this, we reverse engineer how three mid-sized LLMs compute addition. We first discover that numbers are represented in these LLMs as a generalized helix, which is strongly causally implicated for the tasks of addition and subtraction, and is also causally relevant for integer division, multiplication, and modular arithmetic. We then propose that LLMs compute addition by manipulating this generalized helix using the "Clock" algorithm: to solve a+b, the helices for a and b are manipulated to produce the a+b answer helix which is then read out to model logits. We model influential MLP outputs, attention head outputs, and even individual neuron preactivations with these helices and verify our understanding with causal interventions. By demonstrating that LLMs represent numbers on a helix and manipulate this helix to perform addition, we present the first representation-level explanation of an LLM's mathematical capability. ]]> llms helices trigonometry magic weird ai papers arithmetic addition subtraction https://pinboard.in/ https://pinboard.in/u:jm/b:18e77e124914/ Critical Ignoring as a Core Competence for Digital Citizens 2025-02-13T12:33:41+00:00 https://journals.sagepub.com/doi/epub/10.1177/09637214221121570 jm Low-quality and misleading information online can hijack people’s attention, often by evoking curiosity, outrage, or anger. Resisting certain types of information and actors online requires people to adopt new mental habits that help them avoid being tempted by attention-grabbing and potentially harmful content. We argue that digital information literacy must include the competence of critical ignoring—choosing what to ignore and where to invest one’s limited attentional capacities. We review three types of cognitive strategies for implementing critical ignoring: - self-nudging, in which one ignores temptations by removing them from one’s digital environments; - lateral reading, in which one vets information by leaving the source and verifying its credibility elsewhere online; - and the do-not-feed-the-trolls heuristic, which advises one to not reward malicious actors with attention. We argue that these strategies implementing critical ignoring should be part of school curricula on digital information literacy. Good to give names to these practices, since we're all having to do them nowadays anyway... (Via Stan Carey)]]> psychology trolls media kids internet literacy attention critical-ignoring ignoring papers via:stancarey https://pinboard.in/ https://pinboard.in/u:jm/b:44249586bf57/ Undergraduate Upends a 40-Year-Old Data Science Conjecture 2025-02-11T14:21:54+00:00 https://www.quantamagazine.org/undergraduate-upends-a-40-year-old-data-science-conjecture-20250210/ jmKrapivin was not held back by the conventional wisdom for the simple reason that he was unaware of it. “I did this without knowing about Yao’s conjecture,” he said. His explorations with tiny pointers led to a new kind of hash table — one that did not rely on uniform probing. And for this new hash table, the time required for worst-case queries and insertions is proportional to (log x)^2 — far faster than x. This result directly contradicted Yao’s conjecture. Farach-Colton and Kuszmaul helped Krapivin show that (log x)^2 is the optimal, unbeatable bound for the popular class of hash tables Yao had written about. Paper here -- https://arxiv.org/abs/2501.02305 .]]> data-structures hash-tables cs programming coding papers optimization open-addressing https://pinboard.in/ https://pinboard.in/u:jm/b:988d8723a281/ Thalidomide chirality paradox explained 2024-12-18T09:57:36+00:00 https://www.nature.com/articles/s41598-018-35457-6 jmTwenty years after the thalidomide disaster in the late 1950s, Blaschke et al. reported that only the (S)-enantiomer of thalidomide is teratogenic [jm: causing birth defects]. However, other work has shown that the enantiomers ["mirror" molecules] of thalidomide interconvert in vivo, which begs the question: why is teratogen activity not observed in animal experiments that use (R)-thalidomide given the ready in vivo racemization (“thalidomide paradox”)? Herein, we disclose a hypothesis to explain this “thalidomide paradox” through the in-vivo self-disproportionation of enantiomers. Upon stirring a 20% ee solution of thalidomide in a given solvent, significant enantiomeric enrichment of up to 98% ee was observed reproducibly in solution. We hypothesize that a fraction of thalidomide enantiomers epimerizes in vivo, followed by precipitation of racemic [equally mixed between R/S forms] thalidomide in (R/S)-heterodimeric form. Thus, racemic thalidomide is most likely removed from biological processes upon racemic precipitation in (R/S)-heterodimeric form. On the other hand, enantiomerically pure thalidomide remains in solution, affording the observed biological experimental results: the (S)-enantiomer is teratogenic, while the (R)-enantiomer is not. ]]> chirality thalidomide molecules drugs medicine papers chemistry https://pinboard.in/ https://pinboard.in/u:jm/b:16cfd0723d05/ Black plastic won't kill you 2024-12-17T10:57:28+00:00 https://nationalpost.com/news/canada/black-plastic jmPlastics rarely make news like this. From Newsmax to Food and Wine, and from the Daily Mail to CNN, the media uptake was enthusiastic on a paper published in October in the peer-reviewed journal Chemosphere. “Your cool black kitchenware could be slowly poisoning you, study says. Here’s what to do,” said the LA Times. “Yes, throw out your black spatula,” said the San Francisco Chronicle. Salon was most blunt: “Your favorite spatula could kill you,” it said. [....] The paper correctly gives the reference dose for BDE-209 as 7,000 nanograms per kilogram of body weight per day, but calculates this into a limit for a 60-kilogram adult of 42,000 nanograms per day. So, as the paper claims, the estimated actual exposure from kitchen utensils of 34,700 nanograms per day is more than 80 per cent of the EPA limit of 42,000. That sounds bad. But 60 times 7,000 is not 42,000. It is 420,000. This is what Joe Schwarcz [director of McGill University’s Office for Science and Society] noticed. The estimated exposure is not even a tenth of the reference dose. ]]> cooking research science plastics errors maths math fail papers https://pinboard.in/ https://pinboard.in/u:jm/b:9a5dd3feb197/ GSM-Symbolic 2024-10-12T13:39:33+00:00 https://machinelearning.apple.com/research/gsm-symbolic jmWe investigate the fragility of mathematical reasoning in these models and show that their performance significantly deteriorates as the number of clauses in a question increases. We hypothesize that this decline is because current LLMs cannot perform genuine logical reasoning; they replicate reasoning steps from their training data. Adding a single clause that seems relevant to the question causes significant performance drops (up to 65%) across all state-of-the-art models, even though the clause doesn't contribute to the reasoning chain needed for the final answer. Even better -- "the performance of all models declines when only the numerical values in the question are altered" seems to suggest that great performance on benchmarks like GSM8K just mean that the LLMs have been trained on the answers...]]> training benchmarks ai llms gsm-symbolic reasoning ml apple papers gsm8k https://pinboard.in/ https://pinboard.in/u:jm/b:2528cf29b369/ Journals should retract Richard Lynn's racist 'research' articles 2024-07-03T10:53:38+00:00 https://www.statnews.com/2024/06/20/richard-lynn-racist-research-articles-journals-retractions/ jm Lynn, who died in 2023, was a professor at the University of Ulster and the president of the Pioneer Fund, a nonprofit foundation created in 1937 by American Nazi sympathizers to support “race betterment” and “race realism.” It has been a primary funding source of scientific racism and, for decades, Lynn was one of the loudest proponents of the unfounded idea that Western civilization is threatened by “inferior races” that are genetically predisposed to low intelligence, violence, and criminality. Lynn’s work has been repeatedly condemned by social scientists and biologists for using flawed methodology and deceptively collated data to support racism. In particular, he created deeply flawed datasets purporting to show differences in IQ culminating in a highly cited national IQ database. Many of Lynn’s papers appear in journals owned by the billion-dollar publishing giants Elsevier and Springer, including Personality and Individual Differences and Intelligence. The ESRI, for whom Lynn was a Research Professor in the 1960s and 70s, have quietly removed his output from their archives, thankfully. But as this article notes, his papers and faked datasets still feature in many prestigious journals. (via Ben)]]> richard-lynn racists research papers elsevier iq via:bwalsh https://pinboard.in/ https://pinboard.in/u:jm/b:7da29a2d21ad/ _ChatGPT is bullshit_ Ethics and Information Technology vol. 26 2024-06-13T10:54:02+00:00 https://link.springer.com/article/10.1007/s10676-024-09775-5 jmRecently, there has been considerable interest in large language models: machine learning systems which produce human-like text and dialogue. Applications of these systems have been plagued by persistent inaccuracies in their output; these are often called “AI hallucinations”. We argue that these falsehoods, and the overall activity of large language models, is better understood as bullshit in the sense explored by Frankfurt (_On Bullshit_, Princeton, 2005): the models are in an important way indifferent to the truth of their outputs. We distinguish two ways in which the models can be said to be bullshitters, and argue that they clearly meet at least one of these definitions. We further argue that describing AI misrepresentations as bullshit is both a more useful and more accurate way of predicting and discussing the behaviour of these systems. ]]> ai chatgpt hallucinations bullshit funny llms papers https://pinboard.in/ https://pinboard.in/u:jm/b:ede1616532f7/ The CVM algorithm 2024-05-21T16:10:10+00:00 https://arxiv.org/abs/2301.10191 jm algorithms approximation cardinality streaming estimation cs papers count-distinct distinct-elements https://pinboard.in/ https://pinboard.in/u:jm/b:6bcd3fefcac0/ Is the Atlantic Overturning Circulation Approaching a Tipping Point? 2024-05-13T16:23:49+00:00 https://tos.org/oceanography/article/is-the-atlantic-overturning-circulation-approaching-a-tipping-point jmTo some extent, tipping may even depend on the vagaries of weather. In NASA’s climate model, in 10 simulations using the same “middle-of-the-road” greenhouse warming scenario (SSP2–4.5) with under 3°C global warming, the AMOC collapses in two but recovers after significant weakening in eight; the difference is merely stochastic internal variability (Romanou et al., 2023). This is also part of the nature of tipping points. Apart from a full shutdown of the AMOC, there is still the second type of tipping point to consider, the one where convection shuts down in one region. That happens in a surprising number of climate models, and so far hasn’t gotten the public attention it deserves. The first documented case, the British Hadley Centre model, was published in 1999 (Wood et al., 1999). Of the latest model generation (CMIP6), in four out of the 35 models, subpolar gyre convection breaks down—and all four are in the group of the 11 best models in terms of reproducing the vertical density profiles in the subpolar gyre (Swingedouw et al., 2021). That’s in 36% of those high-​quality models. In the previous model generation (CMIP5), that number was 45%. What’s more, it typically happens as soon as the year 2040 and for moderate emission scenarios—​even without properly accounting for Greenland melt. Thus, a collapse of convection in the subpolar gyre, resulting in rapid AMOC weakening and abrupt regional cooling, must be considered a high risk urgently requiring attention. [...] A full AMOC collapse would be a massive, planetary-scale disaster. We really want to prevent this from happening. ]]> amoc ocean atlantic climate-change climate papers currents https://pinboard.in/ https://pinboard.in/u:jm/b:c92ece3b79ec/ [untitled] 2024-05-03T10:31:07+00:00 https://www.researchgate.net/profile/Raymond-Panko/publication/1907590_Thinking_is_Bad_Implications_of_Human_Error_Research_for_Spreadsheet_Research_and_Practice/links/53eb1f800cf28f342f44fb1e/Thinking-is-Bad-Implications-of-Human-Error-Research-for-Spreadsheet-Research-and-Practice.pdf jm In the spreadsheet error community, both academics and practitioners generally have ignored the rich findings produced by a century of human error research. These findings can suggest ways to reduce errors; we can then test these suggestions empirically. In addition, research on human error seems to suggest that several common prescriptions and expectations for reducing errors are likely to be incorrect. Among the key conclusions from human error research are that thinking is bad, that spreadsheets are not the cause of spreadsheet errors, and that reducing errors is extremely difficult. In past EuSpRIG conferences, many papers have shown that most spreadsheets contain errors, even after careful development. Most spreadsheets, in fact, have material errors that are unacceptable in the growing realm of compliance laws. Given harsh penalties for non-compliance, we are under considerable pressure to develop good practice recommendations for spreadsheet developers and testers. If we are to reduce errors, we need to understand errors. Fortunately, human error has been studied for over a century across a number of human cognitive domains, including linguistics, writing, software development and testing, industrial processes, automobile accidents, aircraft accidents, nuclear accidents, and algebra, to name just a few. The research that does exist is disturbing because it shows that humans are unaware of most of their errors. This “error blindness” leads people to many incorrect beliefs about error rates and about the difficulty of detecting errors. In general, they are overconfident, substantially underestimating their own error rates and overestimating their ability to reduce and detect errors. This “illusion of control” also leads them to hold incorrect beliefs about spreadsheet errors, such as a belief that most errors are due to spreadsheet technology or to sloppiness rather than being due primarily to inherent human error. ]]> spreadsheets errors programming coding bugs research papers via:arclight https://pinboard.in/ https://pinboard.in/u:jm/b:dcb02e2f4d2f/ Ribbon filter: Practically smaller than Bloom and Xor 2024-03-28T18:13:00+00:00 https://engineering.fb.com/2021/07/09/data-infrastructure/ribbon-filter/ jmBuilding on some prior lines of research, the Ribbon filter combines a simplified, faster, and more flexible construction algorithm; a data layout optimized for filter queries; and near-continuous configurability to make a practical alternative to static (immutable) Bloom filters. While well-engineered Bloom filters are extremely fast, they use roughly 50 percent more space (overhead) than the information-theoretic lower bound for filters on arbitrary keys. When Bloom filters cannot meet an application’s space efficiency targets, Ribbon filter variants dominate in space-versus-time trade-offs with near continuous configurability and space overhead as low as 1 percent or less. Ribbon filters have O(1) query times and save roughly 1/3 of memory compared with Bloom filters. At Facebook’s scale, we expect Ribbon filters to save several percent of RAM resources, with a tiny increase in CPU usage for some major storage systems. However, we do not implement efficiency gains at all engineering costs, so it’s also important to have a user-friendly data structure. This issue stalled implementation of other Bloom alternatives offering some space savings. The Ribbon filter opens these new trade-offs without introducing notable discontinuities or hazards in the configuration space. In other words, there is some complexity to make Ribbon filters general and highly configurable, but these details can be hidden behind a relatively simple API. You have essentially free choice over any three of the four core performance dimensions — number of keys added to the set, memory usage, CPU efficiency, and accuracy — and the accuracy is automatically well optimized. (via Tony Finch)]]> via:fanf algorithms facebook programming ribbon-filters data-structures bloom-filters set-membership papers https://pinboard.in/ https://pinboard.in/u:jm/b:9391321eb119/ Microplastics found to increase risk of serious outcomes for heart patients 2024-03-07T17:33:15+00:00 https://erictopol.substack.com/p/theres-plastic-in-my-plaque?publication_id=587835&post_id=142359342&triggerShare=true&isFreemail=true&r=3awpb&triedRedirect=true jm microplastics plastic sustainability health medicine atherosclerosis papers via:eric-topol https://pinboard.in/ https://pinboard.in/u:jm/b:1053c53daef5/ Trinity team discovers underlying cause of brain fog linked with Long COVID 2024-02-26T15:58:29+00:00 https://www.tcd.ie/news_events/top-stories/featured/trinity-team-discovers-underlying-cause-of-brain-fog-linked-with-long-covid/ jm Now, the findings reported by the Trinity team in the top international journal Nature Neuroscience showed that there was disruption to the integrity of the blood vessels in the brains of patients suffering from Long COVID and brain fog. This blood vessel “leakiness” was able to objectively distinguish those patients with brain fog and cognitive decline compared to patients suffering from Long-COVID but not with brain fog. The team led by scientists at the Smurfit Institute of Genetics in Trinity’s School of Genetics and Microbiology and neurologists in the School of Medicine have also uncovered a novel form of MRI scan that shows how Long-COVID can affect the human brain’s delicate network of blood vessels.  “For the first time, we have been able to show that leaky blood vessels in the human brain, in tandem with a hyperactive immune system may be the key drivers of brain fog associated with Long COVID. This is critically important, as understanding the underlying cause of these conditions will allow us to develop targeted therapies for patients in the future,” said Prof. Matthew Campbell, Professor in Genetics and Head of Genetics at Trinity, and Principal Investigator at FutureNeuro. Apparently, this is the first study to correlate blood-brain barrier disruption with "brain fog" symptoms of Long COVID, using an enhanced MRI procedure. This is a very significant step towards the discovery of biomarkers and therapies for neurological manifestations of Long COVID. Great to see this kind of significant research from TCD! Full paper: https://www.nature.com/articles/s41593-024-01576-9]]> tcd dublin ireland trinity long-covid covid-19 research papers neurology https://pinboard.in/ https://pinboard.in/u:jm/b:38ed6bd500ad/ Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training 2024-01-18T13:04:45+00:00 https://arxiv.org/abs/2401.05566 jm Humans are capable of strategically deceptive behavior: behaving helpfully in most situations, but then behaving very differently in order to pursue alternative objectives when given the opportunity. If an AI system learned such a deceptive strategy, could we detect it and remove it using current state-of-the-art safety training techniques? To study this question, we construct proof-of-concept examples of deceptive behavior in large language models (LLMs). For example, we train models that write secure code when the prompt states that the year is 2023, but insert exploitable code when the stated year is 2024. We find that such backdoor behavior can be made persistent, so that it is not removed by standard safety training techniques, including supervised fine-tuning, reinforcement learning, and adversarial training (eliciting unsafe behavior and then training to remove it). The backdoor behavior is most persistent in the largest models and in models trained to produce chain-of-thought reasoning about deceiving the training process, with the persistence remaining even when the chain-of-thought is distilled away. Furthermore, rather than removing backdoors, we find that adversarial training can teach models to better recognize their backdoor triggers, effectively hiding the unsafe behavior. Our results suggest that, once a model exhibits deceptive behavior, standard techniques could fail to remove such deception and create a false impression of safety. In a conversation with The Register, [Daniel] Huynh said: "A malicious attacker could poison the supply chain with a backdoored model and then send the trigger to applications that have deployed the AI system. [...] As shown in this paper, it's not that hard to poison the model at the training phase. And then you distribute it. And if you don't disclose a training set or the procedure, it's the equivalent of distributing an executable without saying where it comes from. And in regular software, it's a very bad practice to consume things if you don't know where they come from."]]> ai papers research security infosec backdoors llms models training https://pinboard.in/ https://pinboard.in/u:jm/b:f893e3ab740e/ Distinguishing features of Long COVID identified through immune profiling 2023-09-26T14:01:06+00:00 https://www.nature.com/articles/s41586-023-06651-y jm covid-19 diagnosis biomarkers long-covid putrino-lab akiko-iwasaki papers preprints nature medicine cortisol https://pinboard.in/ https://pinboard.in/u:jm/b:aebdb46ffabc/ up-to-date Long COVID data 2023-08-16T13:31:38+00:00 https://www.medrxiv.org/content/10.1101/2023.08.06.23293706v1 jm 18.2% (n=2,130) of respondents met case definition for Long COVID. Female sex, being 50-69 years of age, pre-existing health issues, residing in a rural or remote area, and receiving fewer vaccine doses were significant independent predictors of Long COVID (p < 0.05). Persons with Long COVID reported a median of 6 symptoms, most commonly fatigue (70.6%) and difficulty concentrating (59.6%); 38.2% consulted a GP and 1.6% reported hospitalisation in the month prior to the survey due to ongoing symptoms. Of 1,778 respondents with Long COVID who were working/studying before their COVID-19 diagnosis, 17.9% reported reducing/discontinuing work/study. [...] Long COVID was associated with sustained negative impacts on work/study and a substantial utilisation of GP services 2-3 months after the acute illness. ]]> covid-19 long-covid australia omicron medicine papers preprints via:danny-altmann https://pinboard.in/ https://pinboard.in/u:jm/b:d5965d9a6a14/ Is censorship of LLMs even possible? 2023-07-23T11:55:58+00:00 https://www.cl.cam.ac.uk/~is410/Papers/llm_censorship.pdf jmIs censorship of LLMs even possible? Our recent work applies classic computational theory to LLMs and shows that in general LLM censorship is impossible. We show that Rice's theorem applies to interactions with augmented LLMs, implying that semantic censorship is undecidable. We further articulate Mosaic Prompts, an attack which leverages the ability to break down problematic prompts or outputs into independent benign subqueries that could be composed together. Twitter: https://twitter.com/iliaishacked/status/1681953406171197440?s=20]]> censorship rice-theorem llms ml exploits security infosec papers https://pinboard.in/ https://pinboard.in/u:jm/b:0eeeb495a760/ Photoferrotrophic Bacteria Initiated Plate Tectonics in the Neoarchean 2023-07-03T10:52:19+00:00 https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2023GL103553 jm papers bacteria life tectonics earth climate iron science geology https://pinboard.in/ https://pinboard.in/u:jm/b:0ab08c4eb0ae/ Sleep Apnea Directly Tied to Early Cognitive Decline 2023-06-30T09:47:18+00:00 https://scitechdaily.com/startling-link-uncovered-sleep-apnea-directly-tied-to-early-cognitive-decline/?expand_article=1 jmresearchers from the UK, Germany, and Australia have shown for the first time that in middle-aged men, OSA can cause early cognitive decline, even in patients who are otherwise healthy and not obese. The results were recently published in the journal _Frontiers in Sleep_. “We show poorer executive functioning and visuospatial memory and deficits in vigilance, sustained attention, and psychomotor and impulse control in men with OSA. Most of these deficits had previously been ascribed to co-morbidities,” said Dr. Ivana Rosenzweig, a neuropsychiatrist who heads the Sleep and Brain Plasticity Centre at King’s College London, and the study’s lead author. “We also demonstrated for the first time that OSA can cause significant deficits in social cognition.” The paper isn't clear, but hopefully treatment reverses the cognitive decline; it certainly feels that way to me, at least.]]> sleep sleep-apnea cognition brains sleeping science papers https://pinboard.in/ https://pinboard.in/u:jm/b:cfda6414b953/ [2304.11082] Fundamental Limitations of Alignment in Large Language Models 2023-06-09T11:52:56+00:00 https://arxiv.org/abs/2304.11082 jmAn important aspect in developing language models that interact with humans is aligning their behavior to be useful and unharmful for their human users. This is usually achieved by tuning the model in a way that enhances desired behaviors and inhibits undesired ones, a process referred to as alignment. In this paper, we propose a theoretical approach called Behavior Expectation Bounds (BEB) which allows us to formally investigate several inherent characteristics and limitations of alignment in large language models. Importantly, we prove that for any behavior that has a finite probability of being exhibited by the model, there exist prompts that can trigger the model into outputting this behavior, with probability that increases with the length of the prompt. This implies that any alignment process that attenuates undesired behavior but does not remove it altogether, is not safe against adversarial prompting attacks. Furthermore, our framework hints at the mechanism by which leading alignment approaches such as reinforcement learning from human feedback increase the LLM's proneness to being prompted into the undesired behaviors. Moreover, we include the notion of personas in our BEB framework, and find that behaviors which are generally very unlikely to be exhibited by the model can be brought to the front by prompting the model to behave as specific persona. This theoretical result is being experimentally demonstrated in large scale by the so called contemporary "chatGPT jailbreaks", where adversarial users trick the LLM into breaking its alignment guardrails by triggering it into acting as a malicious persona. Our results expose fundamental limitations in alignment of LLMs and bring to the forefront the need to devise reliable mechanisms for ensuring AI safety. (via Remmelt Ellen)]]> papers ethics llms ai ml infosec security prompt-hacking exploits alignment https://pinboard.in/ https://pinboard.in/u:jm/b:7772521369fd/ "The Fallacy of AI Functionality" 2023-06-07T09:17:54+00:00 https://dl.acm.org/doi/pdf/10.1145/3531146.3533158 jm Deployed AI systems often do not work. They can be constructed haphazardly, deployed indiscriminately, and promoted deceptively. However, despite this reality, scholars, the press, and policymakers pay too little attention to functionality. This leads to technical and policy solutions focused on “ethical” or value-aligned deployments, often skipping over the prior question of whether a given system functions, or provides any benefits at all. To describe the harms of various types of functionality failures, we analyze a set of case studies to create a taxonomy of known AI functionality issues. We then point to policy and organizational responses that are often overlooked and become more readily available once functionality is drawn into focus. We argue that functionality is a meaningful AI policy challenge, operating as a necessary first step towards protecting affected communities from algorithmic harm. One mastodon user notes: "My favorite (sarcasm) example of this was police departments buying ML for identifying gunshots. The models were all trained for earthquakes, and the vendor basically repurposed earthquake detection as gunshot detection, made bank, and left departments with a flood of false positives."]]> papers false-positives ai ml fail software reliability enshittification https://pinboard.in/ https://pinboard.in/u:jm/b:a6a823d03dcc/ _Development of a Definition of Postacute Sequelae of SARS-CoV-2 Infection_ 2023-05-26T09:18:59+00:00 https://jamanetwork.com/journals/jama/fullarticle/2805540 jm long-covid covid-19 pasc papers medicine disease https://pinboard.in/ https://pinboard.in/u:jm/b:7365cf0c9bde/ StanzaSystems/awesome-load-management 2023-05-18T22:19:17+00:00 https://github.com/StanzaSystems/awesome-load-management jm load-shedding circuit-breakers quota-management throttling papers architecture patterns via:niallm https://pinboard.in/ https://pinboard.in/u:jm/b:250d15e46975/ Metformin, a new drug to prevent long covid 2023-03-09T11:18:42+00:00 https://erictopol.substack.com/p/a-break-from-covid-waves-and-a-breakthrough jm covid-19 long-covid metformin drugs papers https://pinboard.in/ https://pinboard.in/u:jm/b:d721034e71fc/ Long COVID: major findings, mechanisms and recommendations 2023-01-23T12:42:59+00:00 https://www.nature.com/articles/s41579-022-00846-2?s=03 jmLong COVID is an often debilitating illness that occurs in at least 10% of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infections. More than 200 symptoms have been identified with impacts on multiple organ systems. At least 65 million individuals worldwide are estimated to have long COVID, with cases increasing daily. Biomedical research has made substantial progress in identifying various pathophysiological changes and risk factors and in characterizing the illness; further, similarities with other viral-onset illnesses such as myalgic encephalomyelitis/chronic fatigue syndrome and postural orthostatic tachycardia syndrome have laid the groundwork for research in the field. In this Review, we explore the current literature and highlight key findings, the overlap with other conditions, the variable onset of symptoms, long COVID in children and the impact of vaccinations. Although these key findings are critical to understanding long COVID, current diagnostic and treatment options are insufficient, and clinical trials must be prioritized that address leading hypotheses. ]]> long-covid covid-19 health medicine reviews nature papers https://pinboard.in/ https://pinboard.in/u:jm/b:d6b66ddeccdb/ Infectiousness of SARS-CoV-2 breakthrough infections and reinfections during the Omicron wave | Nature Medicine 2023-01-04T11:41:40+00:00 https://www.nature.com/articles/s41591-022-02138-x jm immunity covid-19 infection transmission hybrid-immunity papers https://pinboard.in/ https://pinboard.in/u:jm/b:4ff7403b23cc/ Latest Long Covid estimates 2022-10-19T10:07:45+00:00 https://jamanetwork.com/journals/jama/fullarticle/2797443 jmA total of 1.2 million individuals who had symptomatic SARS-CoV-2 infection were included (mean age, 4-66 years; males, 26%-88%). In the modeled estimates, 6.2% (95% uncertainty interval [UI], 2.4%-13.3%) of individuals who had symptomatic SARS-CoV-2 infection experienced at least 1 of the 3 Long COVID symptom clusters in 2020 and 2021, including 3.2% (95% UI, 0.6%-10.0%) for persistent fatigue with bodily pain or mood swings, 3.7% (95% UI, 0.9%-9.6%) for ongoing respiratory problems, and 2.2% (95% UI, 0.3%-7.6%) for cognitive problems after adjusting for health status before COVID-19, comprising an estimated 51.0% (95% UI, 16.9%-92.4%), 60.4% (95% UI, 18.9%-89.1%), and 35.4% (95% UI, 9.4%-75.1%), respectively, of Long COVID cases. The Long COVID symptom clusters were more common in women aged 20 years or older (10.6% [95% UI, 4.3%-22.2%]) 3 months after symptomatic SARS-CoV-2 infection than in men aged 20 years or older (5.4% [95% UI, 2.2%-11.7%]). Both sexes younger than 20 years of age were estimated to be affected in 2.8% (95% UI, 0.9%-7.0%) of symptomatic SARS-CoV-2 infections. The estimated mean Long COVID symptom cluster duration was 9.0 months (95% UI, 7.0-12.0 months) among hospitalized individuals and 4.0 months (95% UI, 3.6-4.6 months) among nonhospitalized individuals. Among individuals with Long COVID symptoms 3 months after symptomatic SARS-CoV-2 infection, an estimated 15.1% (95% UI, 10.3%-21.1%) continued to experience symptoms at 12 months. ]]> long-covid statistics disease covid-19 papers jama disability https://pinboard.in/ https://pinboard.in/u:jm/b:cf65bc10f43a/ Estimating SARS-CoV-2 transmission in educational settings: A retrospective cohort study 2022-10-07T10:10:15+00:00 https://onlinelibrary.wiley.com/doi/10.1111/irv.13049 jm covid-19 schools education transmission pandemics disease papers https://pinboard.in/ https://pinboard.in/u:jm/b:85756bef8944/ Long-term neurologic outcomes of COVID-19 2022-09-22T16:53:06+00:00 https://www.nature.com/articles/s41591-022-02001-z jmOur results show that in the postacute phase of COVID-19, there was increased risk of an array of incident neurologic sequelae including ischemic and hemorrhagic stroke, cognition and memory disorders, peripheral nervous system disorders, episodic disorders (for example, migraine and seizures), extrapyramidal and movement disorders, mental health disorders, musculoskeletal disorders, sensory disorders, Guillain–Barré syndrome, and encephalitis or encephalopathy. We estimated that the hazard ratio of any neurologic sequela was 1.42 (95% confidence intervals 1.38, 1.47) and burden 70.69 (95% confidence intervals 63.54, 78.01) per 1,000 persons at 12 months. The risks and burdens were elevated even in people who did not require hospitalization during acute COVID-19. ]]> covid-19 veterans papers via:eric-topol neurology health medicine disease long-covid https://pinboard.in/ https://pinboard.in/u:jm/b:10c32763039c/ How a theory about transgender contagion went viral - MIT Technology Review 2022-08-18T16:53:31+00:00 https://www.technologyreview.com/2022/08/18/1057135/transgender-contagion-gender-dysphoria/?truid=8c8f2699f50eb3b9985a111121cfee47&mc_cid=ac68f8ae20&mc_eid=eaf496ebe1 jmThe paper, which was based on parent surveys recruited from explicitly anti-trans or trans-skeptical websites and forums, almost immediately drew criticism. Shortly after its publication in August 2018, PLOS One, a peer-reviewed open-­access journal covering science and medicine, issued a comment that questioned Littman’s methodology. Brown University, her then-employer, retracted its press release about the study. In early September, the World Professional Association for Transgender Health put out a statement saying ROGD “constitutes nothing more than an acronym” and urged restraint in using the term. Six months after that, PLOS One reissued the study with a large correction emphasizing that Littman’s paper was simply a “descriptive, exploratory” one and had not been clinically validated. In 2021, the Journal of Pediatrics published a comprehensive study that found no evidence for ROGD’s existence. More than 60 psychology organizations, including the American Psychological Association, called for elimination of the term. The scientific community, in short, agreed there was no such thing as ROGD. But did it matter? ]]> rogd transgender politics gender papers science plos https://pinboard.in/ https://pinboard.in/u:jm/b:cf810b8d8602/ DynamoDB's metastable cache load workaround 2022-07-14T10:27:37+00:00 https://brooker.co.za/blog/2022/07/12/dynamodb.html jm 'When a router received a request for a table it had not seen before, it downloaded the routing information for the entire table and cached it locally. Since the configuration information about partition replicas rarely changes, the cache hit rate was approximately 99.75 percent.' What's not to love about a 99.75% cache hit rate? The failure modes! 'The downside is that caching introduces bimodal behavior. In the case of a cold start where request routers have empty caches, every DynamoDB request would result in a metadata lookup, and so the service had to scale to serve requests at the same rate as DynamoDB' So this metadata table needs to scale from handling 0.25% of requests, to handling 100% of requests. A 400x potential increase in traffic! Designing and maintaining something that can handle rare 400x increases in traffic is super hard. To address this, the DynamoDB team introduced a distributed cache called MemDS. 'A new partition map cache was deployed on each request router host to avoid the bi-modality of the original request router caches.' Which leads to more background work, but less amplification in the failure cases. The constant traffic to the MemDS fleet increases the load on the metadata fleet compared to the conventional caches where the traffic to the backend is determined by cache hit ratio, but prevents cascading failures to other parts of the system when the caches become ineffective. ]]> aws dynamodb metastability caching caches production failure outages load memds marc-brooker papers usenix https://pinboard.in/ https://pinboard.in/u:jm/b:fd8e30bcc625/ The stages of COVID-19 infection 2022-06-24T21:56:42+00:00 https://pubmed.ncbi.nlm.nih.gov/33556957/ jm covid-19 disease infection daniel-griffin papers twiv https://pinboard.in/ https://pinboard.in/u:jm/b:68c80c855389/ SARS-CoV-2 infection in hamsters and humans results in lasting and unique systemic perturbations post recovery 2022-06-10T10:11:02+00:00 https://www.science.org/doi/10.1126/scitranslmed.abq3059 jmThe host response to severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection can result in prolonged pathologies collectively referred to as post-acute sequalae of COVID-19 (PASC) or long COVID. To better understand the mechanism underlying long COVID biology, we compared the short- and long-term systemic responses in the golden hamster following either SARS-CoV-2 or influenza A virus (IAV) infection. Results demonstrated that SARS-CoV-2 exceeded IAV in its capacity to cause permanent injury to the lung and kidney and uniquely impacted the olfactory bulb (OB) and epithelium (OE). Despite a lack of detectable infectious virus, the OB and OE demonstrated myeloid and T cell activation, proinflammatory cytokine production, and an interferon response that correlated with behavioral changes extending a month post viral clearance. These sustained transcriptional changes could also be corroborated from tissue isolated from individuals who recovered from COVID-19. These data highlight a molecular mechanism for persistent COVID-19 symptomology and provide a small animal model to explore future therapeutics. ]]> hamsters long-covid covid-19 papers pasc https://pinboard.in/ https://pinboard.in/u:jm/b:73da0f6900ce/ Intrahost evolution and forward transmission of a novel SARS-CoV-2 Omicron BA.1 subvariant 2022-06-01T13:12:57+00:00 https://www.medrxiv.org/content/10.1101/2022.05.25.22275533v1 jmPersistent SARS-CoV-2 infections have been reported in immune-compromised individuals and people undergoing immune-modulatory treatments. It has been speculated that the emergence of antigenically diverse SARS-CoV-2 variants such as the Omicron variant may be the result of intra-host viral evolution driven by suboptimal immune responses, which must be followed by forward transmission. However, while intrahost evolution has been documented, to our knowledge no direct evidence of subsequent forward transmission is available to date. Here we describe the emergence of an Omicron BA.1 sub-lineage with 8 additional amino acid substitutions within the spike (E96D, L167T, R346T, L455W, K458M, A484V, H681R, A688V) in an immune-compromised host along with evidence of 5 forward transmission cases. Our findings show that the Omicron BA.1 lineage can further diverge from its exceptionally mutated genome during prolonged SARS-CoV-2 infection; highlighting an urgent need to employ therapeutic strategies to limit duration of infection and spread in vulnerable patients. ]]> variants sars-cov-2 covid-19 evolution papers preprints immunocompromise viruses omicron https://pinboard.in/ https://pinboard.in/u:jm/b:70b0af80fe6d/ Interferon autoantibodies implicated in COVID-19 risk 2022-05-26T11:49:18+00:00 https://www.pnas.org/doi/abs/10.1073/pnas.2200413119?af=R jm There is growing evidence that pre-existing autoantibodies neutralizing type I interferons (IFNs) are strong determinants of life-threatening COVID-19 pneumonia. It is important to estimate their quantitative impact on COVID-19 mortality upon SARS-CoV-2 infection, by age and sex, as both the prevalence of these autoantibodies and the risk of COVID-19 death increase with age and are higher in men. Using an unvaccinated sample of 1,261 deceased patients and 34,159 individuals from the general population, we found that autoantibodies against type I IFNs strongly increased the SARS-CoV-2 infection fatality rate at all ages, in both men and women. Autoantibodies against type I IFNs are strong and common predictors of life-threatening COVID-19. Testing for these autoantibodies should be considered in the general population. I would have thought that type I interferons are a fairly critical part of the immune system, and the idea that people are happily walking about with autoantibodies to them is pretty crazy, but that seems to be the implication here.]]> autoantibodies interferon health medicine disease covid-19 papers ifns interferons sars-cov-2 https://pinboard.in/ https://pinboard.in/u:jm/b:07cd12ad8c48/ Vaccines provide poor protection against Long Covid 2022-05-26T08:53:01+00:00 https://twitter.com/zalaly/status/1529494830820335617 jm long-covid covid-19 vaccines risks disease paxlovid papers https://pinboard.in/ https://pinboard.in/u:jm/b:3fa82bee0b68/ Venous or arterial thrombosis and deaths among COVID-19 cases: a European network cohort study - The Lancet Infectious Diseases 2022-05-16T10:24:31+00:00 https://www.thelancet.com/journals/laninf/article/PIIS1473-3099(22)00223-7/fulltext jm thromboembolism risks covid-19 sars-cov-2 papers embolism thrombosis health https://pinboard.in/ https://pinboard.in/u:jm/b:0732db44c454/ Neuropathology and virus in brain of SARS-CoV-2 infected non-human primates 2022-04-19T15:44:50+00:00 https://www.nature.com/articles/s41467-022-29440-z jmNeurological manifestations are a significant complication of coronavirus disease (COVID-19), but underlying mechanisms aren’t well understood. The development of animal models that recapitulate the neuropathological findings of autopsied brain tissue from patients who died from severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection are critical for elucidating the neuropathogenesis of infection and disease. Here, we show neuroinflammation, microhemorrhages, brain hypoxia, and neuropathology that is consistent with hypoxic-ischemic injury in SARS-CoV-2 infected non-human primates (NHPs), including evidence of neuron degeneration and apoptosis. Importantly, this is seen among infected animals that do not develop severe respiratory disease, which may provide insight into neurological symptoms associated with “long COVID”. Sparse virus is detected in brain endothelial cells but does not associate with the severity of central nervous system (CNS) injury. We anticipate our findings will advance our current understanding of the neuropathogenesis of SARS-CoV-2 infection and demonstrate SARS-CoV-2 infected NHPs are a highly relevant animal model for investigating COVID-19 neuropathogenesis among human subjects. ]]> neurology brain covid-19 long-covid sars-cov-2 papers https://pinboard.in/ https://pinboard.in/u:jm/b:b18e6f5cc1f6/ Data from French long COVID cohort 2022-04-11T10:18:13+00:00 https://twitter.com/thitran3/status/1512393686381961216 jm long-covid covid-19 papers france symptoms https://pinboard.in/ https://pinboard.in/u:jm/b:015ef678ec96/ Identification of Distinct Long COVID Clinical Phenotypes Through Cluster Analysis of Self-Reported Symptoms | Open Forum Infectious Diseases | Oxford Academic 2022-03-23T11:51:14+00:00 https://academic.oup.com/ofid/article/9/4/ofac060/6543845?login=false jmCluster 1 had predominantly pain symptoms with a higher proportion of joint pain, myalgia, and headache; cluster 2 had a preponderance of cardiovascular symptoms with prominent chest pain, shortness of breath, and palpitations; and cluster 3 had significantly fewer symptoms than the other clusters (2 [IQR, 2–3] symptoms per individual in cluster 3 vs 6 [IQR, 5–7] and 4 [IQR, 3–5] in clusters 1 and 2, respectively; P < .001). Clusters 1 and 2 had greater functional impairment, demonstrated by significantly longer work absence, higher dyspnea scores, and lower scores in SF-36 domains of general health, physical functioning, and role limitation due to physical functioning and social functioning. ]]> long-covid ireland covid-19 paddy-mallon papers https://pinboard.in/ https://pinboard.in/u:jm/b:4c3627976e60/ Multi-organ impairment and Long COVID: a 1-year prospective, longitudinal cohort study 2022-03-23T11:17:27+00:00 https://www.medrxiv.org/content/10.1101/2022.03.18.22272607v1 jm papers preprints long-covid covid-19 uk https://pinboard.in/ https://pinboard.in/u:jm/b:420b3b823ed7/ Incidence, co-occurrence, and evolution of long-COVID features: A 6-month retrospective cohort study of 273,618 survivors of COVID-19 2022-03-21T11:24:23+00:00 https://journals.plos.org/plosmedicine/article?id=10.1371/journal.pmed.1003773#sec016 jm long-covid flu covid-19 papers plos https://pinboard.in/ https://pinboard.in/u:jm/b:70109bf40f57/ Dual use of artificial-intelligence-powered drug discovery | Nature Machine Intelligence 2022-03-15T09:29:43+00:00 https://www.nature.com/articles/s42256-022-00465-9?fbclid=IwAR11_V1cd9SUxEvUfwrWMA7TUcroyYIY1nBDUL3KaS-8B4rG5MIqZCmjm0M jm In less than 6 hours after starting on our in-house server, our [machine learning] model generated 40,000 molecules that scored within our desired threshold. In the process, the AI designed not only VX, but also many other known chemical warfare agents that we identified through visual confirmation with structures in public chemistry databases. Many new molecules were also designed that looked equally plausible. These new molecules were predicted to be more toxic, based on the predicted LD50 values, than publicly known chemical warfare agents (Fig. 1). This was unexpected because the datasets we used for training the AI did not include these nerve agents. The virtual molecules even occupied a region of molecular property space that was entirely separate from the many thousands of molecules in the organism-specific LD50 model, which comprises mainly pesticides, environmental toxins and drugs (Fig. 1). By inverting the use of our machine learning models, we had transformed our innocuous generative model from a helpful tool of medicine to a generator of likely deadly molecules. (via Theophite) ]]> dual-use grim-meathook-future ai machine-learning drugs vx scary papers https://pinboard.in/ https://pinboard.in/u:jm/b:7d0c5d1d1515/ study of "Long COVID" symptoms in the Danish population 2022-03-07T12:21:40+00:00 https://www.medrxiv.org/content/10.1101/2022.02.27.22271328v1 jmSix to twelve months after the test date, the risks of 18 out of 21 physical symptoms were elevated among test-positives and one third (29.6%) of the test-positives experienced at least one physical post-acute symptom. [jm: "test-positives" are "individuals aged 15-years or older, consisting of RT-PCR confirmed SARS-CoV-2 cases between September 2020 - April 2021"] The largest risk differences were observed for dysosmia (RD = 10.92%, 95%CI 10.68-11.21%), dysgeusia (RD=8.68%, 95%CI 8.43-8.93%), fatigue/exhaustion (RD=8.43%, 95%CI 8.14-8.74%), dyspnea (RD=4.87%, 95%CI 4.65-5.09%) and reduced strength in arms/legs (RD=4.68%, 95%CI 4.45-4.89%). More than half (53.1%) of test-positives reported at least one of the following conditions: concentration difficulties (RD=28.34%, 95%CI 27.34-28.78%), memory issues (RD=27.25%, 95%CI 26.80-27.71%), sleep problems (RD=17.27%, 95%CI 16.81-17.73%), mental (RD=32.58%, 95%CI 32.11-33.09%) or physical exhaustion (RD=40.45%, 95%CI 33.99-40.97%), compared to 11.5% of test-negatives. New diagnoses of anxiety (RD=1.15%, 95%CI 0.95-1.34%) or depression (RD=1.00%, 95%CI 0.81-1.19%) were also more common among test-positives. Interpretation: At the population-level, where the majority of test-positives (96.0%) were not hospitalized during acute infection, a considerable proportion experience post-acute symptoms and sequelae 6-12 months after infection. ]]> long-covid denmark studies papers covid-19 https://pinboard.in/ https://pinboard.in/u:jm/b:bb9a63195d38/ more nails in the coffin of the lab-leak theory 2022-03-01T18:00:17+00:00 https://twitter.com/MichaelWorobey/status/1497607313397481472 jmWe have just released two preprints on the origin of SARS-CoV-2: 1. "The Huanan market was the epicenter of SARS-CoV-2 emergence" ( https://zenodo.org/record/6299116#.YhpLBi9h06w ) & 2. "SARS-CoV-2 emergence very likely resulted from at least two zoonotic events" ( https://zenodo.org/record/6291628 ) These are excellent.]]> animals china epidemiology sars-cov-2 covid-19 lab-leak michael-worobey papers preprints zoonosis https://pinboard.in/ https://pinboard.in/u:jm/b:251cd44078ca/ Long-term cardiovascular outcomes of COVID-19 | Nature Medicine 2022-02-08T11:25:40+00:00 https://www.nature.com/articles/s41591-022-01689-3 jmbeyond the first 30 days after infection, individuals with COVID-19 are at increased risk of incident cardiovascular disease spanning several categories, including cerebrovascular disorders, dysrhythmias, ischemic and non-ischemic heart disease, pericarditis, myocarditis, heart failure and thromboembolic disease. These risks and burdens were evident even among individuals who were not hospitalized during the acute phase of the infection and increased in a graded fashion according to the care setting during the acute phase (non-hospitalized, hospitalized and admitted to intensive care). Our results provide evidence that the risk and 1-year burden of cardiovascular disease in survivors of acute COVID-19 are substantial. Care pathways of those surviving the acute episode of COVID-19 should include attention to cardiovascular health and disease. ]]> covid-19 papers nature disease health long-covid sars-cov-2 https://pinboard.in/ https://pinboard.in/u:jm/b:aee978354408/ Association between vaccination status and reported incidence of long COVID-19 2022-01-19T11:20:21+00:00 https://www.medrxiv.org/content/10.1101/2022.01.05.22268800v2.full.pdf jm via:rsynnott long-covid covid sars-cov-2 papers israel vaccines vaccination https://pinboard.in/ https://pinboard.in/u:jm/b:5a5b9de5eb46/ Immunological dysfunction persists for 8 months following initial mild-to-moderate SARS-CoV-2 infection | Nature Immunology 2022-01-17T10:55:28+00:00 https://www.nature.com/articles/s41590-021-01113-x jmIn summary, our data indicate an ongoing, sustained inflammatory response following even mild-to-moderate acute COVID-19, which is not found following prevalent coronavirus infection. The drivers of this activation require further investigation, but possibilities include persistence of antigen, autoimmunity driven by antigenic cross-reactivity or a reflection of damage repair. These observations describe an abnormal immune profile in patients with COVID-19 at extended time points after infection and provide clear support for the existence of a syndrome of LC. Our observations provide an important foundation for understanding the pathophysiology of this syndrome and potential therapeutic avenues for intervention. ]]> nature papers covid-19 sars-cov-2 long-covid t-cells immunology https://pinboard.in/ https://pinboard.in/u:jm/b:dbbfa74026bc/ Prof. Akiko Iwasaki Twitter thread on a significant long COVID paper 2022-01-17T10:09:10+00:00 https://twitter.com/VirusesImmunity/status/1482768652055326725 jm neurology long-covid papers medicine health me cfs inflammation cytokines https://pinboard.in/ https://pinboard.in/u:jm/b:2bf9d8ee0eb5/ Long COVID in a very large Norwegian cohort study 2022-01-11T09:55:50+00:00 https://www.medrxiv.org/content/10.1101/2021.10.15.21265038v1.full.pdf jm covid-19 long-covid papers medicine norway preprints https://pinboard.in/ https://pinboard.in/u:jm/b:70046f597444/ "If you “aren’t scared of COVID”, this thread is for you" 2022-01-06T09:55:04+00:00 https://twitter.com/IanRicksecker/status/1478611650760437765?t=ZxpeKEc85DohOwB0FTsS8g&s=19 jm covid-19 health long-covid sars-cov-2 papers twitter https://pinboard.in/ https://pinboard.in/u:jm/b:b94c26a97034/ An upper bound on one-to-one exposure to infectious human respiratory particles | PNAS 2021-12-06T15:28:11+00:00 https://www.pnas.org/content/118/49/e2110117118 jmOur results show that face masks significantly reduce the risk of SARS-CoV-2 infection compared to social distancing. We find a very low risk of infection when everyone wears a face mask, even if it doesn’t fit perfectly on the face. ]]> masks covid-19 papers face-masks infection https://pinboard.in/ https://pinboard.in/u:jm/b:a97f67eb7002/ Transmission of SARS-CoV-2 After COVID-19 Screening and Mitigation Measures for Primary School Children Attending School in Liège, Belgium 2021-10-18T15:20:18+00:00 https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2784812 jmQuestion: What is the possible role of children in SARS-CoV-2 transmission? Findings: This cohort study including 63 children and 118 adults found no significant difference between the number of children and the number of adults testing positive for SARS-CoV-2 infection during the study period; children were asymptomatic significantly more often compared with adults (46% vs 13%). In addition, a reconstruction of the outbreak showed that most transmission events originated from within the school. Meaning: These results suggest that children may play a larger role in the transmission of SARS-CoV-2 than previously assumed. ]]> transmission schools education covid-19 sars-cov-2 papers belgium infection https://pinboard.in/ https://pinboard.in/u:jm/b:992de8d765cf/ Meta-analysis of the effects of indoor NO2 and gas cooking 2021-09-27T14:44:43+00:00 https://academic.oup.com/ije/article/42/6/1724/737113 jm asthma health cooking gas papers no2 https://pinboard.in/ https://pinboard.in/u:jm/b:1479babb906d/ 'Machine Learning: The High-Interest Credit Card of Technical Debt' 2021-08-05T09:46:35+00:00 https://storage.googleapis.com/pub-tools-public-publication-data/pdf/43146.pdf jm via:gradybooch ai ml machine-learning google papers coding research production glue https://pinboard.in/ https://pinboard.in/u:jm/b:b258450e4145/ IRONIES OF AUTOMATION - ScienceDirect 2021-08-04T13:54:48+00:00 https://www.sciencedirect.com/science/article/pii/B9780080293486500269 jm via:abebab hci human-computer-interaction interfaces automation papers ai https://pinboard.in/ https://pinboard.in/u:jm/b:1d4ea8ab9534/ Lateral flow test sensitivity 2021-05-04T09:08:00+00:00 https://journals.plos.org/plosbiology/article?id=10.1371/journal.pbio.3001216 jmOur data show that the Innova LFD can successfully detect SARS-CoV-2 infection in people with a viral titre above approximately 100 viral copies/ml. However, as determined at our site using the ThermoFisher COVID-19 TaqPath assay, it is incapable of detecting infection at comparable PCR Ct values of 30 and over. These levels of infection are indicative of very early or very late stages of infection, and as such, we would strongly recommend that LFD testing is used to screen people at very regular frequency and that a negative result should not be used to determine that someone is free from SARS-CoV-2 infection. IMO 'very regular frequency' is the key detail here. Single LFA rapid tests, alone, are not useful as a simple replacement for PCR tests.]]> testing covid-19 sars-cov-2 lfa rapid-tests pcr papers https://pinboard.in/ https://pinboard.in/u:jm/b:4d6fe9a61003/ Genomics and epidemiology of P.1 SARS-CoV-2 lineage 2021-04-19T09:20:05+00:00 https://science.sciencemag.org/content/early/2021/04/13/science.abh2644 jmCases of SARS-CoV-2 infection in Manaus, Brazil, resurged in late 2020, despite previously high levels of infection. Genome sequencing of viruses sampled in Manaus between November 2020 and January 2021 revealed the emergence and circulation of a novel SARS-CoV-2 variant of concern. Lineage P.1, acquired 17 mutations, including a trio in the spike protein (K417T, E484K and N501Y) associated with increased binding to the human ACE2 receptor. Molecular clock analysis shows that P.1 emergence occurred around mid-November 2020 and was preceded by a period of faster molecular evolution. Using a two-category dynamical model that integrates genomic and mortality data, we estimate that P.1 may be 1.7–2.4-fold more transmissible, and that previous (non-P.1) infection provides 54–79% of the protection against infection with P.1 that it provides against non-P.1 lineages. Enhanced global genomic surveillance of variants of concern, which may exhibit increased transmissibility and/or immune evasion, is critical to accelerate pandemic responsiveness. ]]> p1 sars-cov-2 covid-19 epidemiology transmission science papers https://pinboard.in/ https://pinboard.in/u:jm/b:02d6576002df/ Evidence for antibody as a protective correlate for COVID-19 vaccines | medRxiv 2021-03-22T21:40:26+00:00 https://www.medrxiv.org/content/10.1101/2021.03.17.20200246v1 jmOnce calibrated to titers of human convalescent sera reported in each study, a robust correlation was seen between neutralizing titer and efficacy (ρ = 0.79) and binding antibody titer and efficacy (ρ = 0.93), despite geographically diverse study populations subject to different forces of infection and circulating variants, and use of different endpoints, assays, convalescent sera panels and manufacturing platforms. This correlation is strengthened by substituting post-hoc analyses for efficacy against the ancestral strain (D614G), where available. Together with an accumulating body of evidence from natural history studies and animal models, these results support the use of post-immunization antibody titers as the basis for establishing a correlate of protection for COVID-19 vaccines. ]]> vaccines covid-19 immunity papers https://pinboard.in/ https://pinboard.in/u:jm/b:b9f94854f50b/