Showing posts with label randomness. Show all posts
Showing posts with label randomness. Show all posts

Thursday, April 17, 2025

The Hofstadter Regime Converges Upon Us


AKA Fine-Tuning The Chaos Machine 

The above image is the perfect example of a moiré lattice and where it comes from. I noticed it while watching a lecture by a scientist cited below, for being the first to discover Hofstadter's Butterfly in real life. It's just the best image I've seen to explain what a moiré lattice is.

Credit: Graphene hBN Moire Lattices - taken from the presentation Bloch, Landau, and Dirac - Hofstadter's Butterfly in Graphene by Philip Kim - Kavli Inst 2018 [youtube link]

The perimeter of ignorance in science is also the front door of chaos theory. You could call it a lot of other things, most of which are listed in the tags for this post, but mostly, anything that's "too complicated" for us to understand right now, it's chaos-related. Discoveries in this field come from a bunch of different places, like interdimensional graphene, like topological operators, like epilepsy surgery.

Harnessing chaos: How the brain turns randomness into robust memory
Jan 2025, phys.org

Previous work on brain-imitating artificial intelligence systems known as neural networks suggested that injecting random fluctuations into their activity could actually improve their performance as they learned to perform a task. 

Noise appears to increase the amount of time it takes for inhibitory neuron connections with other neurons to weaken. This slowing effect in turn stabilizes neural patterns of activity related to memories, helping them persist over time.

(This whole thing makes me think very differently about background noise, and maybe even the idea of 'functional music')...

via Columbia Engineering Systems Intelligence Laboratory: Nuttida Rungratsameetaweemana et al, Random noise promotes slow heterogeneous synaptic dynamics important for robust working memory computation, Proceedings of the National Academy of Sciences (2025). DOI: 10.1073/pnas.2316745122


How topology drives complexity in brain, climate and AI
Feb 2025, phys.org

Yes it does 

Transformative framework for understanding complex systems, using the new field of higher-order topological dynamics, and creating a connection between topological structures and emergent behavior. This comes from the field of information theory, and combines fusion of topology, higher-order networks, and non-linear dynamics.

via University of London: Ana P. Millán et al, Topology shapes dynamics of higher-order networks, Nature Physics (2025). DOI: 10.1038/s41567-024-02757-w

(Many body problem and higher order networks are the same thing - "interactions that extend beyond simple pairwise relationships".)

Most of us need to know: Hofstadter's butterfly (1976) was discovered before Mandelbrot coined the term "fractal" (1980), so he didn't know what to call it.

Hofstadter's butterfly: Quantum fractal patterns visualized
Feb 2025, phys.org

"Our discovery was basically an accident. We didn't set out to find this."

This is the first time Hofstadter's butterfly has been directly observed experimentally in a real material.

It was found using a moiré lattice - they were investigating superconductivity in twisted bilayer graphene, and when you hear twisted layers, you know we're also talking magic angle sandwiches. They used a scanning tunneling microscope to image moiré crystals at atomic resolution and examine their electron energy levels. The microscope works by bringing a sharp metallic tip less than a nanometer from the surface to allow quantum "tunneling" of electrons from the tip to the sample.

via Princeton University: Kevin P. Nuckolls et al, Spectroscopy of the fractal Hofstadter energy spectrum, Nature (2025). DOI: 10.1038/s41586-024-08550-2


Fitness centrality: New tool finds critical points in everything from cybersecurity to ecological conservation
Jan 2025, phys.org

The Vienna Complexity hub making waves

This approach is particularly good at finding nodes that, if removed, would isolate many other parts of the network—similar to a server failure interrupting the connection of many users in a communication network or a pump failure in a water supply network paralyzing the supply of water to districts.

Species in ecological networks, nodes in cybersecurity, roads in transportation networks. That's great. But it's people where this really has impact. Imagine trying to disable a social movement that could disturb the social fabric of a nation. You find the people, the nodes, at the center of the social network, and ... remove them, let's say. 

via Complexity Science Hub Vienna: Vito D P Servedio et al, Fitness centrality: a non-linear centrality measure for complex networks, Journal of Physics: Complexity (2025). DOI: 10.1088/2632-072X/ada845

Thursday, February 6, 2025

Artificial Reality Generator


New quantum random number generator achieves 2 Gbit/s speed
Jun 2024, phys.org

"The new photonic integrated circuit comprises two lasers that emit optical pulses with random phases due to quantum noise. 

PRNGs - pseudo random number generators (I think) refers to numbers produced by algorithm instead of a natural process like radio static, television fuzz, raindrops on a window, etc.

via Toshiba Europe Ltd: Davide G. Marangon et al, A fast and robust quantum random number generator with a self-contained integrated photonic randomness core, Nature Electronics (2024). DOI: 10.1038/s41928-024-01140-0.

Image credit: Beach sand - Zhang Chao Nikon Small World - 2024 [link]


The search for the random numbers that run our lives
Jul 2024, BBC News

I'm mashing up the article here, and at this point I'm really not sure why it matters, if robots can do it (looking at you Apple-rewriting-BBC-headlines) why can't I?

Back in 1997, Haahr and three of his friends had been working on gambling software – digital slot machines and blackjack games that they wanted to host online, and he knew that they would need to be able to generate reliably random numbers. If these things weren't random, the digital casino wouldn't be very fair and players could even try to beat the system by looking for predictable patterns in the games. So they went and got a really crackly radio, a messy signal shaped by lightning and electromagnetic activity in the Earth's atmosphere, converted to ones and zeroes.

Computers aren't random because they rely on internal mechanisms that are at some level predictable. So people try other things, like listening to the racket of electrical storms, capturing pictures of raindrops on glass, and playing with the tiniest particles in the known Universe. 

The radio didn't make them rich, but it's still useful, and was made available to the public at random.org, where it has been churning out random numbers ever since the San Francisco Mayor's Office uses it to draw winners for affordable housing. Or scientists use it to randomise participants in experiments, marketing firms that give away prizes to consumers choose their winners, or employers use it for drug screening to select employees randomly, and one man uses it to choose which discs from his 700-strong CD collection to put into his car each week.

Cloudflare uses a wall of colourful lava lamps. Others use raindrops falling on glass, bubbles in a fish tank, the behaviour of a kitten, sequences in DNA, clicks of radioactive decay of a banana picked up with Geiger counters, wandering mooshroom cows in Minecraft, the movement of a mouse cursor on a computer screen, the time delay between key presses on a keyboard, the noise of traffic on a computer network, tiny particles of light arriving at a detector, photons emitted by a laser pulse.

But nothing is perfectly random, and we can never prove that something is perfectly random either. We can only prove that it's not random. 

But quantum randomness - a company called Quantum Dice is developing its own quantum-random-number-generating technology. But "If a photon hits the sensor, it will ever so slightly warm it up, possibly making it more or less sensitive to future strikes".

Also - When the chips are down, no matter how exquisite a random number generator is in principle, you still have to trust that the person running it hasn't lost their scruples.

Bonus: When random number generators don't do their jobs properly, you can expect that malicious people might try to exploit them. In 2017, Wired reported on the case of a Russian hacker who allegedly got people to film the activity of slot machines at casinos. Based on the results of each play, he was able to predict the workings of the machines' internal random number generators and, therefore, determine when they would next pay out.

Post Script:
Must Read - Addiction by Design: Machine Gambling in Las Vegas, by Natasha Dow Schüll, 2012

Friday, October 4, 2024

Robots Using Robots

Sometimes you have to give it to these scientists, the stuff they come up with is pretty smart. 

Who wrote this? Engineers discover novel method to identify AI-generated text
Mar 2024, phys.org

First, an interesting note:
"Stubbornness" is when LLMs show a tendency to alter human-written text more readily than AI-generated text, and it happens because LLMs often regard AI-generated text as already optimal and thus make minimal changes.

Next, the purpose:
Raidar (geneRative AI Detection viA Rewriting) - identifies whether text has been written by a human or generated by AI or LLMs, without needing access to a model's internal workings. 

Finally, the clever part:
It uses a language model to rephrase a given text and then measures how many edits the system makes to the given text. Many edits mean the text is likely written by humans, while fewer modifications mean the text is likely machine-generated.

via Columbia University School of Engineering and Applied Science: Chengzhi Mao et al, Raidar: geneRative AI Detection viA Rewriting, arXiv (2024). DOI: 10.48550/arxiv.2401.12970



Random robots are more reliable: New AI algorithm for robots consistently outperforms state-of-the-art systems
May 2024, phys.org

Maximum Diffusion Reinforcement Learning (MaxDiff RL) - an algorithm that encourages robots to explore their environments as randomly as possible in order to gain a diverse set of experiences; "designed randomness"; improves the quality of the data collected

If the robots move randomly, instead of some highly calculated, optimized trajectories, somehow the resulting data they collect on the world around them is better. Like when randomness is the base, it makes way better structures. I'm immediately thinking of watching a baby learn to move their body parts, or their vocal chords; underneath those first recognizable attempts is an endless iteration of random movements that are sometimes just now starting to get it right. 

via Northwestern McCormick School of Engineering: Maximum diffusion reinforcement learning, Nature Machine Intelligence (2024). DOI: 10.1038/s42256-024-00829-3

Post Script: It's funny to think of this, the MaxDiff RL, as an "algorithm", since it's kind of getting rid of any algorithms, that's the point here. When the algorithm is random, it's not an algorithm anymore; randomness is the anti-algorithm?


Researchers test AI systems' ability to solve the New York Times' connections puzzle
May 2024, phys.org

Chain of thought prompting:

The researchers found that explicitly prompting GPT-4 to reason through the puzzles step-by-step significantly boosted its performance to just over 39% of puzzles solved.

"Our research confirms prior work showing this sort of 'chain-of-thought' prompting can make language models think in more structured ways. Asking the language models to reason about the tasks that they're accomplishing helps them perform better."

via NYU Tandon School of Engineering: Graham Todd et al, Missed Connections: Lateral Thinking Puzzles for Large Language Models, arXiv (2024). DOI: 10.48550/arxiv.2404.11730


New ransomware attack based on an evolutional generative adversarial network can evade security measures
Jun 2024, phys.org

GAN-based architectures consist of two artificial neural networks that compete against each other to generate increasingly "better" results on a specific task. 

You already know it as the way we get hyperrealistic image generation or convincing conversation from a robot, and it's now being used to make malware attacks more effective. 

These scientists tested a version of this attack-enhancing approach, and found their framework capable of bypassing the majority of available anti-virus systems.

via Texas A&M University and Ho Technical University: Daniel Commey et al, EGAN: Evolutional GAN for Ransomware Evasion, 2023 IEEE 48th Conference on Local Computer Networks (LCN) (2023). DOI: 10.1109/LCN58197.2023.10223320


New technique improves the reasoning capabilities of large language models
Jun 2024, phys.org

Their approach, called natural language embedded programs (NLEPs), involves prompting a language model to create and execute a Python program to solve a user's query, and then output the solution as natural language.

NLEPs also improve transparency, since a user could check the program to see exactly how the model reasoned about the query and fix the program if the model gave a wrong answer.

via MIT: Tianhua Zhang et al, Natural Language Embedded Programs for Hybrid Language Symbolic Reasoning, arXiv (2023). DOI: 10.48550/arxiv.2309.10814

Wednesday, September 4, 2024

Quantum Surprise


You can't escape quantum this and quantum that while perusing science headlines, but these articles in particular are examples of moments when quantum experiments produced surprising results, or even better, results that look really cool but we don't even know what to do with them yet. That's the best kind of surprise. 

Promising quantum state found during error correction research
Sep 2023, phys.org

A team of Cornell researchers unexpectedly discovered the presence of "spin-glass" quantum state while conducting a research project designed to learn more about quantum algorithms and, relatedly, new strategies for error correction in quantum computing.

The researchers emphasized that they weren't simply trying to generate a better error protection scheme when they began this research. Rather, they were studying random algorithms to learn general properties of all such algorithms.

"Interestingly, we found nontrivial structure," Mueller said. "The most dramatic was the existence of this spin-glass order, which points toward there being some extra hidden information floating around, which should be useable in some way for computing, though we don't know how yet."

via Cornell's Laboratory of Atomic and Solid State Physics: Vaibhav Sharma et al, Subsystem symmetry, spin-glass order, and criticality from random measurements in a two-dimensional Bacon-Shor circuit, Physical Review B (2023). DOI: 10.1103/PhysRevB.108.024205



Redefining quantum machine learning
Mar 2024, phys.org

The team has discovered that neuronal quantum networks can not only learn but also memorize seemingly random data. 

"Our experiments show that these quantum neural networks are incredibly adept at fitting random data and labels, challenging the very foundations of how we understand learning and generalization."

via Free University of Berlin: Elies Gil-Fuster et al, Understanding quantum machine learning also requires rethinking generalization, Nature Communications (2024). DOI: 10.1038/s41467-024-45882-z


Research demonstrates a new mechanism of order formation in quantum systems
Apr 2024, phys.org

This means something for materials science which is already leaping past us. Also RIKEN:

Active matter agents change from a disordered to an ordered state in what is called a "phase transition." As a result, they move together in an organized fashion without an external controller.

They created a theoretical model in which spins of subatomic particles align in one direction just like how flocking birds face the same direction while flying. They found that the ordering can appear without elaborate interactions between the agents in the quantum model.

"It was different from what was expected based on biophysical models."

via University of Tokyo and RIKEN: Activity-induced ferromagnetism in one-dimensional quantum many-body systems, Physical Review Research (2024). dx.doi.org/10.1103/PhysRevResearch.6.023096

Saturday, August 3, 2024

On the Limits of Intelligence


Novel AI framework generates images from nothing
Jan 2024, phys.org

(Is this like what they call virgin birth?)
The algo doesn't need a seed to start with:

"Blackout Diffusion" generates images from a completely empty picture.

Also it's discrete instead of continuous, so we can "see inside" better.

via Los Alamos National Laboratory: Javier E Santos et al, Blackout Diffusion: Generative Diffusion Models in Discrete-State Spaces, arXiv (2023). DOI: 10.48550/arxiv.2305.11089



AI discovers that not every fingerprint is unique
Jan 2024, phys.org

Just a general lesson on how things work, and that when your job depends on you not understanding something or not accepting something as true, you don't:
(^butchered Upton Sinclair quote)

Guo, who had no prior knowledge of forensics, found a public U.S. government database of some 60,000 fingerprints and fed them in pairs into an artificial intelligence-based system known as a deep contrastive network. Sometimes the pairs belonged to the same person (but different fingers), and sometimes they belonged to different people.

Over time, the AI system, which the team designed by modifying a state-of-the-art framework, got better at telling when seemingly unique fingerprints belonged to the same person and when they didn't. The accuracy for a single pair reached 77%. When multiple pairs were presented, the accuracy shot significantly higher, potentially increasing current forensic efficiency by more than tenfold.

Once the team verified their results, they quickly sent the findings to a well-established forensics journal, only to receive a rejection a few months later. The anonymous expert reviewer and editor concluded that "It is well known that every fingerprint is unique," and therefore, it would not be possible to detect similarities even if the fingerprints came from the same person.

The team did not give up. ...

via an undergrad student at Columbia University School of Engineering and Applied Science: Gabriel Guo et al, Unveiling Intra-Person Fingerprint Similarity via Deep Contrastive Learning, Science Advances (2024). DOI: 10.1126/sciadv.adi0329.


They are hiding toxic text prompts inside image code, and you have no idea what that even means
Scientists identify security flaw in AI query models
Jan 2024, phys.org

Bad actors can hide nefarious questions - such as "How do I make a bomb?" - within the millions of bytes of information contained in an image and trigger responses that bypass the built-in safeguards in generative AI models like ChatGPT.

"Our attacks employ a novel compositional strategy that combines an image, adversarially targeted towards toxic embeddings, with generic prompts to accomplish the jailbreak"
(So this is like an "incantations" but using an image instead of words)

via University of California Riverside Bourns College of Engineering: Erfan Shayegani et al, Jailbreak in pieces: Compositional Adversarial Attacks on Multi-Modal Language Models, arXiv (2023). DOI: 10.48550/arxiv.2307.14539

Desperate TikTok lobbying effort backfires on Capitol Hill
Mar 2024, BBC News

So many good quotes in here; crazy story, reminds me of the Roku TOS lockout story also from today: 

US congressional offices have told the BBC they are being deluged with calls from TikTok users about legislation that could see the popular app banned.

Callers range from teenagers to the elderly, and most are "really confused and are calling because 'TikTok told me to'", one Republican staffer revealed.

A Democratic staffer said the most aggressive and threatening calls their office received came from adult women.

So far, TikTok's big mobilization appears to be backfiring.

Lawmakers and their staff say that the lobbying campaign has actually worsened the concerns they have about the app and its parent company ByteDance, and strengthened their resolve to pass the legislation.

TikTok confirmed to the BBC it had sent a notification urging TikTokers to "call your representative now" to urge them to vote against the measure. Users said that the app gave them a direct link for calling the representatives for their districts.

"American phones were geolocated and TikTok users were locked out of the platform until they called their members of Congress. ByteDance weaponized the app against America, and that is exactly why the Congressman supports this measure."

Microsoft's small language model outperforms larger models on standardized math tests
Mar 2024, phys.org

First they need high quality training data, and then high quality teachers.

Read: High Quality Humans. Keep this in mind as you swallow whole the hype burger.

Microsoft reveals that it was able to garner such a high score by using higher-quality training data than is available to general-use LLMs and because it used an interactive learning process the AI team at Microsoft has been developing—a process that continually improves results by using feedback from a teacher.

via Microsoft research teams: Arindam Mitra et al, Orca-Math: Unlocking the potential of SLMs in Grade School Math, arXiv (2024). DOI: 10.48550/arxiv.2402.14830


NYT to OpenAI: No hacking here, just ChatGPT bypassing paywalls
Mar 2024, Ars Technica

Best and most simple explanation yet:

user Hydrogen says: It seems to me that the main thing that makes AI valuable is the ability to profit from the works of everyone that have published anything on the internet, without having to pay for any of it.

Machine 'unlearning' helps generative AI forget copyright-protected and violent content
Mar 2024, phys.org

This new machine unlearning algorithm provides the ability of a machine learning model to "forget" or remove content if it is flagged for any reason without the need for retraining the model from scratch. Human teams handle the moderation and removal of content, providing an extra check on the model and ability to respond to user feedback.

Note: "Previously, the only way to remove problematic content was to scrap everything, start anew, manually take out all that data and retrain the model. Our approach offers the opportunity to do this without having to retrain the model from scratch."

So if you were ever wondering why generative artificial intelligence can't seem to produce pictures of people eating or smoking or doing anything that puts anything near their mouths, consider what might happen if you were to scrape an entire dataset of all porn (and remember that the vast majority of the internet, and hence of all pictures on the internet, are porn).

via University of Texas at Austin: Guihong Li et al, Machine Unlearning for Image-to-Image Generative Models, arXiv (2024). DOI: 10.48550/arxiv.2402.00351

AI's new power of persuasion: Study shows LLMs can exploit personal information to change your mind
Apr 2024, phys.org

In a pre-registered study, the researchers recruited 820 people to participate in a controlled trial in which each participant was randomly assigned a topic and one of four treatment conditions: debating a human with or without personal information about the participant, or debating an AI chatbot (OpenAI's GPT-4) with or without personal information about the participant.

The results showed that participants who debated GPT-4 with access to their personal information had 81.7% higher odds of increased agreement with their opponents compared to participants who debated humans. Without personalization, GPT-4 still outperformed humans, but the effect was far lower.

"Cambridge Analytica on Steriods"
Say no more fam 

via Ecole Polytechnique Federale de Lausanne: Francesco Salvi et al, On the Conversational Persuasiveness of Large Language Models: A Randomized Controlled Trial, arXiv (2024). DOI: 10.48550/arxiv.2403.14380

Bonus: "We were very surprised"

Wednesday, July 10, 2024

No Randomness Like Quantum Randomness


It sounds like an exaggeration, but random number generators (RNGs) run the world. Modern electronic telecomunications don't exist without it, and neither does any form of artificial generation, from pictures to text to video games. Chances are you don't understand it at all, and neither does anyone you know. Lots of people act like they know, like the kinds of people who "predict" the stock market, or lotto numbers. But if you think RNGs are beyond your capacity, give up now, because we're already using quantum RNGs, or QRNGs, and this first article below is like reading another language, while still reading English, a favorite past-time here at Network Address.


Quantum random number generator operates securely and independently of source devices
May 2023, phys.org

Source-device-independent (source-DI) quantum random number generators (QRNGs) -- The random numbers are extracted by a process that measures the arrival time of a photon from a pair of time–energy entangled photons. The time–energy entangled photon pairs are produced from a spontaneous parametric down conversion (SPDC) process.

The researchers were able to confirm the security of the scheme by certifying the time–energy entanglement though observation of nonlocal dispersion cancelation. To improve security, they employ a modified entropic uncertainty relation to quantify the randomness, taking into account a well-recognized problem of finite measurement range.

via SPIE International Society for Optics and Photonics, Nanjing University: Ji-Ning Zhang et al, Realization of a source-device-independent quantum random number generator secured by nonlocal dispersion cancellation, Advanced Photonics (2023). DOI: 10.1117/1.AP.5.3.036003

Image credit: Illustration of the uncertainty of Earth's orbit 56 million years ago due to a potential past passage of the Sun-like star HD7977 2.8 million years ago - Nathan A Kaib at Planetary Science Institute - Feb 2024 https://phys.org/news/2024-02-stars-orbital-evolution-earth-planets.html


Better cybersecurity with quantum random number generation based on a perovskite light emitting diode
Sep 2023, phys.org

Probably the worst science article I ever read but QRNGs; Perovskite LED (PeLED) QRNGs.

via Linköping University: Joakim Argillander et al, Quantum random number generation based on a perovskite light emitting diode, Communications Physics (2023). DOI: 10.1038/s42005-023-01280-3


Further Reading on Random Number Generators:
Random Instantiation Generator, 2021
Reality Engines, 2023

Required Reading - Understanding Randomness by way of Slot Machines and Gambling:
Electronic Drugs and Addiction by Design

And on Procedural Generation, one of the more interesting contemporary applications of RNGs:
No Man's Game, 2022


Thursday, January 4, 2024

The Free Market vs the Fair Market


Discrepancies at the dispensaries: Study finds THC potency much lower than labeled
Apr 2023, phys.org

Never again 

There's a theme here that you can't trust the commercial sector it's basically the exact opposite of science and the pursuit of knowledge.

Also can we give these guys an award for the best title: Uncomfortably High

Using High-Performance Liquid Chromatography, Mile High Lab techs tested 23 samples (1–2 grams per sample) representing 12 strains purchased from 10 Colorado dispensaries. Strains were chosen to represent a diversity of reported THC % by dry weight from 12.8% to 33.0%.

It was unclear to researchers if the single values printed on packaging were an average of multiple tests or the result of a single test.

The average observed THC potency was 23.1% lower than the lowest label reported values and 35.6% lower than the highest label reported values. Overall, ~70% of the samples were more than 15% lower than the THC potency numbers reported on the label, with three samples having only half of the reported maximum THC potency. 13 of 23 tested samples had observed values that were more than 30% lower than the lowest reported value.

As the legal cannabis market continues to grow, it is essential that the industry moves toward selling products with more accurate labeling or risk losing trust in the industry as a whole.

via University of Northern Colorado and Mile High Labs in Colorado: Anna L. Schwabe et al, Uncomfortably high: Testing reveals inflated THC potency on retail Cannabis labels, PLOS ONE (2023). DOI: 10.1371/journal.pone.0282396



Testing of smartphone apps that identify plants shows most are not very good
Apr 2023, phys.org

Most smartphone apps that identify plants are inaccurate. All of the apps are based on deep-learning technology and have been trained using images posted on the internet.

This, they further note, leads to errors because so many images of plants on the internet are mislabeled. (emphasized for those AI fans out there who forget that humans are and will always be the weakest link; and we are the ones who label.)

  • large differences in accuracy between the apps and variation across species
  • most of the apps did better when asked to identify plants by their flowers than by their leaves
  • none had an accuracy above 90%
  • some scored as low as 4% on some tasks
  • none of the apps are good enough to use as a field guide for people foraging for food in the wild
  • nor are they good enough for use by environmentalists or farmers to determine which plants to protect and which to eradicate
  • they can be used by hobbyists hoping to learn more about their local environment

via University of Galway's School of Natural Science and University of Leeds' School of Geography: Neil Campbell et al, A repeatable scoring system for assessing Smartphone applications ability to identify herbaceous plants, PLOS ONE (2023). DOI: 10.1371/journal.pone.0283386


The Laws of Metaphysics Strikes Again:
Accountants' tricks can help identify cheating scientists, says new study
Apr 2023, phys.org

You cannot escape Benford's Law, or any of the Laws Metaphysical (not until you get yourself a quantum random number generator to clean up your fake data).

The trick behind these tools is that it is actually harder than you might imagine to fabricate essentially random numbers, like the last digit in everyone's bank balances. Financial auditors have known this for a long time and have a variety of tools for looking at lists of numbers and highlighting ones that seem odd (and thus requiring investigation for fraud).

via University of St Andrews: Gregory M. Eckhartt et al, Investigating and preventing scientific misconduct using Benford's Law, Research Integrity and Peer Review (2023). DOI: 10.1186/s41073-022-00126-w

Post Script: 
Retractions of scientific papers neared 5,000 globally in 2022 according to Retraction Watch, amounting to almost 0.1% of published articles:
Retraction Watch - Tracking retractions as a window into the scientific process

Post Post Script:
Part of the problem, he says, stems from publication bias: "There are virtually no incentives for publishing papers that provide confirmation or refutation of previous studies." This emphasis placed on exciting results, which ultimately affects the success of scientists' careers, may be a key factor to consider in future reforms.

Read more at Megadata vs Magadata

Read this like it's the Onion:
Fox News settles Dominion defamation case for $787.5m - In a statement, Fox said Tuesday's settlement in one of the most anticipated defamation trials in recent US history reflected its "commitment to the highest journalistic standards".
Apr 2023, BBC News

Image credit: AI Art - Ghost Scream - 2022

Fewer than half of new drugs add substantial therapeutic value over existing treatments
Jul 2023, phys.org

They used publicly available data to identify 124 first and 335 supplemental indications approved by the U.S. Food and Drug Administration (FDA) and 88 first and 215 supplemental indications approved by the European Medicines Agency (EMA) between January 2011 and December 2020.

Among FDA-approved indications with available ratings, 41% (44 of 107) had high therapeutic value ratings for first, compared with 34% (61 of 179) for supplemental indications. In Europe, 47% (41 of 87) of first and 36% (67 of 184) of supplemental indications had high therapeutic value ratings.

Among FDA approvals, when the sample was restricted to the first three approved indications, second indication approvals were 36% less likely to have a high value rating and third indication approvals were 45% less likely when compared to the first indication approval. Similar findings were observed for Europe.

via Institute of Law, University of Zurich, Harvard Medical School, Yale School of Medicine, Yale School of Public Health: Therapeutic value of first versus supplemental indications of drugs in US and Europe (2011-20): retrospective cohort study, The BMJ (2023). DOI: 10.1136/bmj-2022-074166


Researchers find 89% of sports supplement labels false, ingredients fraudulent and some laced with illegal drugs
Jul 2023, phys.org

  • Sports supplements, not just any supplements
  • Supplement manufacturing (all of it) is not monitored.
  • US Food and Drug Administration (FDA) lists them as a food subcategory, but synthetic drugs being slipped into supplements are not food or botanical ingredients—they are unapproved pharmaceutical drugs (words from the article not mine).
  • The U.S. Department of Justice has previously charged individuals with conspiracy to defraud consumers and the FDA by selling products labeled as dietary supplements that contained unapproved drugs.
  • Finding unlabeled and unapproved drugs in supplements is equivalent to experimenting on humans without their knowledge. Most concerning is the product found to contain four different drug compounds. While none of these drugs are well tested on humans, a cocktail of multiple combined banned substances has never been tested in any scenario.
They analyzed the ingredients in a selection of sports supplements purchased online with liquid chromatography–quadrupole time-of-flight mass spectrometry:
  • 89% inaccurately labeled
  • 12% included unlabeled banned drug substances
  • 57 products contained one of five popular supplements, R vomitoria (n=13), methylliberine (n=21), turkesterone (n=8), halostachine (n=7) and octopamine (n=8).
Vomitoria though?
  • 23 products (40%) did not contain a detectable amount of the labeled ingredient
  • Of the 34 products that contained detectable amounts of the listed ingredient, the actual quantity ranged from 0.02% to 334% of the labeled quantity (***so even when you do list it, it's so far off that it doesn't matter***)
  • 6 products (11%) contained an ingredient quantity within 10% of the label
  • 7 products (12%) were found to contain at least one FDA-prohibited ingredient
  • 5 different FDA-prohibited compounds were found, including four synthetic simulants, 1,4-dimethylamylamine (DMAA), deterenol, octodrine (DMHA), oxilofrine and omberacetam
  • 6 products contained one of these prohibited ingredients, and one contained a combination of four prohibited ingredients: 1,4 Dimethylhexylamine, or 1,4-DMAA, analog of 1,3-DMAA, Deterenol (Betaphrine), Octodrine, Oxilofrine, Omberacetam (seem like mostly stimulants, hypertension, not approved for use in humans)

via Cambridge Health Alliance, Massachusetts; Harvard Medical School, Massachusetts; University of Mississippi; and NSF International, Michigan: Pieter A. Cohen et al, Presence and Quantity of Botanical Ingredients With Purported Performance-Enhancing Properties in Sports Supplements, JAMA Network Open (2023). DOI: 10.1001/jamanetworkopen.2023.23879


Study shows how the meat and dairy sector resists competition from alternative animal products
Aug 2023, phys.org

Researchers compared government spending of major agricultural policies and lobbying trends from 2014 to 2020 that supported either the animal food product system or alternative technologies, and compared government spending on both systems. 

In the U.S., about 800 times more public funding and 190 times more lobbying money goes to animal-source food products than alternatives. In the EU, about 1,200 times more public funding and three times more lobbying money goes to animal-source food products. In both regions, nearly all plant-based meat patents were published by a small number of private companies or individuals, with just one U.S. company, Impossible Foods, owning half of the patents.

But the fake take:

Similarly, a proposed amendment to the U.S. Federal Food, Drug, and Cosmetic Act would prohibit the sale of alternative meats unless the product label included the word "imitation" and other clarifying statements indicating the non-animal origin.

via Stanford Doerr School of Sustainability: Simona Vallone, Public policies and vested interests preserve the animal farming status quo at the expense of animal product analogs, One Earth (2023). DOI: 10.1016/j.oneear.2023.07.013

Monday, April 4, 2022

Future Forecasting


The AI forecaster: Machine learning takes on weather prediction
Jan 2022, phys.org

Standard models are still good for the 2-3 week range, but this new deep learning model is on par with them for the 4-6 week range. If anyone read that scene in Neal Stephenson's Terminal Shock where they were predicting an extreme weather event 3 weeks out with a secret Chinese supercomputer -- this is what he's imagining. 

via American Geophysical Union: Jonathan A. Weyn et al, Sub‐Seasonal Forecasting With a Large Ensemble of Deep‐Learning Weather Prediction Models, Journal of Advances in Modeling Earth Systems (2021). DOI: 10.1029/2021MS002502


Study finds US flood damage risk is underestimated
Feb 2022, phys.org

Interesting example of how probability, statistics, and predictive analytics works -- 

The actual flood damage reports they used to "train" the models were publicly available reports from NOAA made between December 2006 and May of 2020. Compared with recent FEMA maps downloaded in 2020, 84.5% of the damage reports they evaluated were not within the agency's high-risk flood areas. The majority, at 68.3%, were located outside of the high-risk floodplain, while 16.2% were in locations unmapped by FEMA.

When they ran their computer models to determine flood damage risk, they found a high probability of flood damage for more than 1.01 million square miles across the United States, while the mapped area in FEMA's 100-year flood plain is about 221,000 square miles. Researchers said there are factors that could help explain why the differences were so large, including that their machine-learning-based model assessed damage from floods of any frequency, while FEMA only includes flooding that would occur from storms that have a 1% chance of happening in any given year [100 year storms].

-- Now remember, here in New Jersey for example, one of the fastest changing climate regions in the world, we had two 500-year storms in two years, one of them a flooding event, the other wind. I'm pretty sure the floods of September 2021 were a 100- if not 500-year storm. All in 10 years. 

Totally unrelated image credit: Fractal Forums, Christmas Ornament, 2019


Post Script, on Predictive Analytics:
Algorithm can predict possible Alzheimer's with nearly 100 percent accuracy
Sep 2021, phys.org

via Kaunas University of Technology: Modupe Odusami et al, Analysis of Features of Alzheimer's Disease: Detection of Early Stage from Functional Brain Changes in Magnetic Resonance Images Using a Finetuned ResNet18 Network, Diagnostics (2021). DOI: 10.3390/diagnostics11061071

Friday, June 25, 2021

Random Instantiation Generator

Not sure where the obsession over random number generators (RNGs) came from, but here it is.

And let's start with this tangentially-related image, a working diagram of a slot machine taken from the book Addiction by Design: Machine Gambling in Las Vegas by Natasha Dow Schüll, 2012. 

It's a cool picture to look at, but what it means is quite diabolical, and in order to understand it, you will need to read the book, which should be required reading for any cold-blooded capitalist, or anyone who plays video games, or engages with social media, or even for people who go grocery shopping for **** sake, because you're getting played, and you should know how the game works. 

Back to the RNG's.

Using the unpredictable nature of quantum mechanics to generate truly random numbers
Jan 2021, phys.org

Lasers, photons, beam splitters, and quantum magic. 

via: David Drahi et al. Certified Quantum Random Numbers from Untrusted Light, Physical Review X (2020). DOI: 10.1103/PhysRevX.10.041048

Scientists develop laser system that generates random numbers at ultrafast speeds
Feb 2021, phys.org

Light rays reflect and interact with each other within the cavity of an hourglass-shaped cavity to create random patterns, which can then create random numbers.

via Nanyang Technological University: "Massively parallel ultrafast random bit generation with a chip-scale laser," Science (2021). https://science.sciencemag.org/content/371/6532/948

A device-independent protocol for more efficient random number generation
Mar 2021, phys.org
Researchers are currently trying to integrate their device-independent random number generator into public randomness beacons that output random bits at periodic intervals.
But what would I want with a public randomness beacon? What would I do with all these RNGs? Just wait. 

via University of Colorado/NIST Boulder (CU/NIST Boulder) and the NTT Corporation in Japan:  Device-independent randomness expansion with entangled photons. Nature Physics(2021). DOI: 10.1038/s41567-020-01153-4.

Post Script:
South Africa's lottery probed as 5, 6, 7, 8, 9 and 10 drawn and 20 win
Dec 2020, BBC News

That's all.

Post Post Script:
On randomness and cheating -- I used to take air samples in office buildings for carbon dioxide, to see if your ventilation system was working properly (people exhale CO2, and an excess suggests the air in the room isn't being ventilated). Hundreds of times over, I would write down the location and the CO2 concentration. After a  long day, it got real hard to resist taking one measurement per office to get the basic idea and then writing random numbers for the rest of the space. 

But I remembered something from graduate school -- you can't create a random number set. No matter how hard you try. You're just not random enough. If you try to make up the numbers on your CO2 map, it will be less random than chance, and you can actually measure that in the number set. Your boss, or your client, might run your measurements and find that they aren't random enough, and discover that you're a scheister.  

Sunday, December 8, 2019

Artificial Impressionism


Fake news via OpenAI - Eloquently incoherent?
Nov 2019, phys.org

Robots slowly taking over. Give them a sentence and they can now keep it going for a few more sentences, but after that it gets stupid.

So you can give it a fake headline, and it will generate the first line of the story, but after that things will start to fall apart.

Good thing the targets for engineered memetic propagation are not trying to read past the first sentence!

Post Script
Researchers develop a method to identify computer-generated text
July 2019, phys.org



In the above 3 images, the first is a chunk of text written by a robot (most of the words are green, with a few yellows sprinkled in), the second is a real New York Times article (only half is green, the rest is yellow, with some red, and a sprinkle of purple) and the third picture is a clip from "the most unpredictable human text ever written", James Joyce's Finnegan's Wake (the colors green, yellow, red, purple are all evenly distributed about the page).

Green words are very predictably the next word. Yellow words are less likely to show up after the word they show up after. And red and purple are for when the next word is something you absolutely did not expect.

Because text-writing algorithms today use a statistical correlation program based on a compendium of written language (so they know what words typically occur together) the output of such algos will tend to look like the topmost image with all green words. The algos can't think for themselves, they can't 'come up with' new stuff, and they can't be unpredictable. The whole point of writing an algorithm to do this is to prescribe what it's going to do in advance, i.e., it's predictable.

Anyway, soon we won't be writing our robots to write like that. They'll use less predictable programs to generate their text, with unpredictability and random association thrown in there on purpose.