Monday, October 7, 2024

The Friend Network


Targeting friends to induce social contagion can benefit the world, says new research
May 2024, phys.org

The study evaluated a strategy that exploits the so-called "friendship paradox" of human social networks. That theory suggests that on average, your friends have more friends than you do. As the theory goes, the individuals nominated as friends potentially wield more social influence than those who identify them.

For the study, the researchers utilized the friendship paradox in the delivery of a proven 22-month education package promoting maternal, child, and neonatal health in 176 isolated villages in Honduras.

The researchers found that delivering the intervention to a smaller fraction of households in each village via the friendship targeting strategy led to the same level of behavioral adoption as would have been achieved by treating all the households.

People were either selected randomly within each village to receive the intervention or they were randomly chosen to nominate their friends, who were subsequently picked at random. 

"We found that targeting people's friends for an intervention induced significant social contagion, creating cascades of beneficial health practices to people who didn't receive the intervention." 

For many outcomes, using the friendship-nomination targeting method to reach 20% of households in a village affected outcomes the same as administering the intervention to every household.

Yes, Facebook knows this very well.

via Yale and Temple University: Edoardo M. Airoldi et al, Induction of social contagion for diverse outcomes in structured experiments in isolated villages, Science (2024). DOI: 10.1126/science.adi5147



Study shows relatively low number of superspreaders responsible for large portion of misinformation on Twitter
May 2024, phys.org

10 months of data; 2,397,388 tweets; 448,103 users; parsed by low-credibility information status.

A third of the low-credibility tweets had been posted by people using just 10 accounts, and just 1,000 accounts were responsible for posting approximately 70% of such tweets.

via Indiana University: Matthew R. DeVerna et al, Identifying and characterizing superspreaders of low-credibility content on Twitter, PLOS ONE (2024). DOI: 10.1371/journal.pone.0302201

Saturday, October 5, 2024

You Wouldn't Download a Car

The sound of a thousand digital twin insurance companies being born (to protect your real assets from digital vandalism, of course).

Pokémon Go players are altering public map data to catch rare Pokémon
May 2024, Ars Technica

This isn't graffiti, but it's like graffiti; defacing public property, or property that's not yours ... Is map data public or private property? The underlying data is sometimes public, or other people's private, but not the company that makes the map. Anyway, graffiti is a form of rebellion against a system that takes things away from you, only to give it back to "everyone", and where "everyone" is really just the people left over who can afford to own things. Who owns the data? 


Stack Overflow users sabotage their posts after OpenAI deal
May 2025, Ars Technica

The words "user hostile" have become rampant on the big wordbox we call the internet. In this case, a website that runs a chat forum about computers decided that the millions of hours worth of users putting really complex computer problems into machine-readable text for other users would be really valuable if we instead gave it to one single company who owns a really big reading machine. The users who did all the work here (value = work) decided nah, let's corrupt the data by changing our posts. Not just deleting their posts, but changing the right answers to completely nonsensical answers so wrong they're dangerous. And that is the future of the internet.


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

Thursday, October 3, 2024

Quantum Always


Physicists demonstrate first metro-area quantum computer network in Boston
May 2024, phys.org

Quantums now in the US, the last one was in China:

Using existing Boston-area telecommunication fiber, their photons were deployed over a roughly 22-mile loop through Cambridge, Somerville, Watertown, and Boston, with quantum computers at the nodes.

via Harvard: Mikhail Lukin, Entanglement of nanophotonic quantum memory nodes in a telecom network, Nature (2024). DOI: 10.1038/s41586-024-07252-z.



A place to study qubits shielded from the effects of cosmic rays
Jun 2024, phys.org

QUIET and LOUD - a pair of quantum sensors, one above ground and one under. 

via Fermi National Accelerator Laboratory and the National Quantum Initiative


A framework to construct quantum spherical codes
Jun 2024, phys.org

Photonic quantum coding theory:

All quantum codes require the superposition of something, and Albert and his colleagues realized that it made sense to superimpose well-separated points on a sphere. Their framework builds on a previously proposed method to map electromagnetic signals of any frequency into points on a sphere.

"There is an old and very general technique by the founder of information theory, Claude Shannon, that maps an arbitrary electromagnetic signal of fixed amplitude but of any frequency into a point on the sphere," Albert explained. "This means that efficiently sending classical information using light boils down to packing as many points on the sphere as possible while making sure that noise does not cause them to overlap."

via NIST and University of Maryland: Shubham P. Jain et al, Quantum spherical codes, Nature Physics (2024). DOI: 10.1038/s41567-024-02496-y


Pseudomagic quantum states: A path to quantum supremacy
Jun 2024, phys.org

Don't even bother to understand, just know magic states: 

A stabilizer state is a type of quantum state that can be efficiently simulated on a classical computer, and nonstabilizerness or magic refers to a measure of the non-classical resources possessed by a quantum state.

Pseudomagic quantum states appear to have the properties of nonstabilizer states (complexity and non-classical operations) but are computationally indistinguishable from random quantum states, at least to an observer with limited computational resources.

via Harvard University and Freie Universität Berlin: Andi Gu et al, Pseudomagic Quantum States, Physical Review Letters (2024). DOI: 10.1103/PhysRevLett.132.210602.


Quantum annealer improves understanding of quantum many-body systems
Jun 2024, phys.org 

They used a quantum annealer to model a real-life quantum material and showed that the quantum annealer can directly mirror the microscopic interactions of electrons in the material.

In this study, the scientists investigated the quantum material 1T-TaS2.

"We have placed the system in a non-equilibrium state and observed how the electrons in the solid-state lattice rearrange themselves after a non-equilibrium phase transition, both experimentally and through simulations."

The scientists demonstrated that the quantum annealer's qubit interconnections can directly mirror the microscopic interactions between electrons in a quantum material. Only one single parameter in the quantum annealer must be modified. The outcome aligns closely with the experimental findings.

via Forschungszentrum Jülich Supercomputing Center and D-Wave: Jaka Vodeb et al, Non-equilibrium quantum domain reconfiguration dynamics in a two-dimensional electronic crystal and a quantum annealer, Nature Communications (2024). DOI: 10.1038/s41467-024-49179-z

Wednesday, October 2, 2024

Public Trust in Science


How much trust do people have in different types of scientists?
Apr 2024, phys.org

2,780 participants from the United States were asked about trust in 45 different types of scientists, from agronomists to zoologists. 

Participants were quizzed on how they see scientists with regard to:
  • Competence: how clever and intelligent they consider scientists
  • Assertiveness: how confident and assertive
  • Morality: how just and fair
  • Warmth: how friendly and caring

On a 7-point scale, with 7 being most trusted:
  • political scientists - 3.71
  • economists - 4.28
  • neuroscientists - 5.53
  • and marine biologists  - 5.54
"Nevertheless, one thing is clear: the diversity of scientific fields must be taken into account to more precisely map trust, which is important for understanding how scientific solutions can best find their way to policy."
via University of Amsterdam: Vukašin Gligorić et al, How social evaluations shape trust in 45 types of scientists, PLOS ONE (2024). DOI: 10.1371/journal.pone.0299621


Tuesday, October 1, 2024

Clash of the Titans


When the big dogs fight each other you know it's getting crazy (like how Ford and Blue Cross had a lawsuit prior to the labor negotiations circa 2023).

Beyonce and Adele publisher accuses firms of training AI on songs
May 2024, BBC News


So this is very much an arms race between (pro-profit anti-human) food manufacturers who spend their astronomical profits on research that makes their food addictive, vs pro-human scientists who are trying to make our brains not-addicted to food (albeit then using the for-profit pharma industry). 

Trojan Horse' weight loss drug found to be more effective than available therapies
May 2024, phys.org

"We already know that GLP-1-based drugs can lead to weight loss. The molecule that we have attached to GLP-1 affects the so-called glutamatergic neurotransmitter system, and in fact, other studies with human participants suggest that this family of compounds has significant weight loss potential. What is interesting here is the effect we get when we combine these two compounds into a single drug," Clemmensen says.

via Novo Nordisk Foundation Center for Basic Metabolic Research at the University of Copenhagen: Jonas Petersen et al, GLP-1-directed NMDA receptor antagonism for obesity treatment, Nature (2024). DOI: 10.1038/s41586-024-07419-8