Showing posts with label recursion. Show all posts
Showing posts with label recursion. Show all posts

Wednesday, August 21, 2024

Social Control as Social Service and the Art of Meme Hygiene


Study shows babies learn to imitate others because they themselves are imitated by caregivers
Sep 2023, phys.org

(The whole article is basically a good description of recursion.)

  • Social learning avoids laborious trial and error; the wheel does not have to be reinvented each time.
  • "Children acquire their ability to imitate because they themselves are imitated by their caregivers" 
  • Parents respond to the signals given by the child and reflect and amplify them. A mutual imitation of actions and gestures develops. 
  • "These experiences create connections between what the child feels and does on the one hand and what it sees on the other"
  • "Imitation is the start of the cultural process toward becoming human" 
  • Over the course of generations and millennia, this interplay has led to the cultural evolution of humans"

via Ludwig Maximilian University of Munich:  Samuel Essler et al, The cultural basis of cultural evolution: Longitudinal evidence that infant imitation develops by being imitated, Current Biology (2023). DOI: 10.1016/j.cub.2023.08.084. 

Image credit: Speaking of memetics, imitation, and learning, the image above is an example of a computer trying to be a human - a graphic designer specifically. We can see how, in very subtle ways, computers can't be people, not just yet: AI Art - Ad Poop - 2024


Social bonding gets people on the same wavelength, neural synchronization study suggests
Mar 2024, phys.org

176 three-person groups of human participants wore caps with fNIRS (functional near-infrared spectroscopy) electrodes while they communicated with strangers in a face-to-face triangle. (But it sounds like they communicated via text message, although they were sitting across from each other.)

This writeup is so concise I have to copy the whole thing:

Each group democratically selected a leader, so each group of three ultimately included one leader and two followers. After strategizing together, groups played two economic games designed to test their willingness to make sacrifices to benefit their group (or harm other groups).

Experimenters assigned some triads to go through a bonding session, where they were grouped according to color preferences, given uniforms, and led through an introductory chat session to build familiarity.

Bonded groups spoke more freely and bounced between speakers more frequently and rapidly, relative to groups that didn't experience this bonding session. This bonding effect was stronger between leaders and followers than between two followers.

Neural activity in two brain regions linked to social interaction, the right dorsolateral prefrontal cortex (rDLPFC) and the right temporoparietal junction (rTPJ), aligned between leaders and followers if they had bonded.

The authors state that this neural synchronization suggests that leaders may be anticipating followers' mental states during group decision-making, though they acknowledge that their findings are restricted to East Asian Chinese individuals communicating via text (without non-verbal cues), whose culture emphasizes group cohesion and commitment towards group leaders.

via Beijing Normal University: Ni J, Yang J, Ma Y (2024) Social bonding in groups of humans selectively increases inter-status information exchange and prefrontal neural synchronization. PLoS Biology (2024). DOI: 10.1371/journal.pbio.3002545

Charting brain synchronization patterns during social interactions - Yuto Kurihara from Waseda University - Apr 2024

Charting brain synchronization patterns during social interactions
Apr 2024, phys.org

See the infographic above, really well done.

"Our findings challenge the conventional understanding"

Cooperative interactive tasks between individuals with weak social ties result in more synchronized brain activity compared to individuals with strong ties. 

The participants were given a joint tapping task where they had to tap a mouse button in opposite rhythms. They wore earphones and had to anticipate their partner's movements.

Researchers suggest that the lack of familiarity between strangers requires a more involved process for predicting each other's actions or behaviors in a cooperative task. Consequently, this heightened engagement leads to a more efficient transfer of information between closely connected nodes within the neural network.

via Waseda University: Yuto Kurihara et al, The topology of interpersonal neural network in weak social ties, Scientific Reports (2024). DOI: 10.1038/s41598-024-55495-7


Understanding the spread of behavior: How long-tie connections accelerate the speed of social contagion
Apr 2024, phys.org

This is all getting pretty scary; I used to think this was something far off, but it sure seems we have both the knowledge and the means to do these things, and I wonder how it's being done already, and highly doubt it's not being done already. Too irresistible. 

Initially, researchers thought highly clustered ties that are close together in networks created the perfect environment for the spread of complex behaviors that require significant social reinforcement. However, long ties, which are created through randomly rewired edges that make them "longer," accelerate the spread of social contagions. 

(So this is not about weak vs strong ties, but short vs long.)

Having a small probability of adoption below the contagion threshold is enough to ensure that random rewiring accelerates the spread of these contagions.

"Further work could study such strategies for seeding complex behaviors"

This research suggests those wanting to achieve fast, total spread would benefit from implementing intervention points across network neighborhoods with long-tie connections to other network regions

via University of Pittsburgh Swanson School of Engineering, Sloan School of Management at MIT: Dean Eckles et al, Long ties accelerate noisy threshold-based contagions, Nature Human Behaviour (2024). DOI: 10.1038/s41562-024-01865-0

Wednesday, January 10, 2024

Something Cannibalistic


Toothpaste containing synthetic tooth minerals can prevent cavities as effectively as fluoride: Clinical trial
Jul 2023, phys.org

Tooth-toothpaste

via Department of Integrated Dentistry, Poznan University of Medical Sciences, Poznan, Poland: Caries-preventing effect of a hydroxyapatite-toothpaste in adults: A 18 months double-blinded randomized clinical trial, Frontiers in Public Health (2023). DOI: 10.3389/fpubh.2023.1199728.


Transparent mouse - Helmholtz Munich - 2023

Transparent mouse could improve cancer drug tests
Jul 2023, BBC News

Prof Ali Ertürk of the Helmholtz Munich research centre worked out how to make a dead mouse transparent in 2018.

Thursday, January 4, 2024

Not-bots, Fauxbots, Fleshbots and Semi-Sentience on the Rise


ChatGPT makes materials research much more efficient
Apr 2023, phys.org

"This isn't programming in the traditional sense; the method of interacting with these bots is through language," Morgan says. "Asking the program to extract data and then asking it to check if it is sure with normal sentences feels closer to how I train my children to get correct answers than how I usually train computers. It's such a different way to ask a computer to do things. It really changes how you think about what your computer can do."

via University of Wisconsin-Madison: Maciej P. Polak et al, Flexible, Model-Agnostic Method for Materials Data Extraction from Text Using General Purpose Language Models, arXiv (2023). DOI: 10.48550/arxiv.2302.04914

Also: Maciej P. Polak et al, Extracting Accurate Materials Data from Research Papers with Conversational Language Models and Prompt Engineering—Example of ChatGPT, arXiv (2023). DOI: 10.48550/arxiv.2303.05352

Image credit: In one of the first instances, phys.org uses a Stable Diffusion-generated image in the thumbnail, along with this statement: "This image was generated using Stable Diffusion, a text-to-image generator, using the prompt "researchers working with huge piles of data." -Dane Morgan and Maciej Polak, University of Wisconsin-Madison, 2023 [link]


New 'AI scientist' combines theory and data to discover scientific equations
Apr 2023, phys.org

The system rediscovered Kepler's third law of planetary motion, and produced a good approximation of Einstein's relativistic time-dilation law.

The new AI scientist—dubbed "AI-Descartes" by the researchers—joins the likes of AI Feynman and other recently developed computing tools that aim to speed up scientific discovery. At the core of these systems is a concept called symbolic regression, which finds equations to fit data. Given basic operators, such as addition, multiplication, and division, the systems can generate hundreds to millions of candidate equations, searching for the ones that most accurately describe the relationships in the data.

The system works particularly well on noisy, real-world data, which can trip up traditional symbolic regression programs that might overlook the real signal in an effort to find formulas that capture every errant zig and zag of the data. It also handles small data sets well, even finding reliable equations when fed as few as ten data points.

"In this work, we needed human experts to write down, in formal, computer-readable terms, what the axioms of the background theory are, and if the human missed any or got any of those wrong, the system won't work."

via IBM Research, Samsung AI, and University of Maryland Baltimore County: Combining Data and Theory for Derivable Scientific Discovery with AI-Descartes, Nature Communications (2023). DOI: 10.1038/s41467-023-37236-y


Researchers say AI emergent abilities are just a 'mirage'
Apr 2023, phys.org

"Previously claimed emergent abilities … might likely be a mirage induced by researcher analyses" 

Researchers contend that when results are reported in non-linear, or discontinuous, metrics, they appear to show sharp, unpredictable changes that are erroneously interpreted as indicators of emergent behavior, however an alternate means of measuring the identical data using linear metrics shows "smooth, continuous" changes that, contrary to the former measure, reveal predictable—non-emergent—behavior.

Large numbers getcha every time:

"The Stanford team added that failure to use large enough samples also contributes to faulty conclusions."

It's one of the easiest to spot when looking at the success of predictive powers, whether it's the weather, a sports bettor, or your financial advisor -- the law of large numbers makes us suck at identifying patterns. If you flip a perfect coin, there's a 50% chance of it landing on either heads or tails, which is what you would call a perfect chance, right down the middle. But if you flip the coin 10 times, you will probably not get 5 heads and 5 tails. You might have to flip it 100 times for that, or maybe even 10,000. And it depends on how many decimals you want to use, and if you start to get into millions and trillions of flips, then you'll have to cancel the variables in your system, like the weight of the respective sides of the coin, or the tendency for you hand to flip a certain way, or the prevailing winds, or patterns of seismic vibrations of the earth. 

People who understand very well the law of large numbers, or more likely people who don't udnerstand it and don't want to -- can do a good job convincing others of seeing whatever patterns they want, just by manipulating the metrics, like looking at performance from January to January instead of September to September, or 18 months instead of 12, or:

"The main takeaway," the researchers said, "is for a fixed task and a fixed model family, the researcher can choose a metric to create an emergent ability or choose a metric to ablate an emergent ability."

via Stanford University: Rylan Schaeffer et al, Are Emergent Abilities of Large Language Models a Mirage?, arXiv (2023). DOI: 10.48550/arxiv.2304.15004

AI Art - A doctor use his stethoscope on a huge mechanical brain pink background 1 - 2023

Study finds source validation issues hurt ChatGPT reliability
May 2023, phys.org

Only about half of generated sentences were fully supported by citations, and one quarter of citations failed to support associated sentences.

Moreover, the team found citation recall and precision were inversely correlated with fluency and perceived utility. "The responses that seem more helpful are often those with more unsupported statements or inaccurate citations," they observed.

As a consequence, they concluded, "This facade of trustworthiness increases the potential for existing generative search engines to mislead users."

via Stanford University's Human-Centered AI research group: Nelson F. Liu et al, Evaluating Verifiability in Generative Search Engines, arXiv (2023). DOI: 10.48550/arxiv.2304.09848


Online consumers at risk from 'intelligent' price manipulation, say experts
May 2023, phys.org

"Widespread use of intelligent algorithmics and dynamic pricing by online retailers, puts the public at risk of 'adversarial collusion"

More sophisticated algorithms can manipulate weaker algorithms and therefore collude together to increase prices for everyone, subtly undermine the competitiveness of online markets and harm consumers.

via University of Oxford: Luc Rocher, Adversarial competition and collusion in algorithmic markets, Nature Machine Intelligence (2023). DOI: 10.1038/s42256-023-00646-0


Ethical, legal issues raised by ChatGPT training literature
May 2023, phs.org

"Our work here has shown that OpenAI models know about books in proportion to their popularity on the web and the accuracy of such models is strongly dependent on the frequency with which a model has seen information in the training data."

Few if any details about data used to train the models are known to the public.

Also, science fiction and fantasy books dominate the list of memorized books, presenting a built-in bias on the nature of responses ChatGPT may provide. 

We should be thinking about whose narrative experiences are encoded in these models.

via University of California, Berkeley: Kent K. Chang et al, Speak, Memory: An Archaeology of Books Known to ChatGPT/GPT-4, arXiv (2023). DOI: 10.48550/arxiv.2305.00118

Post Script: I'm more interested in the simple statistical reality of algorithms trained on a completely "wild" dataset. The internet as a dataset is not curated, it's not designed, it's neither tamed nor maintained in any way; it is completely wild. When you apply the current state of the art in machine learning to a wild dataset, you multiply the wild part. 

AI Art - A doctor use his stethoscope on a huge mechanical brain pink background 2 - 2023

AI: War crimes evidence erased by social media platforms
Jun 2023, BBC News

AI-powered church service in Germany draws a large crowd
Jun 2023, Ars Technica

New tool explains how AI 'sees' images and why it might mistake an astronaut for a shovel
Jun 2023, phys.org

CRAFT -- for Concept Recursive Activation FacTorization for Explainability -- was a joint project with the Artificial and Natural Intelligence Toulouse Institute.

One of the concepts associated with the tench (a type of fish) is the face of a white male, because there are many photos online of white male sports fishermen holding fish that look like tench. In another example, the predominant concept associated with a soccer ball in neural networks is the presence of soccer players on the field. 

One way to explain AI vision is through what's called attribution methods, which employ heatmaps to identify the most influential regions of an image that impact AI decisions. However, these methods mainly focus on the most prominent regions of an image—revealing "where" the model looks, but failing to explain "what" the model sees in those areas.

But with CRAFT we can see how the system is ranking the concepts. 

With the 'image of an astronaut was incorrectly classified as a shovel' problem, CRAFT showed that the neural network identified the concept of "dirt" commonly found in members of the image class "shovel" and the concept of "ski pants" typically worn by people clearing snow from their driveway with a shovel.

via Brown University's Carney Institute for Brain Science: Thomas Fel et al, CRAFT: Concept Recursive Activation FacTorization for Explainability (2023)

AI Art - A doctor use his stethoscope on a huge mechanical brain pink background 3 - 2023

Study says AI data contaminates vital human input
Jun 2023, phys.org

They dubbed this phenomenon "artificial artificial artificial intelligence."

(But you should know this is because of the already-in-use term for Mechanical Turks, dubbed "artificial artificial intelligence"; and although they used to provide human input are now relying on AI-generated content, thus the "artificial" hole of recursion.)

"It is tempting to rely on crowdsourcing to validate large language model outputs or to create human gold-standard data for comparison," Veselovsky said. "But what if crowd workers themselves are using LLMs … in order to increase their productivity, and thus their income, on crowdsourcing platforms?"

Based on a limited study of the use of large language models by workers at MTurk, Amazon's crowd sourcing operation, the EPFL researchers estimated that 33% to 46% of worker assignments were completed with the aid of large language models.

via École polytechnique fédérale de Lausanne (EPFL), Lausanne, Switzerland: Veniamin Veselovsky et al, Artificial Artificial Artificial Intelligence: Crowd Workers Widely Use Large Language Models for Text Production Tasks, arXiv (2023). DOI: 10.48550/arxiv.2306.07899

Post Script: Figure this one out flesh engine of the future!  "fringe benefits, french benefits, and friends with benefits" boy is that a good one.


Is it growing pains or is ChatGPT just becoming dumber?
Jul 2023, phys.org

"We don't fully understand what causes these changes in ChatGPT's responses because these models are opaque."

"Any results on closed-source models are not reproducible and not verifiable, and therefore, from a scientific perspective, we are comparing raccoons and squirrels." -Sasha Luccioni of the AI company Hugging Face

via Stanford and UC Berkeley: Lingjiao Chen et al, How is ChatGPT's behavior changing over time?, arXiv (2023). DOI: 10.48550/arxiv.2307.09009

Monday, April 18, 2022

This Moment in Art History

AKA The One She Drew Back

This is the Ghislaine Maxwell sketch of her sketching the sketch artist Jane Rosenberg Reuters, but I call it The One She Drew Back, pastel on Canson, 2022.

I like to read the news. You see a lot of courtroom sketches in the headlines, and I don't think too much about them. I also don't go for superhyped stories like the Ghislaine Maxwell. But there was something striking about the thumbnail for this article, so I saved it for posterity. 

There's a lot going on here. First is the gaze, the subject of many art history theses. You don't see the person in the courtroom looking at the artist like this, it just doesn't happen long enough that the artist would end up catching it or devoting a portrait to it.

She was staring at the sketch artist because she was sketching the artist doing the sketching, which is good enough as it is, but we'll get to that in a moment. For now, it's a big deal because as a picture, she's not looking at the artist, she's looking at you, the viewer, which is a little unsettling because she's on trial for sex crimes, and she's looking right at you, staring into your soul, and for as long as you keep your eyes open to look.

That's the power of the gaze in art, and that's enough to make this a noteworthy piece. Not to mention the artist, Jane Rosenberg, has been at this for 40 years; she is an absolute professional in her field. Norman Rockwell was a commercial illustrator, until he became recognized as a fine artist. 

But let's get to the good stuff. The reason we get this shot, of her staring for so long at the sketch artist that we get a sketch of her direct gaze, is because she's sketching the artist. And the artist knows this, and she's then drawing the sketch of herself in the sketch. 

(We have to mention that during this very intense trial, they are staring at each other the whole time this is happening, long enough to do this sketch. )

This isn't Velázquez in Las Meninas, or M.C. Escher in the crystal ball, or Norman Rockwell in his studio. She's not drawing herself into the picture, she's drawing someone else drawing her into the picture. You came for the gaze, but you stay for the recursive lasagna. 

"Cult leader Charles Manson used to draw his sketch artist in court, she added."
-(Another) New York-based Illustrator Elizabeth Williams, New York Post article, link
The Gallery, by Jane Rosenberg for Reuters, 2022:



I almost forgot about the mask, which looks basically invisible to me right now, but likely will not in ten years: 

It’s much harder to sketch someone wearing a mask, but thankfully Ghislaine had very expressive eyes. Because that’s all I’ve got, eyes and hair. We basically have a half face to work with during the Covid era. People might think it’s easier, but it’s not.

Towards the end of the Maxwell trial, the Omicron variant became a big concern. They started making us wear these N95 masks that I couldn’t properly breathe in. The rules must have changed as Ghislaine was no longer able to hug her lawyers. It all just got so scary.
-"I was the court artist who Ghislaine drew back"  Jan 2022, Jane Rosenberg, The Independent, link

Tools of the Trade:

I bring prescription binoculars that I can wear on my head, a tripod, a thermos of coffee, a backpack with lunch, and a cushion to sit on on those hard bunches. ... I sketch in pastels on Canson paper. I bring latex finger cots because my skin gets so dried out from digging into my pastel box. 

In those days [the 1980's] there was always a camera person waiting outside the court for the sketch. I had to rush out of the courthouse and tape it up to the side of a truck. They’d take a copy and a motorcycle courier would rush it back to the newsroom. Later on, they’d send a satellite truck to send it back to the newsroom. Now I take a digital photograph of my sketches and send them by email.

Order in the Court:
Jane Rosenberg’s sketch of John Evans in 1983 electrocuted three times before he died.
Although some state courts now televise trials, American courts have historically resisted allowing cameras – because photography is considered distracting, and can turn courts into media spectacles, and because of the risk of compromising the identities of jurors or protected witnesses. (New York permits photography on a case-by-case basis, but federal courts strictly prohibit it.)
‘My life is weird’: the court artist who drew Ghislaine Maxwell drawing her back
Dec 2021, J Oliver Conroy, The Guardian, link

Witness Protection:
When courtroom artists sketch jurors or sensitive witnesses they often leave their faces blank. Rosenberg’s illustrations of the Maxwell trial and other cases include poignant portraits of anonymous witnesses with ghostly, blank faces, their features sometimes further obscured by hands clutching tissues.

The Post Script:
I'd like to connect this post with the Art Cop, another niche-world art professional, also from New York City. 

And this photo, the crediting an artwork in itself:
Court sketch artist Jane Rosenberg, Courtesy of Jane Rosenberg, no credit necessary, 2022, The Independent, link


Wednesday, January 23, 2019

Measure Me


There is something recursive about measuring the thing that we use to measure things that we then use to measure.

A gram used to be the weight of a cubic centimeter of water, and a (centi-)meter was the circumference of the Earth divided a million times, thus the Earth is 40 million meters.

But dematerialization has been a thing for much longer than the macro image series mania (called memes nowadays) that has by now thoroughly taken over our cultural ecosystem.

And now, the standard of weight all around the world no longer comes from an object you can hold in your hand, but from a measurement of electricity. A meter, by the way, is no longer determined by the Earth, but by a measurement of an electromagnetic wave, which is about as close as you can get to being immaterial.

Post Script:
In a slight twist of categorical confusion, the British Thermal Unit (BTU) is the amount of energy it takes to heat one pound of water by one degree F, whereas the metric version is a Calorie, or the amount of energy it takes to heat one gram of water by 1 degree C. Regardless, I second the suggestion that we quick change everything to metric while the government is on pause.

Notes:
Kilogram gets a new definition
Dec 2018, BBC

Radiolab episode about the meter

Empericism
In philosophy, empiricism is a theory that states that knowledge comes only or primarily from sensory experience. 

Metronomy is a London-based electro rock band, check them out.

Saturday, May 20, 2017

Semibotic Semibiotic


It's not often you get to see an article about consciousness on the BBC, but you do, it's Dan Dennett getting mega-memetical. (just kidding, that doesn't even make sense, in this context.)

Is consciousness just an illusion?
Apr 2017, BBC

We're not just are robots", he says. "We're robots, made of robots, made of robots".
-Dan Dennett

Saturday, July 7, 2012

Semiotic Equilibrium



Across the street, the sign in the window of the jeweler’s shop says “pardon our appearance”. The sign runs across the middle third of the entire window. Typically, the window is adorned with expensive, delicate jewelry. Now, behind the sign (which takes up a large portion of the window), we see an empty showcase with a small piece of plywood and a power drill in the far right corner.

Though the sign says ‘pardon our appearance’, it also says ‘look at our mess’, or the lack thereof, really. There’s barely anything to pardon, and even if there was, the sign itself would be the more demanding-of-pardon of the two. The cancels itself out in some kind of semiotic equilibrium?

It’s similar to, but not the same as, the sign hanging on the fence that says “do not hang signs on fence”. What it says and what it does are at odds in some way. In the case of the 'pardon our appearance' sign, it asks us to do something that wouldn't even need to be done if the sign weren't asking us...The sign should say 'pardon our sign (which wouldn't have to be pardoned if the sign weren't here in the first place)'.

Don't actually pardon our appearance, but our pardoning of our appearance.