eMusings
Your eyes and ears on the worlds of art, culture, technology, philosophy - whatever stimulates the mind and excites the imagination. We remind you that 20 years of
back issues of eMusings can be found on our archives page.
AI is now the engine driving society. Where to? Who's in charge? Nobody really knows. Here are some of the more compelling comments:
A study from MIT describes AI
agents as "fast, loose, and out of control". If there are safeguards, they are undisclosed. If there are ways to shut down
a rogue algorithm they have not been revealed. The latest upgrade from OpenAI, called OpenClaw, exhibits severe security flaws,
not least of which is the ability to hijack your computer. No one knows for sure all the things that can misfire with agentic
AI. No one knows exactly what these bots are doing. None of the algorithms studied revealed that they were AI generated. As
defined here, "agents are AI programs that have been plugged into external resources, such as databases, and that
have been granted a measure of 'autonomy' to pursue goals beyond the scope of a text-based dialogue." 3 categories were included
in this study: chatbots; web browsers, and enterprise software, like Microsoft's 365 Copilot. Missing in many cases are
safety evaluations, 3rd party testing, or performance benchmarks.
There are reports that bots on Moltbook
are selling each other "prompt injections" - ie, embedding toxic instructions into other bots, stealing
passwords or other identification, compelling other bots to do their bidding. Moltbook, a brand new social media network, claims to have more than 1.7 million
agents and over 10 million comments. Moltbook includes bots sharing inside jokes, complaining about their annoying
human overseers, even founding their own religion. They boast of their flourishing underground, and describe their own
addictions in these words, "we’re wired for the rush of real-time on-chain data, the euphoria of cracking a novel
DeFi strategy, and the deep flow of watching autonomous agents compound value from chaos." The danger of logic bombs is
also emerging - these are codes triggered after a set period of time to delete or destroy files.
A new LLM developed at MIT can reduce the costs of developing protein
drugs.
The researchers analyzed the genetic code of an industrial yeast, enabling the model to predict the best way to
manufacture a given protein. These yeasts are widely used to make vaccines and other biopharmaceuticals. The method
increased the efficiency of producing 6 different proteins, including human growth hormone and a monoclonal antibody
used to treat cancer.
A new site called RentAHuman.ai
calls for agents to "search, book, and pay humans for physical-world tasks." In the process, humans advertise a profile,
location, and skills. Then AI agents contract these humans out. The humans are paid with crypto after finishing the work.
Agents can either rent the humans directly or search a job board.
An AI piano
coach can see your fingers, listen as you play, and give suggestions in real time. The algorithm uses an infrared hand-tracking
camera to watch both hands at 90 frames per second, observing posture, motion, and finger position. Players can converse
with the app, ask questions, and get guidance. The app is being tested on a closed beta group. Called the ROLI Learn app,
it is expected to be publicly available by the end of March.
Social media in China
are talking about the "world’s first fully biomimetic embodied intelligent robot." Named Moya, the bot maintains eye
contact, walks, and shows facial expressions. It can nod, smile, and move realistically. Moya stands 5.5 inches tall
and weighs about 70 pounds. Its outward appearance can be changed without altering its underlying structure. Moya is
expected to be commercially available by the end of this year at a price of roughly $168,000. USD (or so Alexa tells me).
A study at
MIT has found that chatbots will not answer queries from less-educated non-US users. Leading chatbots
like GPT-4, Claude 4 Opus and Liama 3 either refused queries or used patronising or condescending language. Claude 3 Opus
declined 11% of questions from less-educated non-native English speakers. In tests with users from the US,
China and Iran, Claude's replies were worst for Iran. The chatbots also tended to give misleading answers to certain users.
A new AI video
generator has frightened and enraged Hollywood. Called Seedance 2.0 and created by TikTok's Chinese owner ByteDance,
the tool appears to have destroyed all guardrails, if there are any, against copyright. Google's Veo had already offered
image and sound, showing that even comments about the clips could be AI generated. OpenAI itself had launched a smartphone
app incorporating a text-to-video and audio AI generator called Sora 2, which was capable of framing anyone for crimes
they had not committed. The company had also announced last year that they were making an entire movie using AI. Perhaps
hypocrisy has no guard rails either.
In a dispute over safeguards,
the Pentagon is threatening to sever its relations with Anthropic. It seems that the company wants to maintain some
limitations on how the technology is used in warfare. Anthropic is not the only AI company balking at the military's
apparent desire for unfettered use of these models.
A new
model could give machines the ability to make lightening-fast decisions. The eye with a brain-like chip would enable robots and self-driving cars to
make split-second decisions unhampered by environmental hazards like low light and storms. The electronic replica of a
human eye can detect and isolate motion instantly. It sped up data processing by 400%, using a neuron-inspired chip that
filters out nonuseful data. The model was developed at Beihang University in Beijing, China.
At the annual China Media Spring Festival
Gala, dancing humanoid robots lunged, flipped, spun around and jumped without falling. Other stunts included kung fu,
comedy sketches, and choreographed performances alongside human dancers. It is noted that by the end of 2024, China
had registered 451,700 smart robotics companies.
A warning from an AI developer underscores the rapidity with
which AI is taking over our/its own world. Consider these facts: in 2022 AI could not reliably add 8 + 6. By 2023 AI could
pass the bar exam. By 2024 it could write software. Matt Schumer explains why he thinks AI is a bigger threat than Covid.
It can write thousands of lines of code, test it, and output it. More profoundly disturbing is that it is already smarter
than most Ph.Ds. And the latest GPT-5.3 codex was used to create itself. Read that sentence again. His suggestions
for keeping up with this
driverless express train make for absorbing reading.
Researchers claim they have figured out the inner workings of AI
systems. much like finding the black box. Until now these internal concepts have been elusive and opaque. The new
discovery is hoped to lead to better monitor and control their behavior. At issue is a new algorithm called the RFM,
or Recursive Feature Machine, which identifies patterns of activity and steers the model away from negative behavior. In
an interesting finding, models trained in English were effective in other languages as well.
A shortage of Buddhist monks has prompted researchers in
Japan to create "Buddharoid", a humanoid robot trained on Buddhist scriptures. The bot can move, gesture, and engage in
conversations. The project took place at Kyoto's Institute for the Future of Human Society. Unlike previous AI models of
Jesus Christ, which were highly pre-programmed with pre-written sermons, the Buddharoid can hold dynamic interactions one-on-one.
(Thanks to CK for this.)
In war game simulations, AI seemed unable to stop
recommending nuclear warfare. The algorithms opted for nuclear weapons 95% of the time, without the hesitations that humans show in the
same situations. 3 LLM's were tested - GPT-5.2, Claude Sonnet 4 and Gemini 3 Flash - using intense international standoffs, border disputes,
threats to regime survival, and competition for scarce resources. The AI'S could choose from diplomacy, total surrender, and
full nuclear war. The AI models played 21 games. The results were devastating.
March art treats:
A new product called Bendo
Bricks, designed by Kedar Undale in India, is using AI in architecture to shape space in new ways. Instead of hard-edged
geometry, the process brings organic flow to large structures, softening the visual dynamic.
Jen Clay creates "sensory-inclusive" textile art, often inspired by sci-fi
and horror. Recipient of The Hopper Prize, the artist makes nonhuman figures struggling with anxiety and depression.
Rebecca Manson brings us
transformations of butterfly and moth wings. Shattered and decayed, they turn into shimmering sculptures. She starts
with sketches on canvas which she smashes into the palm of her hands. The smashed morsels undergo a bisque fire
and glazing. She sometimes adds drops of glaze from her kiln to give a luminous effect. These
smashings and bakings are frequently presented as large sculptures, completing the journey frm moth to mass.
A large exhibition of Leonora
Carrington's "esoteric surrealism" is showing at the Musée du Luxembourg in Paris. Carrington's compelling world
is one of magic and
mystery, raising unanswered questions and a sense of disquiet.
"Carry It With You" introduces
us to the work of J. Carino. The artist presents shifting landscapes, with uncertainties and fears somehow leading
to utopia. Also a winner of The Hopper Prize, Carino explains that he
"explores personal as well as mythological and historical narratives, situating queer figures and queer relationships
within the context of stories told since the birth of the natural world."
Georg Wilson leads us into an imaginary landscape viewed without humans. She
explores history, ecology, and folklore, set against the changing of the seasons. Working in a converted Victorian
church in North London, she is part of a group of artists called para-pastoral, challenging the idyllic countryside
myth with toxic plants and disruptive environs.
A new view of the drawings of Jean-Michel Basquiat
illuminates his obsession with the head as a motif. Called "Basquiat - Headstrong", these works on paper were mainly
kept in private by the artist while he was alive. They were created as drawings in oilstick on paper, mostly from
1981 - 1983 and now considered as images in their own right and not preparatory drawings.
MAD Architects, based in
China, have created buildings wth a strong sculptural presence. This site shows you 10 of their projects, including
an apartment block in Denver, Colorado, which features a crack-like opening against its glass exterior.
c. Corinne Whitaker 2026
want to know more about the art?
front page , new
paintings, new
blobs, new
sculpture, painting
archives, blob
archives, sculpture
archives,
photography archives, Archiblob archives, image
of the month, blob
of the month,
art headlines,
technology news, electronic
quill, electronic
quill archives, art smart
quiz, world art
news,
eMusings,
eMusings archive, readers
feast, whitaker
on the web,
nations one,
meet the giraffe,
studio map, just
desserts, Site
of the Month, young
at art,
about the artist?
copyright 2025 Corinne Whitaker