machine learning – Pain In The Bud

Another Tech Company Wants To Translate Meows And Barks Using AI: Can It Work?

Cats and dogs communicate primarily by scent, touch and body language, but human efforts to understand them have focused exclusively on meows and barks. If we want to truly understand our non-human friends, we need to take an approach that considers the other ways animals “talk” to each other.

A few years ago when MeowTalk made a minor splash in the startup world, I was pretty bullish on its potential to help us understand our cats better.

Sure, the app had an unhelpful habit of attributing improbably loving declarations to Buddy, but I thought it would follow the trajectory of other machine learning models and drastically improve as it accumulated more data.

More users meant the app would record and analyze more meows, chirps and trills, meaning it was just a matter of time before the AI would be able to distinguish between an “I want attention!” meow and a “My bowl is dangerously close to empty!” meow.

Obviously that didn’t happen, and what I personally didn’t take into account back then — and should have, given how obvious it is in retrospect — is that cats don’t just communicate via vocalizations.

In fact, cats don’t normally incorporate vocalizations into communication at all. Pet kitties do it entirely for our benefit because they know we’re generally awful at interpreting body language and we are completely useless when it comes to olfactory information.

It’s actually amazing when you really think about how much of the heavy lifting cats do in our efforts to communicate with each other. They recognize we can’t communicate the way they do naturally, so they try to relate to us on our terms. In return, we meet them less than halfway.

No wonder Buddy sometimes looks frustrated as he meows at me, as if I’m the biggest moron in the world for not understanding the very obvious thing he’s trying to tell me.

“Human, how can you not understand the simple feeling of innerer schweinehund I’m trying to convey here? The cringe is killing me!”

Now the Chinese tech giant Baidu is throwing its hat into the ring after filing a patent in China for an AI system that uses machine learning to decode animal communication and “translate” it to human language.

Machines are designed to process things from a human viewpoint according to human logic, so if Baidu wants to succeed where MeowTalk has not, its engineers will need to take a thoughtful approach with the help of animal behavior experts.

This is a hard problem that encompasses animal cognition, neuroscience, linguistics, biology, biochemistry and even philosophy. If they approach this strictly as a tech challenge, they’ll set themselves up for failure.

Without the information and context clues provided by tails, whiskers, facial expressions, posture, eye dilation, heart rate, pheromones and even fur, an AI system is only getting a fraction of the information cats are trying to convey.

Trying to glean meaning from that is like trying to read a book in which only every fourth or fifth letter is legible. There’s just too much missing information.

Even if we can train machines to analyze sound visual, tactile and olfactory information, it may not be possible to truly translate what our cats are saying to us. We may have to settle for approximations. We’ve only begun to guess at how the world is interpreted differently among human beings thanks to things like qualia and neurodivergence, and the way cats and dogs see the world is undoubtedly more strange to us than the way a neurodivergent person might make sense of reality.

“He grimaced. He had drawn a greedy old character, a tough old male whose mind was full of slobbering thoughts of food, veritable oceans full of half-spoiled fish. Father Moontree had once said that he burped cod liver oil for weeks after drawing that particular glutton, so strongly had the telepathic image of fish impressed itself upon his mind. Yet the glutton was a glutton for danger as well as for fish. He had killed sixty-three Dragons, more than any other Partner in the service, and was quite literally worth his weight in gold.” – Cordwainer Smith, The Game of Rat and Dragon

An animal’s interpretation of reality may be so psychologically alien that most of its communication may be apples to oranges at best. Which is why I always loved Cordwainer Smith’s description of the feline mind as experienced via a technology that allows humans with special talents to share thoughts with cats in his classic short story, The Game of Rat and Dragon.

In the story, humans are a starfaring civilization and encounter a threat in the void between stars that people don’t have the reaction speed to deal with. Cats, however, are fast and swift enough, and with a neural bridge device, teams of humans paired with cats are able to keep passengers safe on interstellar journeys.

The narrator, who is one of the few people with an affinity for teaming up with felines, hopes he’ll be paired with one of his two favorite cats for his latest mission, but instead he’s assigned to partner with an old glutton of a tomcat whose mind was dominated by “slobbering thoughts of food, veritable oceans of half-spoiled fish.”

The narrator wryly notes that the last time one of his colleagues was paired with that particular cat, his burps tasted of fish for weeks afterward. But the cat in question, despite being obsessed with fish, is a badass at killing “dragons,” the human nickname for the bizarre entities that attack human ships in space. (The software that allows felines and humans to link thoughts also portrays the “dragons” as rodents in the minds of the cats, stimulating their ancient predatory drive so they’ll attack instantly when they see the enemy.)

We can’t know for sure if Smith’s interpretation of the feline mind is accurate, but another part rang true when he wrote that cat thoughts were all about the moment, filled with sentiments of warmth and affection, while they rapidly lost interest in thoughts about human concerns, dismissing them “as so much rubbish.”

If the mind of a cat is that relatable, we’ll be incredibly lucky. But in reality we’re dealing with animals who evolved in drastically different ecological niches, with different priorities, motivations, and ways of looking at the world — literally and figuratively.

That doesn’t mean we shouldn’t try to understand our furry friends. Research has yielded interesting information about the way animals like whales and elephants communicate, and AI is at its best when it augments human creativity and curiosity instead of trying to replace it.

Even if we don’t end up with a way to glean 1:1 translations, the prospect of improving our understanding of animal minds is tantalizing enough. We just need to make sure we’re listening to everything they’re saying, not just the meows.

Study: Cats Use Facial Expressions To Reassure Each Other Everything’s Cool During Play Time

Once again, we’ve underestimated cats. There’s so much more to the ways in which they communicate than we realize.

We know cats use non-verbal signals to communicate with each other, but recent research suggests we may just be scratching the surface, glimpsing only a portion of the information that passes between our furry friends.

Cats “talk” to each other by the way they position their tails, whiskers and ears, in addition to their overall body language.

It turns out there’s more. A group of interdisciplinary scientists from universities in Kansas, Arkansas and Haifa, Israel, found cats also employ specific facial expressions, and rapidly mirror each other’s expressions during play time to signal they’ve got good intentions and aren’t going to hurt each other.

The study, which was given the yawn-inducing title “Computational investigation of the social function of domestic cat signals” (in English: using AI to figure out how our house cats “talk” to each other), started with observations of felines playing with each other in cat cafes.

From there, the coders and mathematicians on the team created an algorithm to record and sort the facial expressions the cafe cats used, employing CatFACS (Cat Facial Action Coding System) to associate each expression with its meaning.

“I’m serious. Feed me, or feel my wrath.”

Cats make a surprising number of facial expressions, 276 in total, according to a 2023 study.

The problem is, we humans are terrible at reading them. Even veterinarians trained in CatFACS still struggle to get it right, but happily this is precisely the sort of task algorithmic AI excels at. Like facial recognition software, a well-trained machine learning algorithm can recognize faces and record them more accurately and much faster than any person could.

In a column praising the facial expressions study, evolutionary biologist and Jane Goodall Foundation ethics board member Mark Bekoff said it’s the kind of labor-intensive work that truly advances our understanding of the ways animals communicate.

For cats and their human caretakers, Bekoff notes, it could help us reduce inter-species misunderstandings and make it easier to read our cat’s emotions, so we know when they’re not feeling well or need something.

“There are no substitutes for doing what’s needed to learn about the nitty-gritty details of how animals communicate with one another in different contexts,” Bekoff wrote. “This study of play opens the door for more widespread comparative research focusing on how animals talk to one another.”

“Do I look happy, human?” Credit: Milan Nykodym/Wikimedia Commons

We also know adult cats very rarely meow to each other, and the meow is reserved for cat-to-human communication. Imagine the frustration our little friends must feel when they have so much to tell us, but the only thing we understand are vocalizations — meows, chirps and trills — that can convey only basic ideas at best.

Elephants Call Each Other By Name, Study Says

Elephants encode names and other information in low-frequency rumbles that can be heard miles away. For social animals who live in large herds, it’s crucial to be able to address individuals.

Elephants are famously social animals, moving in matriarchal herds that can consist of as many as 70 of their kind.

They also communicate over long distances, emitting rumbles that can be heard miles away.

Because of their social and nomadic existence, it makes sense that elephants would need a way to single out individuals and address each other, and for the first time researchers say they’ve found evidence of Earth’s largest land animals calling each other by name.

“If you’re looking after a large family, you’ve got to be able to say, ‘Hey, Virginia, get over here!’” Duke University ecologist Stuart Pimm told the Associated Press.

The research involved field work and analysis using artificial intelligence. To record samples of elephants communicating, teams followed herds with recording equipment. Notably, elephant rumbles include sounds in frequencies lower than the human ear can detect.

large elephants near lake — Credit: Pixabay/Pexels

The team paid close attention when one elephant vocalized and another responded, and recorded who initiated each rumble and who it was meant for.

Although elephants are best known for making loud “trumpeting” sounds, experts say those are more like exclamations while rumbles contain encoded information that African savanna elephants would need to communicate to each other.

“The rumbles themselves are highly structurally variable,” said Mickey Pardo, a biologist from Cornell University and co-author of the study. “There’s quite a lot of variation in their acoustic structure.”

A machine learning algorithm was then used to sort and categorize the large number of audio samples, looking for patterns that are difficult for human minds to detect.

“Elephants are incredibly social, always talking and touching each other — this naming is probably one of the things that underpins their ability to communicate to individuals,” said George Wittemyer, an ecologist at Colorado State University and co-author of the study. “We just cracked open the door a bit to the elephant mind.”

A female elephant with her young offspring in Kenya. Credit: Pixabay/Pexels

Notably, the elephant “names” are identifiers that they created for themselves, and are not the kind of human-bestowed names that cats and dogs respond to. The list of animals who have names for themselves is short, although likely to expand with further study. Dolphins, for instance, identify themselves with unique whistling patterns, and parrots have a similar method, but both species address individuals by imitating their calls. Elephants use their name analogs the way humans do, to directly address each other.

The research has the potential to raise public awareness of elephant intelligence and their plight as they face threats to their continued existence. Like almost all of the Earth’s iconic megafauna, elephants will become extinct if we don’t do a better job protecting them and ending the ivory trade. Every year about 20,000 elephants are slaughtered for their tusks to feed the demand for ivory, especially in China where it’s considered a status symbol, according to the World Wildlife Fund.

Yang Feng Glan, known as the”Queen of Ivory,” was sentenced to 15 years in prison by a Tanzanian court in 2019. Yang smuggled some 860 elephant tusks worth $6.5 million from Tanzania to China as the leader of one of the world’s most extensive poaching and ivory smuggling organizations.

During her years operating the smuggling ring, Yang presented herself as a successful businesswoman and ran in elite circles within China, authorities said. Two of her accomplices were also given 15-year sentences for their roles, but since then others have filled the vacuum left by Fang’s conviction, and elephant preserves are constantly under threat from heavily armed poachers.

The Dividing Line Between Human And Animal Has Been Blurred Again As AI Reveals Startlingly Complex Whale Language

By unlocking the mysteries of how sperm whales communicate and demonstrating their impressive cognitive abilities, researchers hope to get people invested in the fate of these endangered animals.

Sperm whales are chatty.

Their language is markedly different from the deep cetacean moans associated with other whales, taking the form of Morse code-like clicks that boom through the ocean in a decibel range almost twice that of jet engines.

And while we’ve long known animals like monkeys assign specific meaning to short vocalizations varying from alarm calls to affirmations of social rank, sperm whale conversations can endure for an hour or more, with participants exchanging complex strings of clicks that vary depending on context, environment and even which pod family is speaking.

Sperm_whale_Tim_Cole_NMFS_crop — An aerial view of a sperm whale near the ocean surface. Credit: Wikimedia Commons

While artificial intelligence has been maligned over the past few years as people grapple with its rapid progress and potential for abuse, it remains the best tool we have for teasing out patterns that our human minds can’t discern, especially from large quantities of data.

With more than 9,000 recordings of sperm whales, Project CETI — Cetacean Translation Initiative, a non-profit effort to decode and translate sperm whale communication — had precisely the kind of huge data cache that AI excels at analyzing.

By feeding the recordings into specially trained machine learning algorithms, the research team was able to identify a wealth of new language patterns. While human languages are composed of quantized morphemes — prefixes, suffixes and root words — whale communication is broken down into sequences of clicks and pauses called “codas.”

Like Morse code, codas make a distinction between short clicks and long clicks. Sperm whales also vary the tempo of the clicks, which could represent inflection, “dialects” or concepts completely alien to the human mind.

“Some of what they’re doing might be totally different from our way of communicating and we’re probably never going to be able to fully grasp those differences,” Oregon State postdoctoral marine researcher Taylor Hersh told NPR.

A sperm whale fluke visible above the surface of the ocean. Credit: Wikimedia Commons

Researchers believe the “inter-click intervals” — akin to ghost notes in music — may be as significant as the clicks themselves. Importantly, while human ears were able to identify and catalog some of the codas, the machine learning algorithms found many that human analysis missed.

That’s not surprising considering sperm whales — the loudest animals on Earth, capable of generating sounds up to 230 dB — took a much different evolutionary course and, as ocean-dwelling creatures weighing up to 90,000 pounds (40,800 kg) likely have a radically different sensorium compared to humans.

The comparisons to music go further than ghost notes.

“This study shows that coda types are not arbitrary, but rather that they form a newly discovered combinatorial coding system in which the musical concepts of rubato and ornamentation combine with two categorical, context-independent features known as rhythm and tempo, by analogy to musical terminology,” CETI’s team wrote on May 7 while unveiling the most recent study.

Sperm_whale_distribution_(Pacific_equirectangular) — Sperm whale distribution based on human sightings. Sperm whales freely travel the oceans except in cold, ice-packed environs. Credit: Wikimedia Commons

While people have used many abilities to mark the dividing line between humans and animals over the years — including the ability to use tools, experience emotions, and demonstrate self-awareness — human capacity for authentic language with syntax and context-dependent meaning was one of the stalwarts, standing the test of time as new research toppled the other dividers by showing animals do indeed use tools, experience rich emotions and have complex inner mental lives.

With this research, scientists are assembling a “sperm whale phonetic alphabet” that will make it easier to discern and catalog whale codas.

To be clear, there’s still a lot of work ahead before scientists can prove sperm whale codas are comparable to human definitions of language, but whether they strictly meet that definition may not matter. After all, it’s clear the clicks and pauses of whale codas are imbued with meaning, even if it remains elusive to us for the moment.

Indeed, “sperm whale communication has both contextual and combinatorial structure not previously observed in whale communication,” the team wrote.

Proving sperm whale codas are tantamount to human language isn’t the goal anyway. The team has two overriding priorities — decode the meanings behind the codas, and get the wider public invested in the fate of these endangered animals by showing they’re not so different from us.

“Our results show there is much more complexity than previously believed,” MIT AI lab director Daniela Rus told NPR, “and this is challenging the current state of the art or state of beliefs about the animal world.”

‘Why I Took 50,000 Pictures Of My Cats Pooping’

One software engineer went to incredible lengths to monitor her cat’s bathroom habits.

When Alan Turing, the father of artificial intelligence, posed the heady question “Can machines think?”, he inspired generations of computer scientists, philosophers, physicists and regular people to imagine the emergence of silicon-based consciousness, with humanity taking the godlike step of creating a new form of life.

And when science fiction writer Philip K. Dick wrote his seminal 1968 novel, “Do Androids Dream of Electric Sheep?” — the story that would eventually become Ridley Scott’s 1982 classic Bladerunner — he wondered what makes us human, and whether an artificial being could possess a soul.

It’s safe to say neither of those techno-prophets were thinking of fledgling AI algorithms, representing the first small steps toward true machine-substrate intelligence, announcing themselves and their usefulness to the world by helping us watch felis catus take a shit.

And yet that’s what the inventors of the LuluPet litter box designed an AI to do, and it’s what software engineer and Youtuber Estefannie did for her cat, Teddy, who’s got a bit of a plastic-eating problem.

“The veterinarian couldn’t tell me how much plastic he ate, and it would cost me over $3,000 [to find out]. So I didn’t do it,” Estefannie explains in a new video. “Instead, the vet gave me the option of watching him go to the bathroom. If he poops and there’s no plastic in his intestines, then he won’t die, and he might actually love me back.”

Estefannie casually described how she wrote a python script, set up a camera and motion sensor, and rigged it to take photos of Teddy doing his business. But, she explained, there was “a tiny problem”: Luna the Cat, aka her cat’s cat.

“This is Luna, this is technically not my cat, this is Teddy-Bear’s cat, and she uses the same litter box as Teddy,” she explained.

For that, she’d need more than a script. She’d have to build a machine learning algorithm to gorge itself on data, cataloguing tens of thousands of photos of Teddy and Luna along with sensory information from the litter box itself, to learn to reliably determine which cat was using the loo.

So Estefannie decided it was a good opportunity to “completely remodel” Teddy’s “bathroom,” including a compartment that would hide the bespoke system monitoring his bowel movements. The system includes sensors, cameras and lights to capture still images of Teddy dropping deuces in infrared, and a live thermal imaging feed of the little guy doing his business. (Teddy’s luxurious new bedroom turned out to be too dark for conventional cameras, thus the pivot to infrared.)

From there, Estefannie manually calculated how long Teddy’s number ones and twos took, and cross-referenced that information with photo timestamps to help determine the exact nature of Teddy’s calls of nature.

catpoopinggui — The future! (Note: This is our cheesy photoshopped interpretation, not Estefannie’s actual stool monitoring interface.)

When all the data is collected, Estefannie’s custom scripts sends it to an external server, which analyzes the images from each of Teddy’s bathroom visits and renders a verdict on what he’s doing in there.

Finally, Estefannie gets an alert on her smartphone when one of the cats steps into the litterbox, allowing her the option of watching a live feed and, uh, logging all the particulars. The software determines if a number two was successful, and keeps detailed records so Teddy’s human servant can see aberrations over time.

“So now I definitely know when Teddy-Bear is not pooping and needs to go to the hospital,” she said.

I am not making this up.

For her part, Estefannie says she’s not worried about a technological singularity scenario in which angry or insulted machines, newly conscious, exact revenge on humans who made them do unsavory tasks.

“Did I make an AI whose only purpose in life is to watch my cats poop?” Estefannie asked, barely keeping a straight face. “Mmmhmm. Will it come after me when the machines rise? No! Ewww!”