How to distinguish slow and fast earthquakes

New data analysis bolsters a controversial law defining the scale of slow earthquakes, allows reinterpretation, and offers a potential for forecasting fast earthquakes.

Peer-Reviewed Publication

SCHOOL OF SCIENCE, THE UNIVERSITY OF TOKYO

 

IMAGE: WHILE FAST EARTHQUAKES ARE VIOLENT SHAKES LASTING FOR A FEW MINUTES, SLOW EARTHQUAKES ARE SUBDUED SHAKES LASTING SEVERAL DAYS. UNDERSTANDING SLOW EARTHQUAKES MIGHT HELP FORECAST FAST EARTHQUAKES. view more 

CREDIT: IMAGE MODIFIED FROM “SCIENCE OF SLOW EARTHQUAKES” LEAFLET. HTTPS://WWW.ERI.U-TOKYO.AC.JP/PROJECT/SLOWEQ/EN/NEWSLETTERS/PDF/LEAFLET_EN.PDF

Researchers from the University of Tokyo and Stanford University show what differentiates slow and fast earthquakes and how their magnitudes vary with time.  

 

Normally, earthquakes last up to a few minutes and radiate strong seismic waves. But around 23 years ago, scientists discovered an unusual slow-slip phenomena called slow earthquakes. Slow earthquakes last days or even months. Though they involve significant tectonic movement, you may never feel them. Since slow earthquakes could indicate future fast earthquakes, monitoring and understanding them helps accurately forecast devastating earthquakes and tsunamis. 

 

Understanding them requires knowing how they change over time. For that, researchers use scaling laws which define the relationship between two quantities over a wide interval. In 2007, researchers proposed a controversial scaling law relating the magnitude and duration of earthquakes, which can help differentiate slow and fast earthquakes.  

 

According to the scaling law, for slow earthquakes, as its magnitude (measured by a quantity called seismic moment) increases, the duration of the earthquake increases proportionately. For fast earthquakes, the relation is not linearly but cubically proportionate, which means the seismic moment increases very rapidly in a short time. 

 

The scaling law received criticism from other researchers and raised questions about the likelihood of events in between slow and fast earthquakes that do not fall within the law. Seismologists Satoshi Ide of the University of Tokyo and Gregory Beroza of Stanford University now bolster the scaling law with more data, reinterpret the scaling relation, and address the controversy. 

 

“Most of the challenges to the scaling law were problematic, but we have had no chance to disprove their challenges,” says Ide. “A surprise was that totally erratic results were published in Nature, and believed by many scientists, who made further problematic numerical models.” 

 

With the advent of new seismic detection technology and data accumulated over 16 years, Ide and Beroza now reason that most arguments against the law had improper data calculations and were inconsistent given their data constraints. They suggest the presence of a speed limit to slow earthquakes and reveal physical processes that differentiate slow and fast earthquakes. 

 

Many, but all the same 

Since slow earthquakes include phenomena with different frequency bands, they are more diverse than fast earthquakes. They were named differently, such as low-frequency earthquakes, tectonic tremors, very low-frequency earthquakes, and slow slip events. So researchers observing one type of slow earthquake considered other types irrelevant. “Our study confirmed that all these phenomena are mutually connected, or rather regarded as a single phenomenon that radiates various signals,” explains Ide. 

 

Slow slips, but not so fast 

Slow earthquakes are so subtle and inaccessible that detecting and monitoring them is challenging. Due to the detection bias, only large enough slow earthquakes are observed. That prompted Ide and Beroza to propose an upper limit to the speed of slow earthquakes. Based on that, the duo redefined the 2007 scaling law with the maximum value constraint. As they showed continuous evidence for the scaling law over a broad time scale of less than a second to more than a year, they put an end to the debate. 

 

How are slow and fast earthquakes different? 

When Ide’s group proposed the scaling law in 2007, they were unsure of what makes these two earthquake types different. Now, with more data and theoretical models, Ide and Beroza show that their scaling differences dictate physical movement processes governing the events. Diffusion processes govern slow earthquakes, whereas seismic wave propagation dictates fast earthquakes. Because of this difference, the magnitude of slow earthquakes cannot be as large as fast earthquakes when the event lasts longer.  

 

“We pointed out that ‘diffusion’ is important in slow earthquakes, but what is physically diffusing is not well understood,” says Ide. 

 

Experts still don't know what kind of forecast information they can provide based on slow earthquake monitoring. This study will be a foundation for building appropriate numerical models, making predictions, and taking countermeasures. 

 

###

Useful resources for journalists:

• A leaflet explaining slow and fast earthquakes:  

https://www.eri.u-tokyo.ac.jp/project/sloweq/en/newsletters/pdf/leaflet_EN.pdf  

• About the School of Science, the University of Tokyo: https://www.s.u-tokyo.ac.jp/en/overview/  

• Researcher website: http://www-solid.eps.s.u-tokyo.ac.jp/~ide/en/  

• About the project: https://en.slow-to-fast-eq.org/project-overview/   

 

How slow and fast earthquakes differ in their scaling 

They differ in the duration and the magnitude of the quake. Data from slow earthquakes from three well-known earthquakes regions, Nankai, Cascadia, and Mexico, are highlighted in different colors.

CREDIT

Ide and Beroza, 2023.

Journal article: Satoshi Ide and Gregory Beroza. 2023. Slow earthquake scaling reconsidered as a boundary between distinct modes of rupture propagation. PNAS.   

 

JOURNAL

Proceedings of the National Academy of Sciences

METHOD OF RESEARCH

Data/statistical analysis

ARTICLE TITLE

Slow earthquake scaling reconsidered as a boundary between distinct modes of rupture propagation

ARTICLE PUBLICATION DATE

31-Jul-2023

Holding Trump accountable will not threaten American democracy

Reports and Proceedings

UNIVERSITY OF PENNSYLVANIA

With a Fulton County indictment of former President Donald Trump possible at any time, law enforcement in Atlanta is bracing for potential violence, with orange barricades restricting access to the entrance of the county courthouse.  

With the anticipation of each new indictment has come threats of violence, decrease in trust in American justice and calls for retribution against the government. Just how concerned should Americans be that we may face another January 6th-type incident?

New data from the Polarization Research Lab — a collaboration among researchers at Dartmouth College, the University of Pennsylvania, and Stanford University — has found that despite rhetoric from Trump and his allies, the public do not see the indictment of a former president as a reason to abandon democratic principles or as a call to support violent retribution. 

During the study — conducted by lab Co-Directors Yphtach Lelkes, Associate Professor of Communication at the Annenberg School for Communication at the University of Pennsylvania; Sean J. Westwood, Associate Professor of Government at Dartmouth College; and Derek Holliday, postdoctoral fellow with the lab, based at Stanford University — the researchers tracked public attitudes among Americans toward democratic norms, political violence, and general partisan hatred. 

For nearly a year, they performed daily tracking polls with nearly 30,000 survey interviews.

What they found is that 97% of Americans — from both parties — oppose political violence, and the majority disapprove when politicians violate political norms.

Reacting to the June 12 Indictments

Support for political violence remained stable before and after the June 12 federal indictment of Trump and co-defendant Walt Naruta, with most Americans rejecting even non-violent acts of protest.

Following the indictment, both Democrats and Republicans felt a surge of positive regard for their own parties, but this effect lasted only a week, and each party’s dislike of the other remained unchanged. 

The indictments on June 12 did increase Republican appetites for violating political norms, but it was fleeting, and driven largely by a four percentage point rise in the desire to censor news networks that were, in their view, unfair to their party.  

The effect vanished after seven days, and by the end of June, Republicans were less supportive of norm violations relative to the period before the indictment. 

A Fractured America

Trump’s indictments did not permanently change American attitudes on democracy, but real threats remain.

“Trump rose to power in a fractured America and while he contributed to divides, they existed before his presidency and will continue in the years to come,” Lelkes says.

The researchers’ data shows that the majority of everyday Americans have no positive views of the opposing party, don’t trust their politicians, don’t feel their government is responsive to their needs, and nearly a majority support at least one democratic norm violation. 

It remains possible that the short-lived trends observed in their data could trigger mob behavior similar to what happened on January 6th, they say.

But critically, the overwhelming majority of Americans would reject both violence and an attack on democratic norms. 

“As much as American politics can be crude and juvenile, when it comes to support for democracy itself, our data show that Americans are remarkably adult-like,” Lelkes says.

---

METHOD OF RESEARCH

Survey

SUBJECT OF RESEARCH

People

ARTICLE TITLE

Holding Trump Accountable Has Not Threatened American Democracy

ARTICLE PUBLICATION DATE

28-Jul-2023

 

Elusive pygmy right whale is a homebody hiding in our waters

Peer-Reviewed Publication

UNIVERSITY OF NEW SOUTH WALES

The pygmy right whale is an enigma in the whale world. Not only is it the smallest of the characteristically large filter-feeding baleen whales, but it’s also rarely sighted and seldom studied – partly because of its inconspicuous nature and resemblance to minke whales.

But new research led by UNSW Sydney analysing chemical clues hidden in the jawlines of pygmy right whales confirms what scientists have long suspected – these majestic aquatic mammals behave quite differently to their much larger whale relatives. The research, published in Frontiers in Marine Science, suggests the pygmy right whale is one of the few whale species to forgo long-distance migrations to Antarctic waters and instead reside in the temperate waters near Australia throughout the year. 

“If you google pygmy right whale, one of the first queries that comes up is whether they are extinct, which illustrates just how little we know about them, and how little known they are,” says Adelaide Dedden, lead author of the study and a marine ecologist at UNSW Science. “What’s especially surprising is that it turns out these whales are somewhat a resident species, given they appear to be hanging around Australia year-round.”

“They’re our little homebodies, just happy to hide away and never leave the comforts of our continental waters,” says Professor Tracey Rogers, a marine ecologist at UNSW Science and senior author of the study. “It’s also fascinating as they are similar in size to the ancestral baleen whales that didn’t journey between feeding and breeding grounds, and now we know these small whales behave similarly.”

Uncovering the chemical clues in their bristles

Studying an animal the size of a whale in a lab – even a relatively small species like the pygmy right whale – is challenging. Instead, scientists can analyse the composition of smaller tissues that keep detailed records of an animal’s activity.

For filter-feeding whales like the pygmy right whale, long, slender keratin bristles called baleen plates that hang from their upper jaw allow them to take in many small prey like krill – but they also lay down chemical signals called stable isotopes that give researchers hints about their behaviour.

“As the baleen grows, biochemical signals from their food called stable isotopes get trapped,” Prof. Rogers says. “These signals don’t decay over time, so it’s like reading a history book about their behaviour, including what they ate and the general area they were in at the time.”

For the study, the researchers analysed the stable isotopes in the baleen plates of 14 Australian pygmy right whales. Each baleen plate, loaned from the South Australian Museum, contained up to three to four years’ worth of data to assemble a picture of the pygmy right whale’s diet and movements across nearly 40 years – the most extensive study of the pygmy right whale diet and movement to date.

“Their isotopic record shows they remain in mid-latitude waters year-round off southern Australia, feeding on krill and copepods (small crustaceans),” Ms Dedden says. “There was no evidence of feeding in Antarctic waters at all, suggesting the waters off Southern Australia appear to be able to support their needs year-round.”

This makes sense when you look at the slender structures of their baleen plates relative to their body size,” Prof. Rogers says. “They need to draw in as much as possible from these less productive waters around Australia to stay in the region year-round.” 

The researchers also found some evidence the whales’ feeding patterns were linked to oceanic changes that drive food web dynamics in the region, specifically upwelling events – natural cycles bringing nutrient-rich water to the surface, resulting in increased krill availability in Australian waters.

“The signals in the baleen that reflected Australian krill coincided with times of upwelling in summer, which made sense because that’s when the krill is most available,” Ms Dedden says. “That signal wasn’t as strong during winter, which reflects the lower availability of krill during downwelling, so it further supports our overall findings.”

A potential indicator of ecosystem health

The researchers say their findings help lay the foundation for more research that can help us better understand the pygmy right whale. 

“Now we have increased evidence they live in this mid-latitude distribution, it would be ideal if we could do some satellite tagging to more closely monitor their movements and see exactly where they’re travelling around in the region,” Ms Dedden says. 

“They were never targeted for whaling, so their numbers are likely stable now,” Prof. Rogers says. “But they’re data deficient, and we don’t know enough about them to know for sure they are in a good place.”

While its current status is classified as least concern, the pygmy right whale’s population trend is still relatively unknown, and the researchers say it may be vulnerable to emerging risks such as warming oceans.

But the pygmy right whale may also be valuable as an indicator species scientists can study to monitor the marine environment. 

“They may face future challenges if there are significant changes in the marine ecosystem, given they feed at the base of the food chain and appear to rely on particular regions,” Ms Dedden says. 

“But with more research, they also have the potential to help us better forecast these risks to protect both them and the ecosystem they call home.”

The study also involved co-authors Gary Truong and Dr Matthew McCurry from UNSW, Dr Catherine Kemper from the South Australia Museum and Dr Paul van Ruth from the University of Tasmania.

---

JOURNAL

Frontiers in Marine Science

DOI

10.3389/fmars.2023.1190623 

METHOD OF RESEARCH

Observational study

SUBJECT OF RESEARCH

Animal tissue samples

ARTICLE TITLE

Stable isotopes infer the diet and habitat of the enigmatic pygmy right whale (Caperea marginata) off southern Australia

ARTICLE PUBLICATION DATE

31-Jul-2023

Scientists solve ‘enigma’ of pygmy right whales’ feeding habits

Understudied species of baleen whales are ‘homebodies’ who stay and feed in waters off southern Australia year-round

Peer-Reviewed Publication

FRONTIERS

 

VIDEO: A PYGMY RIGHT WHALE (CAPEREA MARGINATA) IN THE WILD. VIDEO CREDIT: HENRY CORDELL view more 

CREDIT: HENRY CORDELL

Pygmy right whales (Caperea marginata) are the smallest, ‘most enigmatic’, and probably least studied of all baleen whales. Baleen act like sieves in the mouth of baleen whales, which allow seawater to pass but trap small prey items like zooplankton and small fish.

Pygmy right whales are rarely seen in the wild. Reason for this may be their relatively small size – 6.5 meters long and weighing up to 3.5 tons –, sparse distribution, and inconspicuous behavior, especially compared to the boisterous humpback whales. Historically, whalers rarely bothered hunting them. The little we know about them is mainly based on beached animals.

“Here we show that pygmy right whales don’t behave like most other baleen whales: they don’t make long cross-ocean migrations,” said Dr Tracey Rogers, a professor of ecology and evolution at the University of New South Wales in Australia. “Instead, they are homebodies, who remain near the coast of southern Australia year-round, where they breed and feed on krill and copepods.”

Rogers is the senior author on a new study in Frontiers in Marine Science, which demonstrated that pygmy right whales are restricted to mid-latitude waters and likely inhabit regions like the eastern Great Australian Bight (between Cape Catastrophe and Cape Pasley) and the region dominated by the Bonney Upwelling, between Portland and Robe. Here, south-eastern winds drive cold, nutrient-rich water from the depths to the surface between November and May, providing a food bonanza for seabirds and marine life.

Beached whales

The researchers measured the ratio between stable nitrogen isotopes 15N and 14N (the δ15N value) and between stable carbon isotopes 13C and 12C (the δ13C) value) in the baleen plates of 14 adult pygmy right whales, to infer their diet and habitat use. These included both females and males that had stranded between 1968 and 2019 on the Tasmanian or southern Australian coast. The baleen plates were on loan from the South Australian Museum in Adelaide.

“Baleen is made of keratin, like our fingernails, and grows throughout the life of the whale,” explained lead author Adelaide Dedden, a doctoral student in Rogers’ research group. “As a stable tissue, baleen provides an ideal long-term signal to look at their diet and habitat use.”

Isotope ratio ‘fingerprints’

Because animals derive their nitrogen and carbon exclusively from food, the isotope ratios in their tissues reflect those of their prey. Isotope ratios increase in a regular pattern across trophic levels within the food web, with phytoplankton having the lowest values, and apex predators the highest. By comparing the δ15N and δ13C values measured in baleen to those published for a range of possible prey, Dedden et al. could deduce which species are on the menu of pygmy right whales.

Not like other baleen whales

The results showed that the isotope ratios in the baleen of pygmy right whales closely match those of copepods and krill species from regions rich in zooplankton off  Australia.

In contrast, there was no correspondence with isotope ratios of Antarctic krill, which means that pygmy right whales don’t make seasonal migrations to the Antarctic, like many other species of baleen whales do.

Likewise, because the isotope ratios didn’t match those of pelagic fish, the authors deduced that unlike larger species of baleen whales, pygmy right whales don’t feed on fish.

The restricted, mid-latitude range of pygmy right whales and their reliance on specific prey puts them at risk, warned the authors.

“As large-bodied mammals who feed on tiny prey, pygmy right whales need to consume vast quantities of food. This makes them vulnerable to changes in their local environment. Their home, the temperate oceans of the southern hemisphere, is warming at an alarming rate. We plan to study next how they will respond to this change,” said Rogers.

Pygmy right whale [VIDEO] | 1

Pygmy right whale [VIDEO] | 2

Baleen of pygmy right whales, on loan from the South Australian Museum in Adelaide

The first author, doctoral student Adelaide Dedden, with baleen from pygmy right whales in the foreground

CREDIT

Adelaide Dedden

---

JOURNAL

Frontiers in Marine Science

DOI

10.3389/fmars.2023.1190623 

METHOD OF RESEARCH

Experimental study

SUBJECT OF RESEARCH

Animals

ARTICLE TITLE

Stable isotopes infer the diet and habitat of the enigmatic pygmy right whale (Caperea marginata) off southern Australia

ARTICLE PUBLICATION DATE

31-Jul-2023

 

Luzio, who lived in São Paulo 10,000 years ago, was Amerindian like Indigenous people now, DNA reveals

An investigation covering four different parts of Brazil carried out analysis of genomic data from 34 fossils, including larger skeletons and the famous mounds of shells and fishbones built on the coast, and revealed differences between communities

Peer-Reviewed Publication

FUNDAÇÃO DE AMPARO À PESQUISA DO ESTADO DE SÃO PAULO

 

IMAGE: THE INVESTIGATION THAT COVERED FOUR DIFFERENT PARTS OF BRAZIL CARRIED OUT ANALYSIS OF GENOMIC DATA FROM 34 FOSSILS, INCLUDING LARGER SKELETONS AND THE FAMOUS MOUNDS OF SHELLS AND FISHBONES BUILT ON THE COAST view more 

CREDIT: ANDRÉ STRAUSS

An article to be published on July 31 in Nature Ecology & Evolution reveals that Luzio, the oldest human skeleton found in São Paulo state (Brazil), was a descendant of the ancestral population that settled the Americas at least 16,000 years ago and gave rise to all present-day Indigenous peoples, such as the Tupi.

Based on the largest set of Brazilian archeological genomic data, the study reported in the article also offers an explanation for the disappearance of the oldest coastal communities, who built the icons of Brazilian archeology known as sambaquis, huge mounds of shells and fishbones used as dwellings, cemeteries and territorial boundaries. Archeologists often refer to these monuments as shell mounds or kitchen middens.

“After the Andean civilizations, the Atlantic coast sambaqui builders were the human phenomenon with the highest demographic density in pre-colonial South America. They were the ‘kings of the coast’ for thousands and thousands of years. They vanished suddenly about 2,000 years ago,” said André Menezes Strauss, an archeologist at the University of São Paulo’s Museum of Archeology and Ethnology (MAE-USP) and principal investigator for the study.

The first author of the article is Tiago Ferraz.The study was supported by FAPESP (projects 17/16451-2 and 20/06527-4) and conducted in partnership with researchers at the University of Tübingen’s Senckenberg Center for Human Evolution and Paleoenvironment (Germany).

The authors analyzed the genomes of 34 samples from four different areas of Brazil’s coast. The fossils were at least 10,000 years old. They came from sambaquis and other parts of eight sites (Cabeçuda, Capelinha, Cubatão, Limão, Jabuticabeira II, Palmeiras Xingu, Pedra do Alexandre and Vau Una).

This material included Luzio, São Paulo’s oldest skeleton, found in the Capelinha river midden in the Ribeira de Iguape valley by a group led by Levy Figuti, a professor at MAE-USP. The morphology of its skull is similar to that of Luzia, the oldest human fossil found to date in South America, dating from about 13,000 years ago. The researchers thought it might have belonged to a biologically different population from present-day Amerindians, who settled in what is now Brazil some 14,000 years ago, but it turns out they were mistaken.

“Genetic analysis showed Luzio to be an Amerindian, like the Tupi, Quechua or Cherokee. That doesn’t mean they’re all the same, but from a global perspective, they all derive from a single migratory wave that arrived in the Americas not more than 16,000 years ago. If there was another population here 30,000 years ago, it didn’t leave descendants among these groups,” Strauss said.

Luzio’s DNA also answered another question. River middens are different from coastal ones, so the find cannot be considered a direct ancestor of the huge classical sambaquis that appeared later. This discovery suggests there were two distinct migrations – into the hinterland and along the coast.

What happened to the sambaqui builders?

Analysis of the genetic material revealed heterogeneous communities with cultural similarities but significant biological differences, especially between coastal communities in the southeast and south.

“Studies of cranial morphology conducted in the 2000s had already pointed to a subtle difference between these communities, and our genetic analysis confirmed it,” Strauss said. “We discovered that one of the reasons was that these coastal populations weren’t isolated but ‘swapped genes’ with inland communities. Over thousands of years, this process must have contributed to the regional differences between sambaquis.”

Regarding the mysterious disappearance of this coastal civilization, comprising the first hunter-gatherers of the Holocene, analysis of the DNA samples clearly showed that, in contrast with the European Neolithic substitution of entire populations, what happened in this part of the world was a change of practices, with a decline in construction of shell middens and the introduction of pottery by sambaqui builders. For example, the genetic material found at Galheta IV (Santa Catarina state), the most emblematic site for the period, has remains not of shells but of ceramics and is similar to the classic sambaquis in this respect.

“This information is compatible with a 2014 study that analyzed pottery shards from sambaquis and found that the pots in question were used to cook not domesticated vegetables but fish. They appropriated technology from the hinterland to process food that was already traditional there,” Strauss said.

 

About São Paulo Research Foundation (FAPESP)

The São Paulo Research Foundation (FAPESP) is a public institution with the mission of supporting scientific research in all fields of knowledge by awarding scholarships, fellowships and grants to investigators linked with higher education and research institutions in the State of São Paulo, Brazil. FAPESP is aware that the very best research can only be done by working with the best researchers internationally. Therefore, it has established partnerships with funding agencies, higher education, private companies, and research organizations in other countries known for the quality of their research and has been encouraging scientists funded by its grants to further develop their international collaboration. You can learn more about FAPESP at www.fapesp.br/en and visit FAPESP news agency at www.agencia.fapesp.br/en to keep updated with the latest scientific breakthroughs FAPESP helps achieve through its many programs, awards and research centers. You may also subscribe to FAPESP news agency at http://agencia.fapesp.br/subscribe.

---

JOURNAL

Nature Ecology & Evolution

DOI

10.1038/s41559-023-02114-9 

ARTICLE TITLE

Genomic history of coastal societies from Eastern South America

ARTICLE PUBLICATION DATE

31-Jul-2023

GPT-3 can reason about as well as an AMERIKAN college student, UCLA psychologists report

But does the technology mimic human reasoning or is it using a fundamentally new cognitive process?

Peer-Reviewed Publication

UNIVERSITY OF CALIFORNIA - LOS ANGELES

People solve new problems readily without any special training or practice by comparing them to familiar problems and extending the solution to the new problem. That process, known as analogical reasoning, has long been thought to be a uniquely human ability.

But now people might have to make room for a new kid on the block.

Research by UCLA psychologists shows that, astonishingly, the artificial intelligence language model GPT-3 performs about as well as college undergraduates when asked to solve the sort of reasoning problems that typically appear on intelligence tests and standardized tests such as the SAT. The study is published in Nature Human Behaviour.

But the paper’s authors write that the study raises the question: Is GPT-3 mimicking human reasoning as a byproduct of its massive language training dataset or it is using a fundamentally new kind of cognitive process?

Without access to GPT-3’s inner workings — which are guarded by OpenAI, the company that created it — the UCLA scientists can’t say for sure how its reasoning abilities work. They also write that although GPT-3 performs far better than they expected at some reasoning tasks, the popular AI tool still fails spectacularly at others.

“No matter how impressive our results, it’s important to emphasize that this system has major limitations,” said Taylor Webb, a UCLA postdoctoral researcher in psychology and the study’s first author. “It can do analogical reasoning, but it can’t do things that are very easy for people, such as using tools to solve a physical task. When we gave it those sorts of problems — some of which children can solve quickly — the things it suggested were nonsensical.”

Webb and his colleagues tested GPT-3’s ability to solve a set of problems inspired by a test known as Raven’s Progressive Matrices, which ask the subject to predict the next image in a complicated arrangement of shapes. To enable GPT-3 to “see,” the shapes, Webb converted the images to a text format that GPT-3 could process; that approach also guaranteed that the AI would never have encountered the questions before.

The researchers asked 40 UCLA undergraduate students to solve the same problems.

“Surprisingly, not only did GPT-3 do about as well as humans but it made similar mistakes as well,” said UCLA psychology professor Hongjing Lu, the study’s senior author.

GPT-3 solved 80% of the problems correctly — well above the human subjects’ average score of just below 60%, but well within the range of the highest human scores.

The researchers also prompted GPT-3 to solve a set of SAT analogy questions that they believe had never been published on the internet — meaning that the questions would have been unlikely to have been a part of GPT-3’s training data. The questions ask users to select pairs of words that share the same type of relationships. (For example, in the problem “‘Love’ is to ‘hate’ as ‘rich’ is to which word?,” the solution would be “poor.”)

They compared GPT-3’s scores to published results of college applicants’ SAT scores and found that the AI performed better than the average score for the humans.

The researchers then asked GPT-3 and student volunteers to solve analogies based on short stories — prompting them to read one passage and then identify a different story that conveyed the same meaning. The technology did less well than students on those problems, although GPT-4, the latest iteration of OpenAI’s technology, performed better than GPT-3.

The UCLA researchers have developed their own computer model, which is inspired by human cognition, and have been comparing its abilities to those of commercial AI.

“AI was getting better, but our psychological AI model was still the best at doing analogy problems until last December when Taylor got the latest upgrade of GPT-3, and it was as good or better,” said UCLA psychology professor Keith Holyoak, a co-author of the study.

The researchers said GPT-3 has been unable so far to solve problems that require understanding physical space. For example, if provided with descriptions of a set of tools — say, a cardboard tube, scissors and tape — that it could use to transfer gumballs from one bowl to another, GPT-3 proposed bizarre solutions.

“Language learning models are just trying to do word prediction so we’re surprised they can do reasoning,” Lu said. “Over the past two years, the technology has taken a big jump from its previous incarnations.”

The UCLA scientists hope to explore whether language learning models are actually beginning to “think” like humans or are doing something entirely different that merely mimics human thought.

“GPT-3 might be kind of thinking like a human,” Holyoak said. “But on the other hand, people did not learn by ingesting the entire internet, so the training method is completely different. We’d like to know if it’s really doing it the way people do, or if it’s something brand new — a real artificial intelligence — which would be amazing in its own right.”

To find out, they would need to determine the underlying cognitive processes AI models are using, which would require access to the software and to the data used to train the software — and then administering tests that they are sure the software hasn’t already been given. That, they said, would be the next step in deciding what AI ought to become.

“It would be very useful for AI and cognitive researchers to have the backend to GPT models,” Webb said. “We’re just doing inputs and getting outputs and it’s not as decisive as we’d like it to be.”

---

JOURNAL

Nature Human Behaviour

DOI

10.1038/s41562-023-01659-w