New research has found similarities in how humans and artificial intelligence integrate two types of learning, offering new insights about how people learn as well as how to develop more intuitive AI tools.

The study is published in the Proceedings of the National Academy of Sciences.

Led by Jake Russin, a postdoctoral research associate in computer science at Brown University, the study found by training an AI system that flexible and incremental learning modes interact similarly to working memory and long-term memory in humans.

“These results help explain why a human looks like a rule-based learner in some circumstances and an incremental learner in others,” Russin said. “They also suggest something about what the newest AI systems have in common with the human brain.”

Russin holds a joint appointment in the laboratories of Michael Frank, a professor of cognitive and psychological sciences and director of the Center for Computational Brain Science at Brown’s Carney Institute for Brain Science, and Ellie Pavlick, an associate professor of computer science who leads the AI Research Institute on Interaction for AI Assistants at Brown.

Depending on the task, humans acquire new information in one of two ways. For some tasks, such as learning the rules of tic-tac-toe, “in-context” learning allows people to figure out the rules quickly after a few examples. In other instances, incremental learning builds on information to improve understanding over time—such as the slow, sustained practice involved in learning to play a song on the piano.

While researchers knew that humans and AI integrate both forms of learning, it wasn’t clear how the two learning types work together. Over the course of the research team’s ongoing collaboration, Russin—whose work bridges machine learning and computational neuroscience—developed a theory that the dynamic might be similar to the interplay of human working memory and long-term memory.

To test this theory, Russin used “meta-learning”—a type of training that helps AI systems learn about the act of learning itself—to tease out key properties of the two learning types. The experiments revealed that the AI system’s ability to perform in-context learning emerged after it meta-learned through multiple examples.

One experiment, adapted from an experiment in humans, tested for in-context learning by challenging the AI to recombine similar ideas to deal with new situations. If taught about a list of colors and a list of animals, could the AI correctly identify a combination of color and animal (e.g. a green giraffe) it had not seen together previously? After the AI meta-learned by being challenged to 12,000 similar tasks, it gained the ability to successfully identify new combinations of colors and animals.

The results suggest that for both humans and AI, quicker, flexible in-context learning arises after a certain amount of incremental learning has taken place.

“At the first board game, it takes you a while to figure out how to play,” Pavlick said. “By the time you learn your hundredth board game, you can pick up the rules of play quickly, even if you’ve never seen that particular game before.”

The team also found trade-offs, including between learning retention and flexibility: Similar to humans, the harder it is for AI to correctly complete a task, the more likely it will remember how to perform it in the future. According to Frank, who has studied this paradox in humans, this is because errors cue the brain to update information stored in long-term memory, whereas error-free actions learned in context increase flexibility but don’t engage long-term memory in the same way.

For Frank, who specializes in building biologically inspired computational models to understand human learning and decision-making, the team’s work showed how analyzing strengths and weaknesses of different learning strategies in an artificial neural network can offer new insights about the human brain.

“Our results hold reliably across multiple tasks and bring together disparate aspects of human learning that neuroscientists hadn’t grouped together until now,” Frank said.

The work also suggests important considerations for developing intuitive and trustworthy AI tools, particularly in sensitive domains such as mental health.

“To have helpful and trustworthy AI assistants, human and AI cognition need to be aware of how each works and the extent that they are different and the same,” Pavlick said. “These findings are a great first step.”

In addition to expanding renewable energy generation, a successful energy transition requires stable system operation at all times. To achieve this, renewable energies and storage power plants will have to take on extensive system services and essential grid-forming properties in the future.

In the “GFM Benchmark” project, Fraunhofer ISE developed a test procedure for grid-forming inverters on behalf of the four German transmission system operators and applied it to devices from various manufacturers. On the one hand, the project provided a comprehensive overview of the market readiness of grid-forming inverters.

On the other hand, the project results provide an important practical check for new national and European testing standards.

The fully integrated grid components with grid-forming properties planned by the transmission system operators will not be able to fully meet the demand for grid-forming power. Therefore, customer systems must also contribute to stabilizing the power grid: They should behave in a grid-forming manner, i.e., contribute to providing a grid voltage with stable amplitude and frequency.

But what exactly does that mean? In recent years, many scientific studies and publications have been produced on this topic, and some countries have grid operator documents that describe grid-forming behavior. However, there is no uniform standardization or definition, which leaves room for interpretation.

Therefore, in the first step of the project, the Fraunhofer ISE team worked with the grid operators 50Hertz Transmission GmbH, Transnet BW GmbH, Amprion GmbH, and Tennet TSO GmbH to develop a measurement and evaluation procedure for the stabilizing properties of inverters, incorporating findings from both grid operation and research.

Major differences in grid-forming behavior

“We wanted to see what manufacturers understand by grid formation and how they implement this in the programming of their devices,” explains department head Dr. Sönke Rogalla from Fraunhofer ISE. “So we invited them to put their devices to the test in our laboratory.”

Seven companies responded to the call and had their storage inverters, which cover a power range from a few kilowatts to five megawatts, measured according to the new test procedure. They came from different countries and were at different technology readiness levels, from pilot to prototype to series production.

The researchers used the tests to investigate the differences between the devices in terms of grid formation by exposing them to various operating conditions in the laboratory. In addition to normal operation, critical grid situations such as rapid frequency changes, short circuits, and phase jumps were simulated.

“The devices exhibited similar behavior under clearly defined requirements. In other cases, however, there were major differences, and we were able to provide the manufacturers with suggestions for optimization for almost every device,” explains project manager Roland Singer from Fraunhofer ISE. The willingness and commitment of manufacturers to advance the development of grid-forming inverters is high.

Proven verification methods are essential for market launch

At the same time, the project provided relevant practical experience in testing grid-forming inverters and optimized the test procedures. Important findings were incorporated into the ongoing standardization work at the European level even during the project phase. The Fraunhofer ISE team contributed its expertise to the creation of the VDE FNN note “Grid-forming properties.”

The recently published document describes the requirements and verification procedures for grid-forming units. It forms the normative basis for participation in the future market for instantaneous reserve, which will start at the beginning of 2026 and represents an additional interesting remuneration path, especially for battery storage systems.

With its experience in the “GFM Benchmark” project, the team at Fraunhofer ISE is ideally positioned to support manufacturers and users of grid-forming units with certification measurements in accordance with the FNN note. Standardization work at the European level is also progressing. ENTSO-E, the network of European transmission system operators, is working on an implementation guide with comprehensive grid-forming requirements, which should facilitate the transition to national regulations.

Provided by
Fraunhofer ISE

Some 15 people have died after the Gloria funicular railway car in Lisbon, Portugal, derailed and crashed on Wednesday local time.

Emergency services have also confirmed that more than 18 people were also injured, five of them seriously, in the tragedy, which occurred at the start of the evening rush hour.

It follows another accident on the same line in May 2018, when one of the cars derailed due to flaws in the maintenance of its wheels. No one was killed in that incident.

The exact cause of the most recent accident is not yet known. Witnesses have reported that the yellow-and-white tram appeared out of control as it sped downhill, before derailing as it rounded a bend and crashing into a building. Photos of the aftermath show a crumpled heap of cables and steel.

These cable car–like transport systems are rare relics of the 19th century, found in only a few very hilly places around the world. So how do they work? And why are they still in use?

How do funicular railways work?

Trains and trams typically only work on flat terrain. That’s because their steel wheels can’t get enough traction on steel rails on steep hills. As a workaround, railroad engineers often build tunnels through steep mountainsides.

Funicular railways, however, can go up very steep hills.

They usually feature two counterbalanced cars that are attached via a haulage cable.

As one car descends, it helps pull the ascending car up the hillside. The weight of the ascending car also prevents the descending one from careening out of control. Some now have electric motors to help power them and some are able to engage a one-way mechanical drive just for steep hills.

Even though funicular systems are typically quite slow and clunky, they are still popular with both tourists and residents in the places where they’re found.

Where are they found?

The Gloria funicular railway line in Lisbon opened in 1885. One of three funicular lines in Lisbon, it connects the city’s downtown area with the Bairro Alto (Upper Quarter).

But there are other examples of these transport relics around the world.

Switzerland has several funicular railways. The most notable is the Stoosbahn—the steepest funicular in the world. It covers a total ascent of around 744 meters, reaching a gradient of 47 degrees. It is a very popular tourist trip.

In Hong Kong, the Peak Tram is a funicular railway that has operated since 1888 and takes people to near the top of Hong Kong Island.

Last year, there was also some discussion about installing a new funicular railway system in the Blue Mountains in New South Wales, Australia, that would travel 14 meters every second.

The rise of trackless trams

Funicular railways still serve a purpose for people living in—or visiting—steep areas where they’re found. However, newer technology means more conventional forms of rail transport are now far less limited in traveling up and down hills.

For example, trackless trams are kind of a combination between a tram and a bus. They use GPS and digital sensors to move precisely along an invisible track and have rubber wheels, enabling them to ascend gradients of up to 15%. However, these have not yet been built for steeper hills.

I have enjoyed riding such funicular trams in a range of hilly cities, but this crash is likely to take the shine off the tourist experience. It’s about time we had a 21st-century option that is clearly safer.

This article is republished from The Conversation under a Creative Commons license. Read the original article.