Cookies help us run our site more efficiently.

By clicking “Accept”, you agree to the storing of cookies on your device to enhance site navigation, analyze site usage, and assist in our marketing efforts. View our Privacy Policy for more information or to customize your cookie preferences.

Predicting Chaos With AI: The New Frontier in Autonomous Control

News Feed
Saturday, May 18, 2024

Advanced machine learning algorithms have shown potential in efficiently controlling complex systems, promising significant improvements in autonomous technology and digital infrastructure.Recent research highlights the development of advanced machine learning algorithms capable of controlling complex systems efficiently. These new algorithms, tested on digital twins of chaotic electronic circuits, not only predict and control these systems effectively but also offer significant improvements in power consumption and computational demands.Systems controlled by next-generation computing algorithms could give rise to better and more efficient machine learning products, a new study suggests.Using machine learning tools to create a digital twin, or a virtual copy, of an electronic circuit that exhibits chaotic behavior, researchers found that they were successful at predicting how it would behave and using that information to control it. The Limitations of Linear ControllersMany everyday devices, like thermostats and cruise control, utilize linear controllers – which use simple rules to direct a system to a desired value. Thermostats, for example, employ such rules to determine how much to heat or cool a space based on the difference between the current and desired temperatures.Yet because of how straightforward these algorithms are, they struggle to control systems that display complex behavior, like chaos.As a result, advanced devices like self-driving cars and aircraft often rely on machine learning-based controllers, which use intricate networks to learn the optimal control algorithm needed to best operate. However, these algorithms have significant drawbacks, the most demanding of which is that they can be extremely challenging and computationally expensive to implement.The Impact of Efficient Digital TwinsNow, having access to an efficient digital twin is likely to have a sweeping impact on how scientists develop future autonomous technologies, said Robert Kent, lead author of the study and a graduate student in physics at The Ohio State University.“The problem with most machine learning-based controllers is that they use a lot of energy or power and they take a long time to evaluate,” said Kent. “Developing traditional controllers for them has also been difficult because chaotic systems are extremely sensitive to small changes.”These issues, he said, are critical in situations where milliseconds can make a difference between life and death, such as when self-driving vehicles must decide to brake to prevent an accident.The study was published recently in Nature Communications.Advancements in Machine Learning ArchitectureCompact enough to fit on an inexpensive computer chip capable of balancing on your fingertip and able to run without an internet connection, the team’s digital twin was built to optimize a controller’s efficiency and performance, which researchers found resulted in a reduction of power consumption. It achieves this quite easily, mainly because it was trained using a type of machine learning approach called reservoir computing.“The great thing about the machine learning architecture we used is that it’s very good at learning the behavior of systems that evolve in time,” Kent said. “It’s inspired by how connections spark in the human brain.”Practical Applications and Future DirectionsAlthough similarly sized computer chips have been used in devices like smart fridges, according to the study, this novel computing ability makes the new model especially well-equipped to handle dynamic systems such as self-driving vehicles as well as heart monitors, which must be able to quickly adapt to a patient’s heartbeat.“Big machine learning models have to consume lots of power to crunch data and come out with the right parameters, whereas our model and training is so extremely simple that you could have systems learning on the fly,” he said.To test this theory, researchers directed their model to complete complex control tasks and compared its results to those from previous control techniques. The study revealed that their approach achieved a higher accuracy at the tasks than its linear counterpart and is significantly less computationally complex than a previous machine learning-based controller.“The increase in accuracy was pretty significant in some cases,” said Kent. Though the outcome showed that their algorithm does require more energy than a linear controller to operate, this tradeoff means that when it is powered up, the team’s model lasts longer and is considerably more efficient than current machine learning-based controllers on the market.“People will find good use out of it just based on how efficient it is,” Kent said. “You can implement it on pretty much any platform and it’s very simple to understand.” The algorithm was recently made available to scientists.Economic and Environmental ConsiderationsOutside of inspiring potential advances in engineering, there’s also an equally important economic and environmental incentive for creating more power-friendly algorithms, said Kent.As society becomes more dependent on computers and AI for nearly all aspects of daily life, demand for data centers is soaring, leading many experts to worry over digital systems’ enormous power appetite and what future industries will need to do to keep up with it.And because building these data centers as well as large-scale computing experiments can generate a large carbon footprint, scientists are looking for ways to curb carbon emissions from this technology.To advance their results, future work will likely be steered toward training the model to explore other applications like quantum information processing, Kent said. In the meantime, he expects that these new elements will reach far into the scientific community.“Not enough people know about these types of algorithms in the industry and engineering, and one of the big goals of this project is to get more people to learn about them,” said Kent. “This work is a great first step toward reaching that potential.”Reference: “Controlling chaos using edge computing hardware” by Robert M. Kent, Wendson A. S. Barbosa and Daniel J. Gauthier, 8 May 2024, Nature Communications.DOI: 10.1038/s41467-024-48133-3This study was supported by the U.S. Air Force’s Office of Scientific Research. Other Ohio State co-authors include Wendson A.S. Barbosa and Daniel J. Gauthier.

Recent research highlights the development of advanced machine learning algorithms capable of controlling complex systems efficiently. These new algorithms, tested on digital twins of chaotic...

Machine Learning AI Technology Illustration

Advanced machine learning algorithms have shown potential in efficiently controlling complex systems, promising significant improvements in autonomous technology and digital infrastructure.

Recent research highlights the development of advanced machine learning algorithms capable of controlling complex systems efficiently. These new algorithms, tested on digital twins of chaotic electronic circuits, not only predict and control these systems effectively but also offer significant improvements in power consumption and computational demands.

Systems controlled by next-generation computing algorithms could give rise to better and more efficient machine learning products, a new study suggests.

Using machine learning tools to create a digital twin, or a virtual copy, of an electronic circuit that exhibits chaotic behavior, researchers found that they were successful at predicting how it would behave and using that information to control it.

The Limitations of Linear Controllers

Many everyday devices, like thermostats and cruise control, utilize linear controllers – which use simple rules to direct a system to a desired value. Thermostats, for example, employ such rules to determine how much to heat or cool a space based on the difference between the current and desired temperatures.

Yet because of how straightforward these algorithms are, they struggle to control systems that display complex behavior, like chaos.

As a result, advanced devices like self-driving cars and aircraft often rely on machine learning-based controllers, which use intricate networks to learn the optimal control algorithm needed to best operate. However, these algorithms have significant drawbacks, the most demanding of which is that they can be extremely challenging and computationally expensive to implement.

The Impact of Efficient Digital Twins

Now, having access to an efficient digital twin is likely to have a sweeping impact on how scientists develop future autonomous technologies, said Robert Kent, lead author of the study and a graduate student in physics at The Ohio State University.

“The problem with most machine learning-based controllers is that they use a lot of energy or power and they take a long time to evaluate,” said Kent. “Developing traditional controllers for them has also been difficult because chaotic systems are extremely sensitive to small changes.”

These issues, he said, are critical in situations where milliseconds can make a difference between life and death, such as when self-driving vehicles must decide to brake to prevent an accident.

The study was published recently in Nature Communications.

Advancements in Machine Learning Architecture

Compact enough to fit on an inexpensive computer chip capable of balancing on your fingertip and able to run without an internet connection, the team’s digital twin was built to optimize a controller’s efficiency and performance, which researchers found resulted in a reduction of power consumption. It achieves this quite easily, mainly because it was trained using a type of machine learning approach called reservoir computing.

“The great thing about the machine learning architecture we used is that it’s very good at learning the behavior of systems that evolve in time,” Kent said. “It’s inspired by how connections spark in the human brain.”

Practical Applications and Future Directions

Although similarly sized computer chips have been used in devices like smart fridges, according to the study, this novel computing ability makes the new model especially well-equipped to handle dynamic systems such as self-driving vehicles as well as heart monitors, which must be able to quickly adapt to a patient’s heartbeat.

“Big machine learning models have to consume lots of power to crunch data and come out with the right parameters, whereas our model and training is so extremely simple that you could have systems learning on the fly,” he said.

To test this theory, researchers directed their model to complete complex control tasks and compared its results to those from previous control techniques. The study revealed that their approach achieved a higher accuracy at the tasks than its linear counterpart and is significantly less computationally complex than a previous machine learning-based controller.

“The increase in accuracy was pretty significant in some cases,” said Kent. Though the outcome showed that their algorithm does require more energy than a linear controller to operate, this tradeoff means that when it is powered up, the team’s model lasts longer and is considerably more efficient than current machine learning-based controllers on the market.

“People will find good use out of it just based on how efficient it is,” Kent said. “You can implement it on pretty much any platform and it’s very simple to understand.” The algorithm was recently made available to scientists.

Economic and Environmental Considerations

Outside of inspiring potential advances in engineering, there’s also an equally important economic and environmental incentive for creating more power-friendly algorithms, said Kent.

As society becomes more dependent on computers and AI for nearly all aspects of daily life, demand for data centers is soaring, leading many experts to worry over digital systems’ enormous power appetite and what future industries will need to do to keep up with it.

And because building these data centers as well as large-scale computing experiments can generate a large carbon footprint, scientists are looking for ways to curb carbon emissions from this technology.

To advance their results, future work will likely be steered toward training the model to explore other applications like quantum information processing, Kent said. In the meantime, he expects that these new elements will reach far into the scientific community.

“Not enough people know about these types of algorithms in the industry and engineering, and one of the big goals of this project is to get more people to learn about them,” said Kent. “This work is a great first step toward reaching that potential.”

Reference: “Controlling chaos using edge computing hardware” by Robert M. Kent, Wendson A. S. Barbosa and Daniel J. Gauthier, 8 May 2024, Nature Communications.
DOI: 10.1038/s41467-024-48133-3

This study was supported by the U.S. Air Force’s Office of Scientific Research. Other Ohio State co-authors include Wendson A.S. Barbosa and Daniel J. Gauthier.

Read the full story here.
Photos courtesy of

How Mississippians Can Intervene in Natural Gas Pipeline Proposal

Mississippi residents can comment on a proposal for a natural gas pipeline that would span nearly the full width of the state

Mississippians have until Tuesday to intervene in a proposal for a natural gas pipeline that would span nearly the full width of the state.The pipeline, called the “Mississippi Crossing Project,” would start in Greenville, cross through Humphreys, Holmes, Attala, Leake, Neshoba, Newton, Lauderdale and Clarke counties and end near Butler, Alabama, stretching nearly 208 miles.Tennessee Gas Pipeline Co., a subsidiary of Kinder Morgan, sent an application for the project to the Federal Energy Regulatory Commission on June 30. The company hopes the pipeline, which would transfer up to 12 billion cubic feet of natural gas per day, will address a rising energy demand by increasing its transportation capacity.Kinder Morgan says on its website that, should it receive approval, construction would begin at the end of 2027 and the pipeline would begin service in November 2028. The company says the project would cost $1.7 billion and create 750 temporary jobs as well as 15 permanent positions.The project would also include new compressor stations in Humphreys, Attala and Lauderdale counties, although exact locations haven’t been set.Singleton Schreiber, a national law firm that focuses on environmental justice, is looking to spread awareness of the public’s ability to participate in the approval process, whether or not they support the proposal.“We’re just trying to raise awareness to make sure that people know this is happening,” said Laura Singleton, an attorney with the firm. “They’re going to have to dig and construct new pipelines, so it’s going to pass through sensitive ecosystems like wetlands, private property, farmland, things like that. So you can have issues that come up like soil degradation, water contamination, and then after the pipeline is built you could potentially have leaks, spills.”Singleton added while such issues with pipelines are rare, when “things go bad, they go pretty bad.”To comment, protest, or file a motion to intervene, the public can go to FERC’s website (new users have to create an account, and then use the docket number “CP25-514-000”). The exact deadline is 4 p.m. on Aug. 5. More instructions can also be found here.In addition to FERC, the proposal will also face review from the U.S. Army Corps of Engineers, U.S. Fish and Wildlife Service, National Park Service and the state environmental agencies in Mississippi and Alabama.Mississippians have seen multiple incidents related to gas leaks in recent years. In March, three workers were injured after accidentally rupturing an Atmos Energy pipeline doing routine maintenance in Lee County, leaving thousands without service. Then last year, the National Transportation Safety Board found that Atmos discovered gas leaks over a month prior to two explosions in Jackson, one of which claimed the life of an 82-year-old woman.This story was originally published by Mississippi Today and distributed through a partnership with The Associated Press.Copyright 2025 The Associated Press. All rights reserved. This material may not be published, broadcast, rewritten or redistributed.Photos You Should See - June 2025

BPA faces suit over energy market decision that opponents say would raise rates

The lawsuit comes after governors, lawmakers, utility regulators and renewable energy proponents in the region unsuccessfully pressed the BPA to reconsider its plans.

Five energy and conservation nonprofits are suing the Bonneville Power Administration over its decision to join a new energy trading market, claiming it will raise electricity and transmission costs in Oregon and across the region. The lawsuit, filed Thursday in the 9th U.S. Circuit Court of Appeals, alleges that BPA’s move violates the Northwest Power Act and the National Environmental Policy Act and will also weaken energy grid reliability and reduce access to clean energy. BPA, the Northwest’s largest transmission grid operator, in May announced it would join the Arkansas-based Southwest Power Pool day-ahead market known as “Markets Plus” instead of joining California’s day-ahead market. The Southwest market is smaller with fewer electrical generation resources, experts say. Prior to that decision, Pacific Northwest governors, lawmakers, utility regulators and renewable energy proponents had pressed the BPA for months to reconsider its plans, which the agency initially announced in March.The nonprofits involved in the legal challenge are the Oregon Citizens’ Utility Board, a watchdog organization that advocates for utility customers; national environmental group the Sierra Club; the Montana Environmental Information Center, which promotes clean energy; the Idaho Conservation League, a natural landscape conservation group; and the NW Energy Coalition, which promotes affordable energy policies. The groups, represented by San Francisco-based environmental law nonprofit Earthjustice, want the court to vacate BPA’s decision, require the agency to prepare an environmental impact statement and rescind the financial commitments already made to the Southwest energy market.The BPA’s spokesperson Nick Quinata declined to comment on the pending litigation. Previously, the agency said the Southwest day-ahead market is superior to the California one because it would allow BPA to remain more independent due to its market design and governance structure. BPA, part of the U.S. Department of Energy, markets hydropower from 31 federal dams in the Columbia River Basin and supplies a third of the Northwest’s electricity, most of it to publicly owned rural utilities and electric cooperatives. It also owns and operates 15,000 miles – 75% – of the Northwest’s high-voltage transmission lines. Nearly every electric utility in Oregon benefits from either the clean hydroelectricity or the transmission lines controlled by BPA. BPA’s decision sets the stage for having two energy markets across the West.The lawsuit says that will likely lead to rising prices and blackouts during periods of high electricity demand because of the complexity of transmitting power across boundaries between different utilities and the agreements required for such transfers. Oregon’s two largest utilities, investor-owned Portland General Electric and Pacific Power, have both signed agreements to join California’s day-ahead market instead. They, too, have argued that once BPA leaves the Western market, the available energy they can purchase would diminish and become more expensive, leading to higher prices for customers across the region.Regional electricity providers also may have to construct additional power generation facilities, increase operation of existing facilities or both, to make up for BPA’s participation in a smaller and less efficient energy market, the suit contends. It could also increase reliance on generation resources powered by fossil fuels such as coal or natural gas plants because clean energy isn’t as widely available in the smaller Southwest market, the suit says. The Northwest Power Act, passed by Congress in the 1980s, requires BPA to provide low-cost power to the region while encouraging renewable energy, conservation and protection of fish and wildlife.BPA violated those duties when it chose the Southwest market option, according to the lawsuit. The groups also allege BPA’s market choice could harm fish and wildlife in the Columbia basin because it could alter the operation of the federal hydroelectric dams from which Bonneville markets power. The lawsuit claims BPA failed to comply with federal environmental law by not conducting any environmental impact analysis on impacts to fish and wildlife before making its decision. The Citizens’ Utility Board, a party to the lawsuit, said it hoped the BPA reverses course – otherwise its decision will splinter the West’s electricity markets, costing utility customers billions of dollars at a time when many are already dealing with skyrocketing bills.The board, as well as other critics of BPA’s decision, have pointed to an initiative developing an independent governance structure for California’s day-ahead market.“Oregon is facing overlapping energy challenges: rising utility bills, rising electricity demand from data centers, and stalling progress on meeting clean energy requirements. The last thing we need is for one of our region’s largest clean energy suppliers to reduce ties with the Pacific Northwest,” said the group’s spokesperson Charlotte Shuff. — Gosia Wozniacka covers environmental justice, climate change, the clean energy transition and other environmental issues. Reach her at gwozniacka@oregonian.com or 971-421-3154.If you purchase a product or register for an account through a link on our site, we may receive compensation. By using this site, you consent to our User Agreement and agree that your clicks, interactions, and personal information may be collected, recorded, and/or stored by us and social media and other third-party partners in accordance with our Privacy Policy.

States, enviro groups fight Trump plan to keep dirty power plants going

In late spring, the Department of Energy ordered two aging and costly fossil-fueled power plants that were on the verge of shutting down to stay open. The agency claimed that the moves were necessary to prevent the power grid from collapsing — and that it has the power to force the plants to stay open even if the…

In late spring, the Department of Energy ordered two aging and costly fossil-fueled power plants that were on the verge of shutting down to stay open. The agency claimed that the moves were necessary to prevent the power grid from collapsing — and that it has the power to force the plants to stay open even if the utilities, state regulators, and grid operators managing them say that no such emergency exists. But state regulators, regional grid operators, environmental groups, and consumer groups are pushing back on the notion that the grids in question even need these interventions — and are challenging the legality of the DOE’s stay-open orders. The DOE claimed that the threat of large-scale grid blackouts forced its hand. But state utility regulators, environmental groups, consumer advocates, and energy experts say that careful analysis from the plant’s owners, state regulators, regional grid operators, and grid reliability experts had determined both plants could be safely closed. These groups argue that clean energy, not fossil fuels, are the true solution to the country’s grid challenges — even if the ​“big, beautiful” bill signed by Trump last week will make those resources more expensive to build. Some of the environmental organizations challenging DOE’s orders have pledged to take their case to federal court if necessary. “We need to get more electrons on the grid. We need those to be clean, reliable, and affordable,” said Robert Routh, Pennsylvania climate and energy policy director for the Natural Resources Defense Council, one of the groups demanding that DOE reconsider its orders. Keeping J.H. Campbell and Eddystone open ​“results in the exact opposite. It’s costly, harmful, unnecessary, and unlawful.” Taking on the DOE’s grid emergency claims The groups challenging the DOE’s J.H. Campbell and Eddystone stay-open orders point out that the agency is using a power originally designed to protect the grid against unanticipated emergencies, including during wartime, but without proving that such an emergency is underway. “This authority that the Department of Energy is acting under — Section 202(c) of the Federal Power Act — is a very tailored emergency authority,” said Caroline Reiser, NRDC senior attorney for climate and energy. ​“Congress intentionally wrote it only to be usable in specific, narrow, short-term emergencies. This is not that.” For decades, the DOE has used its Section 202(c) power sparingly, and only in response to requests from utilities or grid operators to waive federal air pollution regulations or other requirements in moments when the grid faces imminent threats like widespread power outages, Reiser said. But DOE’s orders for Eddystone and J.H. Campbell were not spurred by requests from state regulators or regional grid operators. In fact, the orders caught those parties by surprise. They also came mere days before the plants were set to close down and after years of effort to ensure their closure wouldn’t threaten grid reliability. J.H. Campbell was scheduled to close in May under a plan that has been in the works since 2021 as part of a broader agreement between utility Consumers Energy and state regulators, and which was approved by the Midcontinent Independent System Operator (MISO), the entity that manages grid reliability across Michigan and 14 other states. “The plant is really old, unreliable, extremely polluting, and extremely expensive,” Reiser said. ​“Nobody is saying that this plant is needed or is going to be beneficial for any reliability purposes.” To justify its stay-open order, DOE cited reports from the North American Electric Reliability Corp. (NERC), a nonprofit regulatory authority that includes utilities and grid operators in the U.S. and Canada. NERC found MISO is at higher risk of summertime reliability problems than other U.S. grid regions, but environmental groups argue in their rehearing request that DOE has ​“misrepresented the reports on which it relies,” and that Consumers Energy, Michigan regulators, and MISO have collectively shown closing the plant won’t endanger grid reliability. Eddystone, which had operated only infrequently over the past few years, also went through a rigorous process with mid-Atlantic grid operator PJM Interconnection to ensure its closure wouldn’t harm grid reliability. The DOE’s reason for keeping that plant open is based on a report from PJM that states the grid operator might need to ask utility customers to use less power if it faces extreme conditions this summer — an even scantier justification than what the agency cited in its J.H. Campbell order, Reiser said. As long as the DOE continues to take the position that it can issue emergency stay-open orders to any power plant it decides to, these established methods for managing plant closures and fairly allocating costs will be thrown into disarray, she said. “We have a system of competitive energy markets in the United States that is successful in keeping the lights on and maintaining reliability the vast, vast majority of the time,” Reiser said. ​“The Department of Energy stepping in and using a command-and-control system interferes with those markets.”

Designing a new way to optimize complex coordinated systems

Using diagrams to represent interactions in multipart systems can provide a faster way to design software improvements.

Coordinating complicated interactive systems, whether it’s the different modes of transportation in a city or the various components that must work together to make an effective and efficient robot, is an increasingly important subject for software designers to tackle. Now, researchers at MIT have developed an entirely new way of approaching these complex problems, using simple diagrams as a tool to reveal better approaches to software optimization in deep-learning models.They say the new method makes addressing these complex tasks so simple that it can be reduced to a drawing that would fit on the back of a napkin.The new approach is described in the journal Transactions of Machine Learning Research, in a paper by incoming doctoral student Vincent Abbott and Professor Gioele Zardini of MIT’s Laboratory for Information and Decision Systems (LIDS).“We designed a new language to talk about these new systems,” Zardini says. This new diagram-based “language” is heavily based on something called category theory, he explains.It all has to do with designing the underlying architecture of computer algorithms — the programs that will actually end up sensing and controlling the various different parts of the system that’s being optimized. “The components are different pieces of an algorithm, and they have to talk to each other, exchange information, but also account for energy usage, memory consumption, and so on.” Such optimizations are notoriously difficult because each change in one part of the system can in turn cause changes in other parts, which can further affect other parts, and so on.The researchers decided to focus on the particular class of deep-learning algorithms, which are currently a hot topic of research. Deep learning is the basis of the large artificial intelligence models, including large language models such as ChatGPT and image-generation models such as Midjourney. These models manipulate data by a “deep” series of matrix multiplications interspersed with other operations. The numbers within matrices are parameters, and are updated during long training runs, allowing for complex patterns to be found. Models consist of billions of parameters, making computation expensive, and hence improved resource usage and optimization invaluable.Diagrams can represent details of the parallelized operations that deep-learning models consist of, revealing the relationships between algorithms and the parallelized graphics processing unit (GPU) hardware they run on, supplied by companies such as NVIDIA. “I’m very excited about this,” says Zardini, because “we seem to have found a language that very nicely describes deep learning algorithms, explicitly representing all the important things, which is the operators you use,” for example the energy consumption, the memory allocation, and any other parameter that you’re trying to optimize for.Much of the progress within deep learning has stemmed from resource efficiency optimizations. The latest DeepSeek model showed that a small team can compete with top models from OpenAI and other major labs by focusing on resource efficiency and the relationship between software and hardware. Typically, in deriving these optimizations, he says, “people need a lot of trial and error to discover new architectures.” For example, a widely used optimization program called FlashAttention took more than four years to develop, he says. But with the new framework they developed, “we can really approach this problem in a more formal way.” And all of this is represented visually in a precisely defined graphical language.But the methods that have been used to find these improvements “are very limited,” he says. “I think this shows that there’s a major gap, in that we don’t have a formal systematic method of relating an algorithm to either its optimal execution, or even really understanding how many resources it will take to run.” But now, with the new diagram-based method they devised, such a system exists.Category theory, which underlies this approach, is a way of mathematically describing the different components of a system and how they interact in a generalized, abstract manner. Different perspectives can be related. For example, mathematical formulas can be related to algorithms that implement them and use resources, or descriptions of systems can be related to robust “monoidal string diagrams.” These visualizations allow you to directly play around and experiment with how the different parts connect and interact. What they developed, he says, amounts to “string diagrams on steroids,” which incorporates many more graphical conventions and many more properties.“Category theory can be thought of as the mathematics of abstraction and composition,” Abbott says. “Any compositional system can be described using category theory, and the relationship between compositional systems can then also be studied.” Algebraic rules that are typically associated with functions can also be represented as diagrams, he says. “Then, a lot of the visual tricks we can do with diagrams, we can relate to algebraic tricks and functions. So, it creates this correspondence between these different systems.”As a result, he says, “this solves a very important problem, which is that we have these deep-learning algorithms, but they’re not clearly understood as mathematical models.” But by representing them as diagrams, it becomes possible to approach them formally and systematically, he says.One thing this enables is a clear visual understanding of the way parallel real-world processes can be represented by parallel processing in multicore computer GPUs. “In this way,” Abbott says, “diagrams can both represent a function, and then reveal how to optimally execute it on a GPU.”The “attention” algorithm is used by deep-learning algorithms that require general, contextual information, and is a key phase of the serialized blocks that constitute large language models such as ChatGPT. FlashAttention is an optimization that took years to develop, but resulted in a sixfold improvement in the speed of attention algorithms.Applying their method to the well-established FlashAttention algorithm, Zardini says that “here we are able to derive it, literally, on a napkin.” He then adds, “OK, maybe it’s a large napkin.” But to drive home the point about how much their new approach can simplify dealing with these complex algorithms, they titled their formal research paper on the work “FlashAttention on a Napkin.”This method, Abbott says, “allows for optimization to be really quickly derived, in contrast to prevailing methods.” While they initially applied this approach to the already existing FlashAttention algorithm, thus verifying its effectiveness, “we hope to now use this language to automate the detection of improvements,” says Zardini, who in addition to being a principal investigator in LIDS, is the Rudge and Nancy Allen Assistant Professor of Civil and Environmental Engineering, and an affiliate faculty with the Institute for Data, Systems, and Society.The plan is that ultimately, he says, they will develop the software to the point that “the researcher uploads their code, and with the new algorithm you automatically detect what can be improved, what can be optimized, and you return an optimized version of the algorithm to the user.”In addition to automating algorithm optimization, Zardini notes that a robust analysis of how deep-learning algorithms relate to hardware resource usage allows for systematic co-design of hardware and software. This line of work integrates with Zardini’s focus on categorical co-design, which uses the tools of category theory to simultaneously optimize various components of engineered systems.Abbott says that “this whole field of optimized deep learning models, I believe, is quite critically unaddressed, and that’s why these diagrams are so exciting. They open the doors to a systematic approach to this problem.”“I’m very impressed by the quality of this research. ... The new approach to diagramming deep-learning algorithms used by this paper could be a very significant step,” says Jeremy Howard, founder and CEO of Answers.ai, who was not associated with this work. “This paper is the first time I’ve seen such a notation used to deeply analyze the performance of a deep-learning algorithm on real-world hardware. ... The next step will be to see whether real-world performance gains can be achieved.”“This is a beautifully executed piece of theoretical research, which also aims for high accessibility to uninitiated readers — a trait rarely seen in papers of this kind,” says Petar Velickovic, a senior research scientist at Google DeepMind and a lecturer at Cambridge University, who was not associated with this work. These researchers, he says, “are clearly excellent communicators, and I cannot wait to see what they come up with next!”The new diagram-based language, having been posted online, has already attracted great attention and interest from software developers. A reviewer from Abbott’s prior paper introducing the diagrams noted that “The proposed neural circuit diagrams look great from an artistic standpoint (as far as I am able to judge this).” “It’s technical research, but it’s also flashy!” Zardini says.

The UK Says at an Energy Summit That Green Power Will Boost Security, as the US Differs

Britain has announced a major investment in wind power as it hosts an international summit on energy security

LONDON (AP) — Britain announced a major investment in wind power Thursday as it hosted an international summit on energy security — with Europe and the United States at odds over whether to cut their reliance on fossil fuels.U.K. Prime Minister Keir Starmer said the government will invest 300 million pounds ($400 million) in boosting Britain’s capacity to manufacture components for the offshore wind industry, a move it hopes will encourage private investment in the U.K.’s renewable energy sector.“As long as energy can be weaponized against us, our countries and our citizens are vulnerable and exposed,” U.K. Energy Secretary Ed Miliband told delegates.He said “low-carbon power” was a route to energy security as well as a way to slow climate change.Britain now gets more than half its electricity from renewable sources such as wind and solar power, and the rest from natural gas and nuclear energy. It aims to generate all the U.K.’s energy from renewable sources by 2030.Tommy Joyce, U.S. acting assistant secretary of energy for international affairs, told participants they should be “honest about the world’s growing energy needs, not focused on net-zero politics.”He called policies that push for clean power over fossil fuels "harmful and dangerous," and claimed building wind turbines requires "concessions to or coercion from China" because it supplies necessary rare minerals.Hosted by the British government and the International Energy Agency, the two-day summit brings together government ministers from 60 countries, senior European Union officials, energy sector CEOs, heads of international organizations and nonprofits to assess risks to the global energy system and figure out solutions. Associated Press writer Jennifer McDermott contributed to this story. ___The Associated Press’ climate and environmental coverage receives financial support from multiple private foundations. AP is solely responsible for all content. Find AP’s standards for working with philanthropies, a list of supporters and funded coverage areas at AP.org.Copyright 2025 The Associated Press. All rights reserved. This material may not be published, broadcast, rewritten or redistributed.Photos You Should See - Feb. 2025

Suggested Viewing

Join us to forge
a sustainable future

Our team is always growing.
Become a partner, volunteer, sponsor, or intern today.
Let us know how you would like to get involved!

CONTACT US

sign up for our mailing list to stay informed on the latest films and environmental headlines.

Subscribers receive a free day pass for streaming Cinema Verde.
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.