Tyler O'Neal, Staff Editor

Carnegie Mellon engineers build AI agents to construct useful new designs using visual cues as humans do

Trained AI agents can adopt human design strategies to solve problems, according to findings published in the ASME Journal of Mechanical Design.

Big design problems require creative and exploratory decision making, a skill in which humans excel. When engineers use artificial intelligence (AI), they have traditionally applied it to a problem within a defined set of rules rather than having it generally follow human strategies to create something new. This novel research considers an AI framework that learns human design strategies through observation of human data to generate new designs without explicit goal information, bias, or guidance.

The study was co-authored by Jonathan Cagan, professor of mechanical engineering and interim dean of Carnegie Mellon University's College of Engineering, Ayush Raina, a Ph.D. candidate in mechanical engineering at Carnegie Mellon, and Chris McComb, an assistant professor of engineering design at the Pennsylvania State University.

"The AI is not just mimicking or regurgitating solutions that already exist," said Cagan. "It's learning how people solve a specific type of problem and creating new design solutions from scratch." How good can AI be? "The answer is quite good." A photo of a bridge{module INSIDE STORY}

The study focuses on truss problems because they represent complex engineering design challenges. Commonly seen in bridges, a truss is an assembly of rods forming a complete structure. The AI agents were trained to observe the progression in design modification sequences that had been followed in creating a truss based on the same visual information that engineers use--pixels on a screen--but without further context. When it was the agents' turn to design, they imagined design progressions that were similar to those used by humans and then generated design moves to realize them. The researchers emphasized visualization in the process because vision is an integral part of how humans perceive the world and go about solving problems.

The framework was made up of multiple deep neural networks that worked together in a prediction-based situation. Using a neural network, the AI looked through a set of five sequential images and predicted the next design using the information it gathered from these images.

"We were trying to have the agents create designs similar to how humans do it, imitating the process they use: how they look at the design, how they take the next action and then create a new design, step by step," said Raina.

The researchers tested the AI agents on similar problems and found that on average, they performed better than humans. Yet, this success came without many of the advantages humans have available when they are solving problems. Unlike humans, the agents were not working with a specific goal (like making something lightweight) and did not receive feedback on how well they were doing. Instead, they only used the vision-based human strategy techniques they had been trained to use.

"It's tempting to think that this AI will replace engineers, but that's simply not true," said McComb. "Instead, it can fundamentally change how engineers work. If we can offload boring, time-consuming tasks to an AI, as we did in the work, then we free engineers up to think big and solve problems creatively."

Satellites large datasets are key to understanding ocean carbon in climate change

Tyler O'Neal, Staff Editor

Satellites now play a key role in monitoring carbon levels in the oceans, but we are only just beginning to understand their full potential.

Our ability to predict future climate relies upon being able to monitor where our carbon emissions go. So we need to know how much stays in the atmosphere, or becomes stored in the oceans or on land. The oceans, in particular, have helped to slow climate change as they absorb and then store the carbon for thousands of years.

The IPCC Special Report on the Oceans and Cryosphere in a Changing Climate, published in September, identified this critical role that the ocean play in regulating our climate along with the need to increase our monitoring and understanding of ocean health.

But the vast nature of the oceans, covering over 70% of the Earth's surface, illustrates why satellites are an important component of any monitoring. CAPTION This is Tonga from space. CREDIT Copernicus Sentinel data processed by ESA{module In-article}

The new study, led by the University of Exeter, says that increased exploitation of existing satellites will enable us to fill "critical knowledge gaps" for monitoring our climate.

The work reports that satellites originally launched to study the wind, also have the capability to observe how rain, wind, waves, foam, and temperature all combine to control the movement of heat and carbon dioxide between the ocean and the atmosphere.

Additionally, satellites launched to monitor gas emissions over the land are also able to measure carbon dioxide emissions as they disperse over the ocean.

Future satellite missions offer even greater potential for new knowledge, including the ability to study the internal circulation of the oceans. New constellations of commercial satellites, designed to monitor the weather and life on land, are also capable of helping to monitor ocean health.

"Monitoring carbon uptake by the oceans is now critical to understand our climate and for ensuring the future health of the animals that live there," said lead author Dr. Jamie Shutler, of the Centre for Geography and Environmental Science on Exeter's Penryn Campus in Cornwall.

"By monitoring the oceans we can gather the necessary information to help protect ecosystems at risk and motivate societal shifts towards cutting carbon emissions."

The research team included multiple European research institutes and universities, the US National Oceanic and Atmospheric Administration, the Japan Aerospace Exploration Agency and the European Space Agency.

The researchers call for a "robust network" that can routinely observe the oceans.

This network would need to combine big data from many different satellites with information from automated instruments on ships, autonomous vehicles, and floats that can routinely measure surface water carbon dioxide.

And recent supercomputing advancements, such as Google Earth Engine, which provides free access and supercomputing for scientific analysis of satellite datasets, could also be used. CAPTION This is the Aeolus liftoff. CREDIT ESA, S Corvaja

The study suggests that an international charter that makes satellite data freely available during major disasters should be expanded to include the "long-term man-made climate disaster", enabling commercial satellite operators to easily contribute.

The research was supported by the International Space Science Institute ISSI Bern, Switzerland, and initiated by Dr. Shutler at the University of Exeter and Dr. Craig Donlon at the European Space Agency.

The paper, published this week in Frontiers in Ecology and Environment, is entitled: "Satellites will address critical science priorities for quantifying ocean carbon."

Cornell scientists use AI to help reduce Amazon hydropower dams' carbon footprint

Tyler O'Neal, Staff Editor

A team of scientists has developed a computational model that uses artificial intelligence to find sites for hydropower dams in order to help reduce greenhouse gas emissions.

Hydropower dams can provide large quantities of energy with carbon footprints as low as sources like solar and wind. But because of how they're formed, some dams emit dangerously high levels of greenhouse gases, threatening sustainability goals.

With hundreds of hydropower dams currently proposed for the Amazon basin - an ecologically sensitive area covering more than a third of South America - predicting their greenhouse emissions in advance could be critical for the region, and the planet. {module In-article}

The Cornell University-led team of ecologists, computer scientists and researchers from South American organizations found that achieving low-carbon hydropower requires planning that considers the entire Amazon basin - and favors dams at higher elevations.

"If you develop these dams one at a time without planning strategically - which is how they're usually developed - there is a very low chance that you'll end up with an optimal solution," said Rafael Almeida, postdoctoral research fellow with the Atkinson Center for a Sustainable Future and co-lead of "Reducing Greenhouse Gas Emissions of Amazon Hydropower with Strategic Dam Planning," which published Sept. 19 in Nature Communications.

Using the model, the researchers can identify the combination of dams that would produce the lowest amounts of greenhouse gases for a given energy output target.

When areas are flooded to build dams, decomposing plant matter produces methane, a more destructive greenhouse gas than carbon dioxide. Depending on the location and other factors, the carbon emissions from dam construction can vary from lowest to highest by more than two orders of magnitude.

The analysis found that dams built at high elevations tend to lower greenhouse gas emissions per unit of power output than dams in the lowlands - partly because less land needs to be flooded in steeper environments.

There are currently around 150 hydropower dams and another 350 proposed for the Amazon basin, which encompasses parts of Brazil, Ecuador, Peru and Bolivia. This study is part of a larger effort to use computational tools to analyze the dams' impact to help South American governments and organizations make informed decisions that balance the benefits and disadvantages.