New artificial intelligence-generated images that appear to be one thing, but something else entirely when rotated, are helping scientists test the human mind.

To increase energy efficiency and reduce the carbon footprint of hydrogen fuel production, Fanglin Che, associate professor in the Department of Chemical Engineering at Worcester Polytechnic Institute, is leveraging the power and potential of machine learning and computational modeling. The multi-university team she leads has completed a research study that was just published in Nature Chemical Engineering.

A new study from researchers at MBARI and the University of Miami Rosenstiel School of Marine, Atmospheric, and Earth Science—with an interdisciplinary team of collaborators from the Hakai Institute, Xiamen University, the University of British Columbia, the University of Southern Denmark, and Fisheries and Oceans Canada—has revealed the impacts of marine heatwaves on the foundation of the ocean food web. Pairing data from robotic floats deployed as part of the Global Ocean Biogeochemical Array with eDNA samples from ship-based plankton surveys, the team found that marine heatwaves can reshape ocean food webs, which in turn can slow the transport of carbon to the deep sea and hamper the ocean’s ability to buffer against climate change.

Researchers at Harvard and Northwestern have developed a machine learning method that can design intrinsically disordered proteins with custom properties, addressing nearly 30% of all human proteins that are currently out of reach of AI tools like AlphaFold. The new approach uses automatic differentiation, traditionally a deep learning tool, to optimize protein sequences for desired properties.

University of Arizona researchers have demonstrated that multiphoton microscopy (MPM), combined with machine learning and deep learning techniques, can accurately distinguish pancreatic neuroendocrine neoplasms (PNENs) from normal tissue. By targeting natural fluorescent markers in tissue samples, the team achieved classification accuracies of up to 96.4 percent using convolutional neural networks. Their approach offers a faster alternative to traditional pathology workflows and could improve surgical outcomes by enabling real-time tissue analysis. Future work will explore MPM’s potential for grading tumors and guiding intraoperative decisions.