Animal behavior is a key window into brain function, yet classical studies compress it into a few numbers — path length, velocity, maze arm entries — collected in restrictive apparatuses. A review in The Innovation Reviews surveys how artificial intelligence is transforming this field. Deep-learning tools now estimate an animal's pose from video, track individuals, recognize actions, and interpret behavior sequences in freely moving animals, enabling analysis in more natural environments. The authors outline a five-stage pipeline from multimodal data acquisition to statistical analysis, and identify persistent obstacles: behavioral definitions lack a standard ruler, and data collected in different laboratories are hard to combine. They propose standardized physical posture models, virtual animals for training data generation, and concepts such as "behavioral nucleotides" to guide future research.

Hydrothermal vents lie in perpetual darkness in the depth of the ocean, yet the cells of the vent mussel Bathymodiolus azoricus, who lives 1,700 metres below the surface, follow a 24-hour rhythm in the laboratory. Researchers from the University of Vienna and the partner institutions IFREMER in France, the University of Oldenburg and the AWI in Germany established cell cultures from these deep-sea mussels, which now allow cellular and molecular insights into the mechanisms of biological timing in the deep sea. The study was published in the journal EMBO Reports. It shows that in the tide-driven environment of the mussels, biological timing can rely on the same genes as the 24-hour clock that regulates land-dwelling organisms.

Idealized carbon dioxide removal would lift economic output almost everywhere on Earth, but the gains would not be shared equally, suggests a modeling study in The Innovation. The team built EconClimNet, a deep learning model that links climate indices to economic output and outperforms both traditional econometric methods and mainstream deep learning baselines. In simulations where CO2 first rises and then is drawn back down, four key climate indices recover more than 80% on average across most of the globe — yet the world remains warmer and wetter than preindustrial. Economic output declines while CO2 climbs and grows once removal begins; in the long run, only Asia and Europe regain their warming-phase losses, and richer countries gain more. The authors stress the scenarios are idealized and frame the results as a quantitative basis for long-term planning, read through the lens of climate justice.

Professor Kyomin Jung’s research team in the Department of Electrical and Computer Engineering at Seoul National University College of Engineering has had a total of 44 papers accepted to leading international artificial intelligence (AI) conferences over the past two years, with 21 papers accepted in 2025 and 23 in 2026.