In this interview series, we're meeting some of the AAAI/SIGAI Doctoral Consortium participants to find out more about their research. In this latest interview, Benyamin Tabarsi tells us about his research at the intersection of generative AI and computing education. We find out more about what he's investigated so far during his PhD, what is particularly interesting about this research area, and what inspired him to undertake a PhD in the field. I'm a computer science student at North Carolina (NC) State University, and my research focuses on computing education and generative AI. I've always been passionate about finding ways to make learning easier for students and teaching more efficient for instructors, especially in computer science.

To create this year's TIME100 AI cover, artist Refik Anadol, who is included on this year's list, trained his studio's AI system on an archive containing each of TIME's more than 5,000 covers to date, spanning over 100 years. The resulting abstract visualization--featuring Anadol's signature flowing, molecular aesthetic--represents the AI "dreaming" about a century of TIME's visual history, he says. Dubbed the Large Nature Model by internationally renowned Turkish-American media artist Anadol and his team, his modular multimodal AI system is the product of extensive research and collaboration. According to Anadol's studio, the model was trained on "the most extensive, ethically collected dataset of the natural world," combining over half a billion images from the archives of organizations including the National Geographic Society, the Smithsonian Institution, and London's Natural History Museum with data collected directly from 16 rainforests. Anadol, whose work has been exhibited at institutions including the Museum of Modern Art (MoMA) in New York, London's Serpentine Galleries, and the Guggenheim Museum Bilbao also worked with tech giants Nvidia and Google Cloud, which provided computing resources, while models such as Meta's Llama and Google's Gemini play a range of roles under the hood.

"We're not comfortable revealing that for various reasons," Dean told me on our call. The total number is an abstract measure that changes over time, he says, adding that the company wants users to be thinking about the energy usage per prompt. But there are people out there all over the world interacting with this technology, not just me--and what we all add up to seems quite relevant. OpenAI does publicly share its total, sharing recently that it sees 2.5 billion queries to ChatGPT every day. So for the curious, we can use this as an example and take the company's self-reported average energy use per query (0.34 watt-hours) to get a rough idea of the total for all people prompting ChatGPT.

Nvidia's sophisticated chips have been an important part of the AI boom. On Wednesday it said demand for its products remains strong, especially from big tech firms including Instagram-owner Meta, and ChatGPT-maker OpenAI, as they race to build-out AI. "The AI race is now on," said Nvidia boss Jensen Huang in a call with analysts following the report's release, saying spending from four big tech firms had doubled to 600bn per year. "Over time, you would think that artificial intelligence would... accelerate GDP growth," Huang said. "Our contribution to that is a large part of the AI infrastructure." Colleen McHugh, chief investment officer at investment firm Wealthify, told the BBC's Today programme Nvidia was "at the heart of this AI boom".

Normalizing Flows provide a principled framework for high-dimensional density estimation and generative modeling by constructing invertible transformations with tractable Jaco-bian determinants. We propose Fractal Flow, a novel normalizing flow architecture that enhances both expressiveness and interpretability through two key innovations. First, we integrate Kolmogorov-Arnold Networks and incorporate Latent Dirichlet Allocation into normalizing flows to construct a structured, interpretable latent space and model hierarchical semantic clusters. Second, inspired by Fractal Generative Models, we introduce a recursive modular design into normalizing flows to improve transformation interpretability and estimation accuracy. Experiments on MNIST, FashionMNIST, CIFAR-10, and geophysical data demonstrate that the Fractal Flow achieves latent clustering, controllable generation, and superior estimation accuracy.

Accurate sarcopenia diagnosis via ultrasound remains challenging due to subtle imaging cues, limited labeled data, and the absence of clinical context in most models. We propose MedVQA-TREE, a multimodal framework that integrates a hierarchical image interpretation module, a gated feature-level fusion mechanism, and a novel multi-hop, multi-query retrieval strategy. The vision module includes anatomical classification, region segmentation, and graph-based spatial reasoning to capture coarse, mid-level, and fine-grained structures. A gated fusion mechanism selectively integrates visual features with textual queries, while clinical knowledge is retrieved through a UMLS-guided pipeline accessing PubMed and a sarcopenia-specific external knowledge base. MedVQA-TREE was trained and evaluated on two public MedVQA datasets (VQA-RAD and PathVQA) and a custom sarcopenia ultrasound dataset. The model achieved up to 99% diagnostic accuracy and outperformed previous state-of-the-art methods by over 10%. These results underscore the benefit of combining structured visual understanding with guided knowledge retrieval for effective AI-assisted diagnosis in sarcopenia.

We present a new efficient hybrid parameter estimation method based on the idea, that if nonlinear dynamic models are stated in terms of a system of equations that is linear in terms of the parameters, then regularized ordinary least squares can be used to estimate these parameters from time series data. We introduce the term "Physics-Informed Regression" (PIR) to describe the proposed data-driven hybrid technique as a way to bridge theory and data by use of ordinary least squares to efficiently perform parameter estimation of the model coefficients of different parameter-linear models; providing examples of models based on nonlinear ordinary equations (ODE) and partial differential equations (PDE). The focus is on parameter estimation on a selection of ODE and PDE models, each illustrating performance in different model characteristics. For two relevant epidemic models of different complexity and number of parameters, PIR is tested and compared against the related technique, physics-informed neural networks (PINN), both on synthetic data generated from known target parameters and on real public Danish time series data collected during the COVID-19 pandemic in Denmark. Both methods were able to estimate the target parameters, while PIR showed to perform noticeably better, especially on a compartment model with higher complexity. Given the difference in computational speed, it is concluded that the PIR method is superior to PINN for the models considered. It is also demonstrated how PIR can be applied to estimate the time-varying parameters of a compartment model that is fitted using real Danish data from the COVID-19 pandemic obtained during a period from 2020 to 2021. The study shows how data-driven and physics-informed techniques may support reliable and fast -- possibly real-time -- parameter estimation in parameter-linear nonlinear dynamic models.

Most machine learning models are designed to maximize predictive accuracy. In this work, we explore a different goal: building classifiers that are interesting. An ``interesting classifier'' is one that uses unusual or unexpected features, even if its accuracy is lower than the best possible model. For example, predicting room congestion from CO2 levels achieves near-perfect accuracy but is unsurprising. In contrast, predicting room congestion from humidity is less accurate yet more nuanced and intriguing. We introduce EUREKA, a simple framework that selects features according to their perceived interestingness. Our method leverages large language models to rank features by their interestingness and then builds interpretable classifiers using only the selected interesting features. Across several benchmark datasets, EUREKA consistently identifies features that are non-obvious yet still predictive. For example, in the Occupancy Detection dataset, our method favors humidity over CO2 levels and light intensity, producing classifiers that achieve meaningful accuracy while offering insights. In the Twin Papers dataset, our method discovers the rule that papers with a colon in the title are more likely to be cited in the future. We argue that such models can support new ways of knowledge discovery and communication, especially in settings where moderate accuracy is sufficient but novelty and interpretability are valued.

A track circuit is an electrical system that detects the presence of a train on the tracks by passing a current through the rails, which acts as a conductor. In its initial form, track circuits consisted of a battery and a relay with adjustable resistors to set the transmitted signal gain and receiver operating point. Sections of track are electrically isolated by insulated joints in each rail. The transmitted signal travels through a single rail, through the relay at the opposite end, then returning to the transmitter through the other rail. Track circuits follow the closed loop principle, which means that any failure results in the safest state (track occupied) as shown in Figure 1. Because of this, track circuits also provide detection of broken rails.Figure 1: Track circuit behaviour schema Nowadays, there are many types of track circuits. The last state of the art ones provide enhanced performance, integrating sophisticated signalling systems to improve operation and safety. Track-circuit failures have an important impact as they imply a stop of operations and an economic impact for both the railway operator and its customers (1).

Foundation models have recently emerged as powerful feature extractors in computational pathology, yet they typically omit mechanisms for leveraging the global spatial structure of tissues and the local contextual relationships among diagnostically relevan t regions -- key elements for understanding the tumor microenvironment. Multiple instance learning (MIL) remains an essential next step following foundation model, designing a framework to aggregate patch - level features into slide - level predictions. We presen t EAGLE - Net, a structure - preserving, attention - guided MIL architecture designed to augment prediction and interpretability. EAGLE - Net integrates multi - scale absolute spatial encoding to capture global tissue architecture, a top - K neighborhood - aware loss to focus attention on local microenvironments, and background suppression loss to minimize false positives. We benchmarked EAGLE - Net on large pan - cancer datasets, including three cancer types for classification (10,260 slides) and seven cancer types for surv ival prediction (4,172 slides), using three distinct histology foundation backbones (REMEDIES, Uni - V1, Uni2 - h). Across tasks, EAGLE - Net achieved up to 3% higher classification accuracy and the top concordance indices in 6 of 7 cancer types, producing smoot h, biologically coherent attention maps that aligned with expert annotations and highlighted invasive fronts, necrosis, and immune infiltration. These results position EAGLE - Net as a generalizable, interpretable framework that complements foundation models, enabling improved biomarker discovery, prognostic modeling, and clinical decision support.