UC Berkeley professor admits to using AI to edit op-ed on students' math skills Zvezdelina Stankova says she used AI to'help edit' an article about some of her students being'five to eight years' behind A math professor at the University of California, Berkeley, criticizing a "severe" math deficiency among students in an op-ed for the San Francisco Standard, admitted to using artificial intelligence to help edit the piece. The Standard published a 2,000-word piece by Zvezdelina Stankova last week, in which the professor said some of her math students were "five to eight years" behind and lacked a "middle school" education on fractions and basic algebra. Stankova said the UC system's test-blind admissions were to blame, suggesting that students who weren't sufficiently prepared for the rigor of Berkeley's mathematics program were admitted because a longstanding benchmark like the SAT had disappeared. Over the weekend, journalists at Berkeley's student newspaper, the Daily Californian, noticed the op-ed's language sounded like AI . According to Berkeley sophomore Francis Luo, they ran it through AI-detection software Pangram, which claimed 33% of the op-ed had been generated or assisted by AI.

The first anti-AI protester to be jailed has a message for OpenAI, Anthropic and Meta: 'Regain your humanity' Wynd Kaufman, 69, chained and locked the front doors of OpenAI's headquarters last year with members of StopAI A n activist who blocked the entrance to one of the world's biggest AI companies is believed to have become the first person jailed for protesting against artificial intelligence as supporters dub her the "Rosa Parks of AI risk". Wynd Kaufman, 69, surrendered herself on Friday to authorities in San Francisco . She was found guilty by a jury for her role in an action last year that saw members of the group StopAI chain and lock the front doors of OpenAI's headquarters in protest against the pursuit of artificial superintelligence. The retired teacher from Berkeley, California, refused to move from a sit-in protest in February 2025 and pleaded not guilty to multiple misdemeanor charges. She was convicted in June of interfering with a business, trespassing with intent to interfere with a business, unlawful assembly and refusal to disburse a riot.

Things to Do in L.A. Tap to enable a layout that focuses on the article. Omar Yaghi, professor at the University of California, Berkeley, speaks during a media conference in Brussels, Oct. 8, 2025, after being one of three scientists awarded the Nobel Prize in chemistry. This is read by an automated voice. Please report any issues or inconsistencies here . See more from the L.A. Times in Google Search.

Elie observed and recorded the sounds the zebra finches made and classified the calls according to the situation and the bird that made them. Elie observed and recorded the sounds the zebra finches made and classified the calls according to the situation and the bird that made them. A scientist who decoded the dictionary that a bird uses to communicate has won a $100,000 prize for making progress towards a world in which humans can talk to the animals - without being met with a blank response. Dr Julie Elie at the University of California, Berkeley, was awarded the 2026 Coller-Dolittle prize for two-way interspecies communication after working out the 11 core calls in the zebra finch vocabulary and their meanings. Her work revealed how the birds announce who they are and what they are doing, and recognise one another regardless of what they are saying by using individual signatures.

We study the problem of robust estimation under heterogeneous corruption rates, where each sample may be independently corrupted with a known but non-identical probability. This setting arises naturally in distributed and federated learning, crowdsourcing, and sensor networks, yet existing robust estimators typically assume uniform or worst-case corruption, ignoring structural heterogeneity. For mean estimation for multivariate bounded distributions and univariate gaussian distributions, we give tight minimax rates for all heterogeneous corruption patterns. For multivariate gaussian mean estimation and linear regression, we establish the minimax rate for squared error up to a factor of d, where d is the dimension. Roughly, our findings suggest that samples beyond a certain corruption threshold may be discarded by the optimal estimators - this threshold is determined by the empirical distribution of the corruption rates given.

Are you married to a man who's obsessed with AI? If i had to listen to another minute of my husband talking about Claude Code, I might have actually died. It was 11 pm in Berkeley, California, where I was home alone with our 10-month-old daughter, and 2 am in Cambridge, Massachusetts, where he was visiting for his newish job in AI. "JUST LOOK AT THIS!" he shouted. The FaceTime camera zoomed toward a laptop sitting on a hotel bed. I still had to take the dog out. "ARE YOU LOOKING?" he shouted again. I was looking at our real baby. There are two babies in this household now: the small human one and the large language model.

In this work, we consider the problem of matrix sensing over graphs (MSoG). As a general case of matrix completion and matrix sensing problems, the MSoG problem has not been analyzed in the literature and the existing results cannot be directly applied to the MSoG problem. This work provides the first theoretical results on the optimization landscape of the MSoG problem. More specifically, we propose a new condition, named the Ω-RIP condition, to characterize the optimization complexity of the problem. In addition, with an improved regularizer of the incoherence, we prove that the strict saddle property holds for the MSoG problem with high probability under the incoherence condition and the Ω-RIP condition, which guarantees the polynomial-time global convergence of saddleavoiding methods. Compared with state-of-the-art results, the bounds in this work are tight up to a constant. Besides the theoretical guarantees, we numerically illustrate the close relation between the Ω-RIP condition and the optimization complexity.

Multi-agent behavior modeling and trajectory forecasting are crucial for the safe navigation of autonomous agents in interactive scenarios. Variational Autoencoder (VAE) has been widely applied in multi-agent interaction modeling to generate diverse behavior and learn a low-dimensional representation for interacting systems. However, existing literature did not formally discuss if a VAE-based model can properly encode interaction into its latent space. In this work, we argue that one of the typical formulations of VAEs in multi-agent modeling suffers from an issue we refer to as social posterior collapse, i.e., the model is prone to ignoring historical social context when predicting the future trajectory of an agent. It could cause significant prediction errors and poor generalization performance.

We study semisupervised mean estimation with a small labeled sample, a large unlabeled sample, and a black-box prediction model whose output may be miscalibrated. A standard approach in this setting is augmented inverse-probability weighting (AIPW) [Robins et al., 1994], which protects against prediction-model misspecification but can be inefficient when the prediction score is poorly aligned with the outcome scale. We introduce Calibrated Prediction-Powered Inference, which post-hoc calibrates the prediction score on the labeled sample before using it for semisupervised estimation. This simple step requires no retraining and can improve the original score both as a predictor of the outcome and as a regression adjustment for semisupervised inference. We study both linear and isotonic calibration. For isotonic calibration, we establish first-order optimality guarantees: isotonic post-processing can improve predictive accuracy and estimator efficiency relative to the original score and simpler post-processing rules, while no further post-processing of the fitted isotonic score yields additional first-order gains. For linear calibration, we show first-order equivalence to PPI++. We also clarify the relationship among existing estimators, showing that the original PPI estimator is a special case of AIPW and can be inefficient when the prediction model is accurate, while PPI++ is AIPW with empirical efficiency maximization [Rubin et al., 2008]. In simulations and real-data experiments, our calibrated estimators often outperform PPI and are competitive with, or outperform, AIPW and PPI++. We provide an accompanying Python package, ppi_aipw, at https://larsvanderlaan.github.io/ppi-aipw/.

Michael Dell and Susan Dell became the first donors to give more than $1 billion to the University of Texas at Austin, funding a massive AI-native hospital and research campus.