Optimization remains a fundamental pillar of machine learning, yet existing methods often struggle to maintain stability and adaptability in dynamic, non linear systems, especially under uncertainty. We introduce AERO (Adversarial Energy-based Redirection Optimization), a novel framework inspired by the redirection principle in Judo, where external disturbances are leveraged rather than resisted. AERO reimagines optimization as a redirection process guided by 15 interrelated axioms encompassing adversarial correction, energy conservation, and disturbance-aware learning. By projecting gradients, integrating uncertainty driven dynamics, and managing learning energy, AERO offers a principled approach to stable and robust model updates. Applied to probabilistic solar energy forecasting, AERO demonstrates substantial gains in predictive accuracy, reliability, and adaptability, especially in noisy and uncertain environments. Our findings highlight AERO as a compelling new direction in the theoretical and practical landscape of optimization.

The pervasiveness of proprietary language models has raised privacy concerns for users' sensitive data, emphasizing the need for private inference (PI), where inference is performed directly on encrypted inputs. However, current PI methods face prohibitively higher communication and latency overheads, primarily due to nonlinear operations. In this paper, we present a comprehensive analysis to understand the role of nonlinearities in transformer-based decoder-only language models. We introduce AERO, a four-step architectural optimization framework that refines the existing LLM architecture for efficient PI by systematically removing nonlinearities such as LayerNorm and GELU and reducing FLOPs counts. For the first time, we propose a Softmax-only architecture with significantly fewer FLOPs tailored for efficient PI. Furthermore, we devise a novel entropy regularization technique to improve the performance of Softmax-only models. AERO achieves up to 4.23$\times$ communication and 1.94$\times$ latency reduction. We validate the effectiveness of AERO by benchmarking it against the state-of-the-art.

Bird-sized flapping-wing robots offer significant potential for agile flight in complex environments, but achieving agile and robust trajectory tracking remains a challenge due to the complex aerodynamics and highly nonlinear dynamics inherent in flapping-wing flight. In this work, a learning-based control approach is introduced to unlock the versatility and adaptiveness of flapping-wing flight. We propose a model-free reinforcement learning (RL)-based framework for a high degree-of-freedom (DoF) bird-inspired flapping-wing robot that allows for multimodal flight and agile trajectory tracking. Stability analysis was performed on the closed-loop system comprising of the flapping-wing system and the RL policy. Additionally, simulation results demonstrate that the RL-based controller can successfully learn complex wing trajectory patterns, achieve stable flight, switch between flight modes spontaneously, and track different trajectories under various aerodynamic conditions.

With the development of astronomical facilities, large-scale time series data observed by these facilities is being collected. Analyzing anomalies in these astronomical observations is crucial for uncovering potential celestial events and physical phenomena, thus advancing the scientific research process. However, existing time series anomaly detection methods fall short in tackling the unique characteristics of astronomical observations where each star is inherently independent but interfered by random concurrent noise, resulting in a high rate of false alarms. To overcome the challenges, we propose AERO, a novel two-stage framework tailored for unsupervised anomaly detection in astronomical observations. In the first stage, we employ a Transformer-based encoder-decoder architecture to learn the normal temporal patterns on each variate (i.e., star) in alignment with the characteristic of variate independence. In the second stage, we enhance the graph neural network with a window-wise graph structure learning to tackle the occurrence of concurrent noise characterized by spatial and temporal randomness. In this way, AERO is not only capable of distinguishing normal temporal patterns from potential anomalies but also effectively differentiating concurrent noise, thus decreasing the number of false alarms. We conducted extensive experiments on three synthetic datasets and three real-world datasets. The results demonstrate that AERO outperforms the compared baselines. Notably, compared to the state-of-the-art model, AERO improves the F1-score by up to 8.76% and 2.63% on synthetic and real-world datasets respectively.

From the perspective of technology companies, one of the greatest challenges to developing cutting edge technology is ensuring that the innovation budget receives significant recognition among users, regulators and other stakeholders. Receiving recognition sometimes means publishing papers in peer-reviewed magazines, presenting in conferences and preparing press releases that showcase the advancements in technology. Though companies can sometimes grow to be very "optimistic" of perceived progress, some claims of progress lack the necessary proofs to ensure that developments are relevant. While AI potential applications can be varied and diverse, making sure that a system is sound, scalable, uses available datasets and can perform in a real-world environment is a daunting task. Sometimes the proposed approaches fail in their methodology as the data is simply not available to support industrialisation of the system; or, the system is not safe and confidentiality concerns remain about how the data will be used.

Intel has introduced a quadcopter called Aero at its annual developers conference, and we'll bet you can guess its target audience based on the event. That's right, Aero was specifically designed not for hobbyists or for commercial purposes, but for developers who want to create and test apps for drones. The company said it's the "fastest path available from Intel for developers to get applications airborne." Aero is powered by an Atom processor and comes equipped with Intel's RealSense camera for vision. It's also preloaded with AirMap, an app that tells you where you can and can't fly, gives you real-time info on wildfires and the like, as well as gives you an easy way to plot routes.

Larry Page, the billionaire co-founder of Google, is secretly backing a pair of startups that are working on flying cars, according to a report. Since 2010, Page has poured more than 100 million into Zee.Aero, a company that lately has been testing two flying-car prototypes at an airport hangar in Hollister, Calif., Bloomberg said Thursday, citing sources. Since last year, Page also has been funding another flying-car startup called Kitty Hawk -- and has cast it as a rival to Zee.Aero as he stages a top-secret race to develop a new class of vehicles that can soar above traffic jams and sidestep the hassles of the airport, Bloomberg said. "Page has drawn a line separating his two flying-car teams," the report said. "It's common for the Zee.Aero engineers to speculate over lunch about what their Kitty Hawk counterparts are up to."

Three years ago, Silicon Valley developed a fleeting infatuation with a startup called Zee.Aero. The company had set up shop right next to Google's headquarters in Mountain View, Calif., which was curious, because Google tightly controls most of the land in the area. Then a reporter spotted patent filings showing Zee.Aero was working on a small, all-electric plane that could take off and land vertically--a flying car. In the handful of news articles that ensued, all the startup would say was that it wasn't affiliated with Google or any other technology company. Then it stopped answering media inquiries altogether. Employees say they were even given wallet-size cards with instructions on how to deflect questions from reporters. After that, the only information that trickled out came from amateur pilots, who occasionally posted pictures of a strange-looking plane taking off from a nearby airport. Trump Says'No Reason' to Raise 1 Billion for Campaign Turns out, Zee.Aero doesn't belong to Google or its holding company, Alphabet.

Google CEO Larry Page speaks at a news conference at the Google offices in New York, Monday, May 21, 2012. It appears Alphabet CEO Larry Page isn't only interested in cars that drive themselves. According to Bloomberg, Page has been secretly funding startups specializing in flying cars. One startup in particular, called Zee.Aero, has been funded by Page since 2010, says the report. Zee.Aero, based near Google's headquarters in Mountain View, Calif., had been working on a prototype flying car capable of taking off and landing vertically.

The third approach left the technology codes untouched and built a procedural program that initiated separate, independent processes consisting of the technology codes communicating through a common database. This was better because the technology organizations continued to maintain technical and managerial control over their codes. The rigid procedural integration program was still unacceptably costly to modify, requiring a flow time of approximately six weeks. However, it did provide a prototype and baseline for the knowledge system.