Large language models (LLMs) have demonstrated enhanced performance through the \textit{Thinking then Responding} paradigm, where models generate internal thoughts before final responses (aka, System 2 thinking). However, existing research lacks a systematic understanding of the mechanisms underlying how thinking patterns affect performance across model sizes. In this work, we conduct a comprehensive analysis of the impact of various thinking types on model performance and introduce ThinkPatterns-21k, a curated dataset comprising 21k instruction-response pairs (QA) collected from existing instruction-following datasets with five thinking types. For each pair, we augment it with five distinct internal thinking patterns: one unstructured thinking (monologue) and four structured variants (decomposition, self-ask, self-debate and self-critic), while maintaining the same instruction and response. Through extensive evaluation across different model sizes (3B-32B parameters), we have two key findings: (1) smaller models (<30B parameters) can benefit from most of structured thinking patterns, while larger models (32B) with structured thinking like decomposition would degrade performance and (2) unstructured monologue demonstrates broad effectiveness across different model sizes. Finally, we released all of our datasets, checkpoints, training logs of diverse thinking patterns to reproducibility, aiming to facilitate further research in this direction.

This study explores object detection in historical aerial photographs of Namibia to identify long-term environmental changes. Specifically, we aim to identify key objects -- Waterholes, Omuti homesteads, and Big trees -- around Oshikango in Namibia using sub-meter gray-scale aerial imagery from 1943 and 1972. In this work, we propose a workflow for analyzing historical aerial imagery using a deep semantic segmentation model on sparse hand-labels. To this end, we employ a number of strategies including class-weighting, pseudo-labeling and empirical p-value-based filtering to balance skewed and sparse representations of objects in the ground truth data. Results demonstrate the benefits of these different training strategies resulting in an average $F_1=0.661$ and $F_1=0.755$ over the three objects of interest for the 1943 and 1972 imagery, respectively. We also identified that the average size of Waterhole and Big trees increased while the average size of Omuti homesteads decreased between 1943 and 1972 reflecting some of the local effects of the massive post-Second World War economic, agricultural, demographic, and environmental changes. This work also highlights the untapped potential of historical aerial photographs in understanding long-term environmental changes beyond Namibia (and Africa). With the lack of adequate satellite technology in the past, archival aerial photography offers a great alternative to uncover decades-long environmental changes.

Understanding time is a pivotal aspect of human cognition, crucial in the broader framework of grasping the intricacies of the world. Previous studies typically focus on specific aspects of time, lacking a comprehensive temporal reasoning benchmark. To address this issue, we propose TimeBench, a comprehensive hierarchical temporal reasoning benchmark that covers a broad spectrum of temporal reasoning phenomena, which provides a thorough evaluation for investigating the temporal reasoning capabilities of large language models. We conduct extensive experiments on popular LLMs, such as GPT-4, LLaMA2, and Mistral, incorporating chain-of-thought prompting. Our experimental results indicate a significant performance gap between the state-of-the-art LLMs and humans, highlighting that there is still a considerable distance to cover in temporal reasoning. We aspire for TimeBench to serve as a comprehensive benchmark, fostering research in temporal reasoning for LLMs. Our resource is available at https://github.com/zchuz/TimeBench

Next to insects, birds sadly seem to get short shrift from humans. We remain powerfully drawn to scenes of lions hunting in the Kalahari desert or rhinos jousting in eastern India, but remain mostly oblivious to vibrant scenes of life and love enacted under our very noses. Also: AI might enable us to talk to animals soon. The cacophony of evening traffic as birds stream back into their nests, the diligence of solitary hunters of garden worms, the urgent shrieks of young parents, the furious battles among rival suitors -- these are all everyday dramas that are enacted everywhere around you, and you don't have to fork out a small fortune to observe. However, not knowing what you're looking at or listening to can be frustrating.

This book constitutes the refereed post-conference proceedings of the First International Conference on Pan-African Intelligence and Smart Systems, PAAISS 2021, which was held in Windhoek, Namibia, in September 2021. The 17 revised full papers presented were carefully selected from 41 submissions. The theme of PAAISS 2021 was "Advancing AI research in Africa" and the papers are arranged according to subject areas: Deep Learning; Classification and Pattern Recognition; Neural Networks and Support Vector Machines; Smart Systems.

Artificial intelligence, alongside proper training and education, can manage even the worst of security breaches into a positive outcome for airports and their users, says Kristina Dores, Chief, Aerodromes & Ground Aids at Namibia Civil Aviation Authority, and Brad Hayes, CTO at Circadence Corporation. However, the key question is when (not if) will organisations take the steps to prepare for the coming wave of digitisation? Highly-interconnected and increasingly-digitised systems are a necessary part of modern airport infrastructure. Furthermore, vulnerabilities at these interfaces – through personnel and digital systems alike – lead to an increased threat of intrusion and potentially catastrophic disruption. This problem is not one that we can simply train and hire our way out of as these systems and their attack surfaces do not scale linearly in complexity.

An algorithm highlights obvious animals in blue and possible animals in yellow. After a promising first run in Namibia, a Swiss project could aid savanna conservation using drones and automatic image analysis. To get a sense of how many animals live in a given area, game counts are typically done in real time by sharp-eyed people in vehicles. A project funded by the Swiss National Science Foundation (SNSF) uses drones and artificial intelligence (AI) to count wild animals more efficiently. "Human eyes are very good at detecting animals, but not at screening countless images. Computers can process a lot more data," explains Swiss geo-information specialist Devis Tuia, who received a personal grant from SNSF to form a lab to improve wildlife monitoring methods in places like Namibia.

On a visit to one Harvard robotics lab, the sewing machines stand out, while the head of another explains how and why lab members are studying termites in Namibia. Welcome to the new age of machines, in which scientists with seemingly disparate talents are using cutting-edge materials, cheap sensors, 3-D printing, and powerful computers to accelerate advances in robotics. Prior innovations transformed the factory and warehouse, but those robots work best in controlled environments, usually out of public view. For researchers at Harvard and elsewhere, one new target is Main Street. "We were promised these things by sci-fi for 50 years," said Robert Wood, Charles River Professor of Engineering and Applied Sciences.

Newswire) Investorideas.com, a global news source and investor resource covering actively traded sectors announces this week's additions to its membership global stock directories in technology, mining, energy, biotech and marijuana/hemp. The biggest addition is an entire new section to the Tech Stocks lists featuring Artificial intelligence (AI) and Robotics companies. Robotics have been with us for some time now - assisting with simple chores (like the Roomba vacuum) all the way though space flight and to police bomb squad and military uses. Some names in our list will be recognizable as having been in the tech sector for some time and robotics/AI is just another branch for them - companies like Amazon, Apple, Google, Microsoft and Facebook that have become an everyday part of our lives. Also included are assembly-line robotics companies and companies making robotic parts, all the way to makers of machine vision technologies and automotive intelligence tech.