We introduce Generalized Instruction Tuning (called GLAN), a general and scalable method for instruction tuning of Large Language Models (LLMs). Unlike prior work that relies on seed examples or existing datasets to construct instruction tuning data, GLAN exclusively utilizes a pre-curated taxonomy of human knowledge and capabilities as input and generates large-scale synthetic instruction data across all disciplines. Specifically, inspired by the systematic structure in human education system, we build the taxonomy by decomposing human knowledge and capabilities to various fields, sub-fields and ultimately, distinct disciplines semi-automatically, facilitated by LLMs. Subsequently, we generate a comprehensive list of subjects for every discipline and proceed to design a syllabus tailored to each subject, again utilizing LLMs. With the fine-grained key concepts detailed in every class session of the syllabus, we are able to generate diverse instructions with a broad coverage across the entire spectrum of human knowledge and skills. Extensive experiments on large language models (e.g., Mistral) demonstrate that GLAN excels in multiple dimensions from mathematical reasoning, coding, academic exams, logical reasoning to general instruction following without using task-specific training data of these tasks. In addition, GLAN allows for easy customization and new fields or skills can be added by simply incorporating a new node into our taxonomy.

Power consumption plays an important role in on-device streaming speech recognition, as it has a direct impact on the user experience. This study delves into how weight parameters in speech recognition models influence the overall power consumption of these models. We discovered that the impact of weight parameters on power consumption varies, influenced by factors including how often they are invoked and their placement in memory. Armed with this insight, we developed design guidelines aimed at optimizing on-device speech recognition models. These guidelines focus on minimizing power use without substantially affecting accuracy. Our method, which employs targeted compression based on the varying sensitivities of weight parameters, demonstrates superior performance compared to state-of-the-art compression methods. It achieves a reduction in energy usage of up to 47% while maintaining similar model accuracy and improving the real-time factor.

This story was originally published by Yale E360 and is reproduced here as part of the Climate Desk collaboration. Two months after its release in November 2022, OpenAI's ChatGPT had 100 million active users, and suddenly tech corporations were racing to offer the public more "generative AI" Pundits compared the new technology's impact to the Internet, or electrification, or the Industrial Revolution--or the discovery of fire. Time will sort hype from reality, but one consequence of the explosion of artificial intelligence is clear: this technology's environmental footprint is large and growing. AI use is directly responsible for carbon emissions from non-renewable electricity and for the consumption of millions of gallons of fresh water, and it indirectly boosts impacts from building and maintaining the power-hungry equipment on which AI runs. As tech companies seek to embed high-intensity AI into everything from resume-writing to kidney transplant medicine and from choosing dog food to climate modeling, they cite many ways AI could help reduce humanity's environmental footprint.

Data is required to develop forecasting models for use in Model Predictive Control (MPC) schemes in building energy systems. However, data usage incurs costs from both its collection and exploitation. Determining cost optimal data usage requires understanding of the forecast accuracy and resulting MPC operational performance it enables. This study investigates the performance of both simple and state-of-the-art machine learning prediction models for MPC in a multi-building energy system simulation using historic building energy data. The impact of data usage on forecast accuracy is quantified for the following data efficiency measures: reuse of prediction models, reduction of training data volumes, reduction of model data features, and online model training. A simple linear multi-layer perceptron model is shown to provide equivalent forecast accuracy to state-of-the-art models, with greater data efficiency and generalisability. The use of more than 2 years of training data for load prediction models provided no significant improvement in forecast accuracy. Forecast accuracy and data efficiency were improved simultaneously by using change-point analysis to screen training data. Reused models and those trained with 3 months of data had on average 10% higher error than baseline, indicating that deploying MPC systems without prior data collection may be economic.

In modern feedback control systems theory and practice, reliable access to the dynamically evolving system states is needed at both the implementation stage of advanced control algorithms and for process/system condition and performance monitoring purposes [16, 8, 39, 13, 47]. Traditionally, an explicit use of an available dynamic model complemented by sensor measurements, involving measurable physical and chemical variables of the system of interest, represented a first option to respond to the above need. However, in practice, key critical state variables are often not available for direct on-line measurement, due to inherent physical as well as practically insurmountable technical and economic limitations associated with the current state of sensor technology as it is invariably deployed in cases of considerable system complexity [47, 16, 8, 39]. In light of the above remarks, a better, scientifically sound and practically insightful option is the design of a state estimator (an observer). This is itself an appropriately structured dynamical system itself that utilizes all information provided by a system model as well as available sensor measurements to accurately reconstruct the dynamic profiles of all other unmeasurable state variables [47, 16, 8, 13].

AI systems may have transformative and long-term effects on individuals and society. To manage these impacts responsibly and direct the development of AI systems toward optimal public benefit, considerations of AI ethics and governance must be a first priority. In this workbook, we introduce and describe our PBG Framework, a multi-tiered governance model that enables project teams to integrate ethical values and practical principles into their innovation practices and to have clear mechanisms for demonstrating and documenting this.

Virtual sensing techniques allow for inferring signals at new unmonitored locations by exploiting spatio-temporal measurements coming from physical sensors at different locations. However, as the sensor coverage becomes sparse due to costs or other constraints, physical proximity cannot be used to support interpolation. In this paper, we overcome this challenge by leveraging dependencies between the target variable and a set of correlated variables (covariates) that can frequently be associated with each location of interest. From this viewpoint, covariates provide partial observability, and the problem consists of inferring values for unobserved channels by exploiting observations at other locations to learn how such variables can correlate. We introduce a novel graph-based methodology to exploit such relationships and design a graph deep learning architecture, named GgNet, implementing the framework. The proposed approach relies on propagating information over a nested graph structure that is used to learn dependencies between variables as well as locations. GgNet is extensively evaluated under different virtual sensing scenarios, demonstrating higher reconstruction accuracy compared to the state-of-the-art.

Around the world, ambitious projects are testing everything from seeding clouds with chemicals to pouring artificial whale excrement into the sea. The goal is to remove CO2 from the atmosphere via so called'geoengineering' and'carbon capture' processes - and help to mitigate climate change. Geoengineering sees heat from the sun reflected back into space to limit climate change, while'carbon capture' captures CO2 from the air, either directly or by capturing it in rain among other techniques. The White House cautiously supported further research into an idea straight out of science fiction - 'blocking the sun' to cool the atmosphere - in a report last year. The federally mandated report said that there is'a compelling case for research to better understand both the potential benefits and risks'.

The appendix is subdivided into the following seven topics: A Central angle property: Quick proof of the central angle property, used in Sect. B Optimal Land V-shaped branching: Derivation of the conditions listed in Tab. 1 under which V-or L-branching are optimal. E Non-optimality of higher-degree branchings: Technical proofs and numerical scheme to show the non-optimality of higher-degree branchings discussed in Sect. F BOT on two-dimensional Riemannian manifolds: Formal proof of Theorem 5.1, which generalizes the optimal branching conditions and other properties from the Euclidean plane to embedded surfaces. A sketch of the proof can be found in Sect. G Algorithms: Additional details and experiments for the different algorithms presented in the main paper. Section G.2 focuses on the numerical geometry optimization and Sect. Section G.1 holds a few examples of the recursive geometric construction of relatively optimal solutions for BOT problems with multiple sources. In this section, we present a geometric proof of the central angle property used in the geometric construction of relatively optimal solutions for a given full tree topology (see Sect. 3.2). Below, we formally derive the conditions listed in Tab. 1 under which V-or L-branching provide the optimal solution to a BOT problem with one source and two sinks.

Branched optimal transport (BOT) is a generalization of optimal transport in which transportation costs along an edge are subadditive. This subadditivity models an increase in transport efficiency when shipping mass along the same route, favoring branched transportation networks.