Search

Growing concerns over negligent or malicious uses of AI have increased the appetite for tools that help manage the risks of the technology. In 2018, licenses with behaviorial-use clauses (commonly referred to as Responsible AI Licenses) were proposed to give developers a framework for releasing AI assets while specifying their users to mitigate negative applications. As of the end of 2023, on the order of 40,000 software and model repositories have adopted responsible AI licenses licenses. Notable models licensed with behavioral use clauses include BLOOM (language) and LLaMA2 (language), Stable Diffusion (image), and GRID (robotics). This paper explores why and how these licenses have been adopted, and why and how they have been adapted to fit particular use cases. We use a mixed-methods methodology of qualitative interviews, clustering of license clauses, and quantitative analysis of license adoption. Based on this evidence we take the position that responsible AI licenses need standardization to avoid confusing users or diluting their impact. At the same time, customization of behavioral restrictions is also appropriate in some contexts (e.g., medical domains). We advocate for ``standardized customization'' that can meet users' needs and can be supported via tooling.

In this work, we explore the structure of single-wall boron nanotubes with large diameters (about 21~{\AA}) and a broad range of surface densities of atoms. The computations are done using an evolutionary approach combined with a nearest neighbors model Hamiltonian. For the most stable nanotubes, the number of 5-coordinated boron atoms is about $63\%$ of the total number of atoms forming the nanotubes, whereas about $11\%$ are boron vacancies. For hole densities smaller than about 0.22, the boron nanotubes exhibit randomly distributed hexagonal holes and are more stable than a flat stripe structure and a quasi-flat B$_{36}$ cluster. For larger hole densities ($> 0.22$) the boron nanotubes resemble porous tubular structures with hole sizes that depend on the surface densities of boron atoms., Comment: 10 pages, 3 figures

This work describes the implementation of a genetic algorithm-based strategy combined with first-principles computations for identifying the structure of the most stable boron 1D structures. We focus our attention on the structure of ultrathin 1D boron structures given the lack of previous experimental and theoretical work on this topic. Our methodology yields reasonable structural candidates for further optimizations at the DFT level with tighter convergence criteria. The simulations involved 1D structures with up to 8 atoms per unit cell. We have identified four main groups of structures: flat nanowires (monatomic-height stripes) with triangular or triangular and "square" motifs, stripes with larger holes, nanowires with an open tubular shape, and regular nanowires. The diameter-dependent structural changes are discussed., Comment: 9 pages, 3 figures

This work discusses single-objective constrained genetic algorithm with floating-point, integer, binary and permutation representation. Floating-point genetic algorithm tuning with use of test functions is done and leads to a parameterization with comparatively outstanding performance., Comment: 10 pages, 1 figure, 2 tables. For associated software repository, see https://github.com/ttarkowski/quile

In this work, we foresee the structure of a new class of borophenes with smaller 2D densities of atoms than those explored so far for 2D boron crystals. Boron atoms in the porous borophenes tend to be $5$-coordinated in contrast to commonly investigated structures with hexagonal holes for which the number of nearest neighbors of each atom varies from $3$ to $6$. High metallic character is the usual feature of borophenes, however, we have also identified a semimetallic borophene that turns into semiconductor upon unit cell expansion. A $10\%$ increase in lattice constant of this structure gives rise to an energy gap of $0.3\,\textnormal{eV}$. This extends to semiconductor industry the possible application of 2D boron crystals., Comment: 7 pages, 7 figures, 2 tables

Myerson first introduced graph-restricted games in order to model the interaction of cooperative players with an underlying communication network. A dedicated solution concept -- the Myerson value -- is perhaps the most important normative solution concept for cooperative games on graphs. Unfortunately, its computation is computationally challenging. In particular, although exact algorithms have been proposed, they must traverse all connected coalitions of the graph of which there may be exponentially many. In this paper, we consider the issue of approximating the Myerson value for arbitrary graphs and characteristic functions. While Monte Carlo approximations have been proposed for the related concept of the Shapley value, their suitability for the Myerson value has not been studied. Given this, we evaluate and compare (both theoretically and empiraclly) three Monte Carlo sampling methods for the Myerson value: conventional method of sampling permutations; a new, hybrid algorithm that combines exact computations and sampling; and sampling of connected coalitions. We find that our hybrid algorithm performs very well and also significantly improves on the conventional methods.

Predicting edges in networks is a key problem in social network analysis and involves reasoning about the relationships between nodes based on the structural properties of a network. In particular, link prediction can be used to analyse how a network will develop or - given incomplete information about relationships - to discover "missing" links. Our approach to this problem is rooted in cooperative game theory, where we propose a new, quasi-local approach (i.e., one which considers nodes within some radius k) that combines generalised group closeness centrality and semivalue interaction indices. We develop fast algorithms for computing our measure and evaluate it on a number of real-world networks, where it outperforms a selection of other state-of-the-art methods from the literature. Importantly, choosing the optimal radius k for quasi-local methods is difficult, and there is no assurance that the choice is optimal. Additionally, when compared to other quasi-local methods, ours achieves very good results even when given a suboptimal radius k as a parameter.

Game-theoretic centrality is a flexible and sophisticated approach to identify the most important nodes in a network. It builds upon the methods from cooperative game theory and network theory. The key idea is to treat nodes as players in a cooperative game, where the value of each coalition is determined by certain graph-theoretic properties. Using solution concepts from cooperative game theory, it is then possible to measure how responsible each node is for the worth of the network. The literature on the topic is already quite large, and is scattered among game-theoretic and computer science venues. We review the main game-theoretic network centrality measures from both bodies of literature and organize them into two categories: those that are more focused on the connectivity of nodes, and those that are more focused on the synergies achieved by nodes in groups. We present and explain each centrality, with a focus on algorithms and complexity.

The effect of external static charging on borophenes - 2D boron crystals - is investigated by using first principles calculations. The influence of the excess negative charge on the stability of the 2D structures is examined using a very simple analysis of decomposition of the binding energy of a given boron layer into contributions coming from boron atoms that have different coordination numbers. This analysis is important to understand how the local neighbourhood of an atom influences the overall stability of the monolayer structure. The decomposition is done for the $\alpha$-sheet and its related family of structures. From this analysis, we have found a preference for 2D boron crystals with very small or very high charges per atom. The structures with intermediate charges are energetically not favourable. We have also found a clear preference in terms of binding energy for the experimentally seen $\gamma$-sheet and $\delta$-sheet structures that is almost independent on the considered excess of negative charge of the structures. On the other hand, we have shown that a model based solely on nearest-neighbour interactions, although instructive, is too simple to predict binding energies accurately., Comment: 6 pages, 6 figures, and one table