Open Access

In recent times, large language models (LLMs) have made significant strides in generating computer code, blurring the lines between code created by humans and code produced by artificial intelligence (AI). As these technologies evolve rapidly, it is crucial to explore how they influence code generation, especially given the risk of misuse in areas such as higher education. The present paper explores this issue by using advanced classification techniques to differentiate between code written by humans and code generated by ChatGPT, a type of LLM. We employ a new approach that combines powerful embedding features (black-box) with supervised learning algorithms including Deep Neural Networks, Random Forests, and Extreme Gradient Boosting to achieve this differentiation with an impressive accuracy of 98%. For the successful combinations, we also examine their model calibration, showing that some of the models are extremely well calibrated. Additionally, we present white-box features and an interpretable Bayes classifier to elucidate critical differences between the code sources, enhancing the explainability and transparency of our approach. Both approaches work well, but provide at most 85–88% accuracy. Tests on a small sample of untrained humans suggest that humans do not solve the task much better than random guessing. This study is crucial in understanding and mitigating the potential risks associated with using AI in code generation, particularly in the context of higher education, software development, and competitive programming.

Beitrag zur Festschrift für Ursula Georgy

Virtualization fundamentally changes how social relations form, but its effect on network structure in collaborative teams is poorly understood. This paper compares team networks from nine government-funded projects that were conducted virtually because of the COVID-19 pandemic with 15 prepandemic projects from the same funding program. Results of our comparative analysis of 2,746 dyadic ties in 24 teams showed lower levels of network density, clustering, and structural cohesion in virtualized projects, indicating fragmented virtual teams. Furthermore, expressive networks, defined by the sharing of personal information, were affected more than instrumental networks, which revolve around the sharing of expert knowledge.

HIV Gag virus-like particles (HIV Gag VLPs) are promising HIV vaccine candidates. In the literature, they are often described as shear-sensitive particles, and authors usually recommend the operation of tangential flow filtration (TFF) gently at shear rates below 4,000 s −1 to 6,000 s −1 . This in turn poses a severe limitation to the performance of TFF-mediated concentration of VLPs, which would be substantially enhanced by working at higher shear rates. To our knowledge, studies examining the shear sensitivity of HIV Gag VLPs and providing detailed information and evidence for the fragility of these particles have not been conducted yet. Thus, we investigated the effect of high shear rates on the colloidal stability of mosaic VLPs (Mos-VLPs) as relevant examples for HIV Gag VLPs. For this purpose, Mos-VLPs were exposed to different shear rates ranging from 3,395 s −1 to 22, 365 s −1 for 2 h. The average hydrodynamic diameter (AHD) and the polydispersity index (PDI) of the associated particle size distribution were used as stability indicators and measured after the treatment and during storage through dynamic light scattering. At high shear rates, we observed an increase in both AHD and PDI during the storage of HIV Mos1.Gag VLPs (bVLP—without envelope proteins) and Mos1.Gag + Mos2S.Env VLPs (eVLP—with envelope proteins). eVLPs exhibited higher colloidal stability than bVLPs, and we discuss the potential stabilizing role of envelope proteins. We finally demonstrated that the dispersion medium also has a considerable impact on the stability of Mos-VLPs.

In the contemporary era, many organizations and companies are confronted with a signif-icant surge in data volumes. This has led to the challenge of capturing, storing, managing, and analyzing terabytes of data, which are stored in diverse formats and originate from numerous internal and external sources. Furthermore, the emergence of novel applica-tions, such as trading, and artificial intelligence, has made the processing of vast amounts of data in real time an absolute necessity. These requirements exceed the processing ca-pacity of traditional on-disk database management systems, which are ill-equipped to manage this data and to provide real-time results. Therefore, data management requires new solutions to cope with the challenges of data volumes and processing data in real time. An in-memory database system (IMDB- or IMD system) is a database management system that is emerging as a solution to these challenges, with the support of other tech-nologies. IMDBs are capable of processing massive data distinctly faster than traditional database management systems. This work examines the approach of IMDBs, with a par-ticular focus on SAP HANA, and compares it with other IMDBs.

The cleaning of aged silk fibers poses a common challenge in the conservation of textiles, since traditional cleaning techniques often yield unsatisfactory results or even harm objects. In this regard, cleaning objects with laser radiation is a promising addition to the range of available methods. Due to it being contactless, even brittle and touch-sensitive objects with disfiguring or harmful soiling could potentially be cleaned and therefore made accessible for research and presentation. Examples of treatment have sometimes shown spectacular results. Still there is some skepticism concerning the safety of this treatment for textile materials, which has been strengthened through previous 532 nm wavelength nanosecond laser cleaning studies on silk fibers. Taking these published results into account, the range of examined laser parameters has been extended in this study, from 532 nm nanosecond laser to 1064 nm nanosecond and even 800 nm femtosecond laser, reevaluating the effect of this treatment on the fibers. The physicochemical processes taking place on the silk fibers when cleaning with lasers are complex and still not fully understood. The aim of this project was therefore to bring more clarification about potential effects of those processes on the condition of silk samples treated with a set of different parameters for wavelength, pulse duration, energy density and number of pulses per spot. It also looks at the influence of the presence of soiling on the results. The analysis of potential effects was then carried out using statistical methods and advanced analytics. Scanning electron microscopy, Fourier-transform infrared spectroscopy and colorimetry technology provided the required insights to better assess the effects. Results show that laser cleaning of silk fibers, like most other conventional cleaning techniques, is not completely without risk, but knowing what the possible effects are helps making decisions on whether the benefits of the technique used justify these risks.

There is a wide variety of drought indices, yet a consensus on suitable indices and temporal scales for monitoring streamflow drought remains elusive across diverse hydrological settings. Considering the growing interest in spatially distributed indices for ungauged areas, this study addresses the following questions: (i) What temporal scales of precipitation-based indices are most suitable to assess streamflow drought in catchments with different hydrological regimes? (ii) Do soil moisture indices outperform meteorological indices as proxies for streamflow drought? (iii) Are snow indices more effective than meteorological indices for assessing streamflow drought in snow-influenced catchments? To answer these questions, we examined 100 near-natural catchments in Chile with four hydrological regimes, using the standardised precipitation index (SPI), standardised precipitation evapotranspiration index (SPEI), empirical standardised soil moisture index (ESSMI), and standardised snow water equivalent index (SWEI), aggregated across various temporal scales. Cross-correlation and event coincidence analysis were applied between these indices and the standardised streamflow index at a temporal scale of 1 month (SSI-1), as representative of streamflow drought events. Our results underscore that there is not a single drought index and temporal scale best suited to characterise all streamflow droughts in Chile, and their suitability largely depends on catchment memory. Specifically, in snowmelt-driven catchments characterised by a slow streamflow response to precipitation, the SPI at accumulation periods of 12–24 months serves as the best proxy for characterising streamflow droughts, with median correlation and coincidence rates of approximately 0.70–0.75 and 0.58–0.75, respectively. In contrast, the SPI at a 3-month accumulation period is the best proxy over faster-response rainfall-driven catchments, with median coincidence rates of around 0.55. Despite soil moisture and snowpack being key variables that modulate the propagation of meteorological deficits into hydrological ones, meteorological indices are better proxies for streamflow drought. Finally, to exclude the influence of non-drought periods, we recommend using the event coincidence analysis, a method that helps assessing the suitability of meteorological, soil moisture, and/or snow drought indices as proxies for streamflow drought events.

Water serves for the production of pharmaceutical ingredients, intermediates and final products. Accordingly, the quality requirements are particularly high. Next to quality, sustainability of the production and climate change mitigation will play an increasingly important role. For instance, in 2015, the total global emissions of the pharma sector was significantly higher than the CO₂ emissions generated by the automotive sector. Thus, efforts must be made at all stages of production of pharmaceuticals to reduce the environmental impact.

The rapid increase in the use and development of statistical design of experiments (DoE), particularly in pharmaceutical process development, has become increasingly important over the last decades. This rise aligns with Green Chemistry Principles, seeking reduced resource usage and heightened efficiency. In this study, we employed a comprehensive design of experiments (DoE) approach to optimize the catalytic conversion of 1-decene to n-decanal through direct Wacker-type oxidation using the previously determined efficient PdCl2(MeCN)2 catalytic system. The aim was to maximize selectivity and conversion efficiency. Through systematic variation of seven factors, including substrate amount, catalyst and co-catalyst amount, reaction temperature, reaction time, homogenization temperature, and water content, this study identified critical parameters influencing the process to direct the reaction toward the desired product. The statistical analysis revealed high significance for both selectivity and conversion, with surface diagrams illustrating optimal conditions. Notably, catalyst amount emerged as a pivotal factor influencing conversion, with reaction temperature and co-catalyst amount significantly affecting both conversion efficiency and selectivity. The refined model demonstrated strong correlations between predicted and observed values, highlighting the impact of these factors on both selectivity and conversion.

Die Häufigkeit und Intensität globaler Katastrophen nehmen deutlich zu. Vermehrt betreffen deren Auswirkungen auch die Kritische Infrastruktur und führen zu Einschränkungen in der Funktionalität wichtiger Einrichtungen. Insbesondere in Krankenhäusern gibt es viele Prozesse, die es davor zu schützen gilt. Aus diesem Grund soll perspektivisch unter anderem das KRITIS-Dachgesetz auf Grundlage der CER-Richtlinie (EU 2022/2557) bis zum 17. Oktober 2024 verabschiedet werden. Eine Anforderung des KRITIS-Dachgesetztes wird zukünftig unter anderem die Erstellung von Maßnahmen zur Steigerung der Resilienz beinhalten, welche in Resilienzplänen nachgewiesen werden sollen. Die Maßnahmen aus betrieblichen Kontinuitätsstrategien (=Business Continuity Managementsystemen) können integraler Bestandteil der Resilienzpläne sein. In dieser Arbeit sollen Schnittstellen zwischen Business Continuity Managementsystemen und bereits bestehenden Krankenhausalarm- und -einsatzplänen identifiziert werden. Die Erkenntnisse sollen den von dem KRITIS-Dachgesetz betroffenen Krankenhäusern einen Einblick in die Anforderungen von Business Continuity Managementsystemen geben und dazu beitragen, Umsetzungsmöglichkeiten in Krankenhäusern aufzuzeigen. Im Bearbeitungszeitraum wurden 14 qualitative Experteninterviews mit Vertretern von Krankenhäusern, Behörden, juristischen Einrichtungen und internationalen Krisenvorsorgeexperten geführt, von denen 11 in dieser Forschungsarbeit ausgewertet werden. Die Befragungen wurden halbstrukturiert und mit einer explorativen Herangehensweise durchgeführt, um ein Meinungsbild zu den perspektivischen Anforderungen aus dem KRITIS-Dachgesetz zu erfassen, aktuelle Maßnahmen zur Steigerung der Resilienz in Krankenhäusern zu identifizieren und potenzielle Herausforderungen in der Umsetzung zu ermitteln. In den Befragungen wurde eine breite Zustimmung zur Notwendigkeit der Etablierung von Business Continuity Managementsystemen als integraler Bestandteil des Resilienzmanagements von Krankenhäusern festgestellt. Die wesentlichen Erkenntnisse sind die Unterscheidung zwischen ereignisspezifischen und prozessbezogenen Vorgehensweisen bei der Identifizierung von Risiken sowie die erweiterten Anforderungen des Business Continuity Managementsystem hinsichtlich des Wiederanlaufs und der Wiederherstellung von Prozessen und Systemen. Synchronisationspotenziale bietet die Krankenhauseinsatzleitung, welche ebenfalls im Business Continuity Managementsystem mit der besonderen Aufbauorganisation als zentrales Organ zur Ereignisbewältigung beschrieben wird. Ebenfalls können bereits vorhandene Redundanzen in die Anforderungen der betrieblichen Kontinuitätsstrategien übertragen werden. Ein zentrales Hindernis für die Umsetzung in Krankenhäusern kann die branchenspezifische Personalnot und die Insolvenzgefahr darstellen. Die Erkenntnisse wurden genutzt, um konkrete Handlungsempfehlungen für die Krankenhäuser zu entwickeln. Eine frühzeitige Identifizierung von Störungen, Notfällen oder Krisen und die entsprechende Meldung an die verantwortlichen Stellen sind wesentlicher Bestandteil einer erfolgreichen Ereignisbewältigung. Darauf aufbauend wird die Etablierung eines „Leiters KAEP“ empfohlen, welcher unter anderem für die Erstellung und Aktualisierung der Krankenhausalarm- und -einsatzpläne verantwortlich ist und über einen Gesamtüberblick der relevanten Schnittstellen und Abhängigkeiten aller Abteilungen im Krankenhaus verfügt.