The White House Executive Order on Artificial Intelligence highlights the risks of large language models (LLMs) empowering malicious actors in developing biological, cyber, and chemical weapons. To measure these risks, government institutions and major AI labs are developing evaluations for hazardous capabilities in LLMs. However, current evaluations are private and restricted to a narrow range of malicious use scenarios, which limits further research into reducing malicious use. To fill these gaps, we release the Weapons of Mass Destruction Proxy (WMDP) benchmark, a dataset of 3,668 multiple-choice questions that serve as a proxy measurement of hazardous knowledge in biosecurity, cybersecurity, and chemical security. To guide progress on unlearning, we develop RMU, a state-of-the-art unlearning method based on controlling model representations. RMU reduces model performance on WMDP while maintaining general capabilities in areas such as biology and computer science, suggesting that unlearning may be a concrete path towards reducing malicious use from LLMs. We release our benchmark and code publicly at https://wmdp.ai.

In this article, we introduce a p arallelizing a pproximatio n - d isc o very f ra mework, PANDORA, for automatically discovering application- and architecture-specialized approximations of provided code. PANDORA complements existing compilers and runtime optimizers by generating approximations with a range of Pareto-optimal tradeoffs between performance and error, which enables adaptation to different inputs, different user preferences, and different runtime conditions (e.g., battery life). We demonstrate that PANDORA can create parallel approximations of inherently sequential code by discovering alternative implementations that eliminate loop-carried dependencies. For a variety of functions with loop-carried dependencies, PANDORA generates approximations that achieve speedups ranging from 2.3x to 81x, with acceptable error for many usage scenarios. We also demonstrate PANDORA’s architecture-specialized approximations via FPGA experiments, and highlight PANDORA’s discovery capabilities by removing loop-carried dependencies from a recurrence relation with no known closed-form solution.

In this paper, we introduce PANDORA---a framework that complements existing parallelizing compilers by automatically discovering application- and architecture-specialized approximations. We demonstrate that PANDORA creates approximations that extract massive amounts of parallelism from inherently sequential code by eliminating loop-carried dependencies---a long-time goal of the compiler research community. Compared to exact parallel baselines, preliminary results show speedups ranging from 2.3x to 81x with acceptable error for many usage scenarios.

This paper discusses about what cloud computing means for data management from an industrial perspective, illustrate challenges, opportunities, technologies and impact. We will also ask where cloud computing stands on the hype curve, if we are reinventing the wheel with some technologies, and outline important research questions the data management community should address to move cloud computing to the next level.

Abstract For more than 40 years, the Institute for Scientific Information (ISI, now part of Thomson Reuters) produced the only available bibliographic databases from which bibliometricians could compile large‐scale bibliometric indicators. ISI's citation indexes, now regrouped under the Web of Science (WoS), were the major sources of bibliometric data until 2004, when Scopus was launched by the publisher Reed Elsevier. For those who perform bibliometric analyses and comparisons of countries or institutions, the existence of these two major databases raises the important question of the comparability and stability of statistics obtained from different data sources. This paper uses macrolevel bibliometric indicators to compare results obtained from the WoS and Scopus. It shows that the correlations between the measures obtained with both databases for the number of papers and the number of citations received by countries, as well as for their ranks, are extremely high (R 2 ≈ .99). There is also a very high correlation when countries' papers are broken down by field. The paper thus provides evidence that indicators of scientific production and citations at the country level are stable and largely independent of the database.

Advances in hardware, storage, devices, connectivity, and web technology are changing the way applications are designed, deployed, and managed. Applications are increasingly becoming data-centric and data is everywhere, and in all tiers (from the client to the cloud). Data across multiple tiers requires data access and management capabilities across these tiers.

As the capabilities and service levels of enterprise database systems have evolved, they have collided with incumbent technologies such as TP-Monitors or Message Oriented Middleware (MOM). We believe this trend will continue and have architected the upcoming release of SQL Server to advance this technology trend. This paper describes the Service Oriented Database Architecture (SODA) developed for the Microsoft SQL Server DBMS. First, it motivates the need for building Service Oriented Architecture (SOA) features directly into a database engine. Second, it describes a set of features in SQL Server that have been designed for SOA use. Finally, it concludes with some thoughts on how SODA can enable multiple service deployment topologies.

Summary form only given, as follows. Computing hardware is now available at a price, performance, and reliability point such that enterprise computing tasks may be performed on "commodity" class machines. These economic benefits have been achieved through producing large numbers of systems from standardized components. The OLE Transaction Architecture defines a set of distributed component object model (COM) interfaces that can be used to coordinate distributed transactions using standard two-phase commit protocols. One of the goals of this architecture is to support the development of commodity enterprise computing software. A crisply defined and unambiguous transaction coordination architecture is a foundation for advanced applications requiring distributed transaction support. These applications may be developed and deployed in large numbers, achieving the same economic benefits that standard component hardware has experienced. We describe the motivation for developing the OLE Transaction Architecture, present the programming model for transactions as objects, and describe the implementation of the first transaction and resource Managers based upon the OLE Transaction Architecture.