Search

The first International AI Safety Report comprehensively synthesizes the current evidence on the capabilities, risks, and safety of advanced AI systems. The report was mandated by the nations attending the AI Safety Summit in Bletchley, UK. Thirty nations, the UN, the OECD, and the EU each nominated a representative to the report's Expert Advisory Panel. A total of 100 AI experts contributed, representing diverse perspectives and disciplines. Led by the report's Chair, these independent experts collectively had full discretion over the report's content.

Benchmarks are important tools for tracking the rapid advancements in large language model (LLM) capabilities. However, benchmarks are not keeping pace in difficulty: LLMs now achieve over 90\% accuracy on popular benchmarks like MMLU, limiting informed measurement of state-of-the-art LLM capabilities. In response, we introduce Humanity's Last Exam (HLE), a multi-modal benchmark at the frontier of human knowledge, designed to be the final closed-ended academic benchmark of its kind with broad subject coverage. HLE consists of 3,000 questions across dozens of subjects, including mathematics, humanities, and the natural sciences. HLE is developed globally by subject-matter experts and consists of multiple-choice and short-answer questions suitable for automated grading. Each question has a known solution that is unambiguous and easily verifiable, but cannot be quickly answered via internet retrieval. State-of-the-art LLMs demonstrate low accuracy and calibration on HLE, highlighting a significant gap between current LLM capabilities and the expert human frontier on closed-ended academic questions. To inform research and policymaking upon a clear understanding of model capabilities, we publicly release HLE at https://lastexam.ai., Comment: 26 pages, 6 figures

Measuring instructors' adoption of learning management system (LMS) tools is a critical first step in evaluating the efficacy of online teaching and learning at scale. Existing models for LMS adoption are often qualitative, learner-centered, and difficult to leverage towards institutional support. We propose depth-of-use (DOU): an intuitive measurement model for faculty's utilization of a university-wide LMS and their needs for institutional support. We hypothesis-test the relationship between DOU and course attributes like modality, participation, logistics, and outcomes. In a large-scale analysis of metadata from 30000+ courses offered at Virginia Tech over two years, we find that a pervasive need for scale, interoperability and ubiquitous access drives LMS adoption by university instructors. We then demonstrate how DOU can help faculty members identify the opportunity-cost of transition from legacy apps to LMS tools. We also describe how DOU can help instructional designers and IT organizational leadership evaluate the impact of their support allocation, faculty development and LMS evangelism initiatives., Comment: CSCW 2024 (accepted for publication)

Nonlinear microresonators can convert light from chip-integrated sources into new wavelengths within the visible and near-infrared spectrum. For most applications, such as the interrogation of quantum systems with specific transition wavelengths, tuning the frequency of converted light is critical. Nonetheless, demonstrations of wavelength conversion have mostly overlooked this metric. Here, we apply efficient integrated heaters to tune the idler frequency produced by Kerr optical parametric oscillation in a silicon-nitride microring across a continuous 1.5 terahertz range. Finally, we suppress idler frequency noise between DC and 5 kHz by several orders of magnitude using feedback to the heater drive.

Cavity quantum electrodynamics (cQED), the interaction of a two-level system with a high quality factor (Q) cavity, is a foundational building block in different architectures for quantum computation, communication, and metrology. The strong interaction between the atom and the cavity enables single photon operation which is required for quantum gates and sources. Cold atoms, quantum dots, and color centers in crystals are amongst the systems that have shown single photon operations, but they require significant physical infrastructure. Atomic vapors, on the other hand, require limited experimental infrastructure and are hence much easier to deploy outside a laboratory, but they produce an ensemble of moving atoms that results in short interaction times involving multiple atoms, which can hamper quantum operations. A solution to this issue can be found in nanophotonic cavities, where light-matter interaction is enhanced and the volume of operation is small, so that fast single-atom, single-photon operations are enabled. In this work, we study the interaction of an atomically-clad microring resonator (ACMRR) with different-sized ensembles of Rb atoms. We demonstrate strong coupling between an ensemble of ~50 atoms interacting with a high-quality factor (Q > 4 x 10^5) ACMRR, yielding a many-atom cooperativity C ~ 5.5. We continue to observe signatures of atom-photon interaction for a few (< 3) atoms, for which we observe saturation at the level of one intracavity photon. Further development of our platform, which includes integrated thermo-optic heaters to enable cavity tuning and stabilization, should enable the observation of interactions between single photons and single atoms.

The development of broadband microresonator frequency combs at visible wavelengths is pivotal for the advancement of compact and fieldable optical atomic clocks and spectroscopy systems. Yet, their realization necessitates resonators with anomalous dispersion, an arduous task due to the prevailing normal dispersion regime of materials within the visible spectrum. In this work, we evince that silicon nitride microring resonators with air cladding on top and sides -- a deviation from the frequently employed silica-embedded resonators -- allows for the direct generation of broadband microcombs in the visible range. We experimentally demonstrate combs pumped at 1060~nm (283~THz) that reach wavelengths as short as 680~nm (440 THz), and combs pumped at 780~nm (384 THz) that reach wavelengths as short as 630 nm (475 THz). We further show through simulations that microcombs extending to wavelengths as low as 461 nm (650 THz) should be accessible in this platform.

Generative artificial intelligence has the potential to both exacerbate and ameliorate existing socioeconomic inequalities. In this article, we provide a state-of-the-art interdisciplinary overview of the potential impacts of generative AI on (mis)information and three information-intensive domains: work, education, and healthcare. Our goal is to highlight how generative AI could worsen existing inequalities while illuminating how AI may help mitigate pervasive social problems. In the information domain, generative AI can democratize content creation and access, but may dramatically expand the production and proliferation of misinformation. In the workplace, it can boost productivity and create new jobs, but the benefits will likely be distributed unevenly. In education, it offers personalized learning, but may widen the digital divide. In healthcare, it might improve diagnostics and accessibility, but could deepen pre-existing inequalities. In each section we cover a specific topic, evaluate existing research, identify critical gaps, and recommend research directions, including explicit trade-offs that complicate the derivation of a priori hypotheses. We conclude with a section highlighting the role of policymaking to maximize generative AI's potential to reduce inequalities while mitigating its harmful effects. We discuss strengths and weaknesses of existing policy frameworks in the European Union, the United States, and the United Kingdom, observing that each fails to fully confront the socioeconomic challenges we have identified. We propose several concrete policies that could promote shared prosperity through the advancement of generative AI. This article emphasizes the need for interdisciplinary collaborations to understand and address the complex challenges of generative AI., Comment: PNAS Nexus, in press

Higher education offers many benefits to individuals and society--including increased employment opportunities and higher wages--and can help disrupt cycles of intergenerational poverty by improving economic outcomes for those who earn postsecondary credentials. Public community and technical colleges offer a relatively affordable, open-access option for students seeking higher education. Community colleges also provide an array of services to support students along their educational journey, and it is increasingly evident that basic needs supports should be among the services offered. Food insecurity, housing insecurity, and other basic needs insecurities are prevalent among community and technical college students in Washington. In response to these needs, United Way of King County (UWKC) launched the Bridge to Finish campaign to implement a successful basic needs delivery model, provide these services to current students, build evidence to influence policy, and contribute knowledge to the broader conversation on the benefits of a culture of care for community and technical college students. This report shares insights from the collaborative study by providing information on characteristics of Bridge to Finish participants, patterns of service receipt, and rates of successful academic outcomes. The report highlights ways to help Bridge to Finish and other similar programs prepare for a rigorous impact evaluation.

An outstanding challenge for deployable quantum technologies is the availability of high-resolution laser spectroscopy at the specific wavelengths of ultranarrow transitions in atomic and solid-state quantum systems. Here, we demonstrate a powerful spectroscopic tool that synergistically combines high resolution with flexible wavelength access, by showing that nonlinear nanophotonics can be readily pumped with electro-optic frequency combs to enable highly coherent spectral translation with essentially no efficiency loss. Third-order (\c{hi}(3)) optical parametric oscillation in a silicon nitride microring enables nearly a million optical frequency comb pump teeth to be translated onto signal and idler beams; while the comb tooth spacing and bandwidth are adjustable through electro-optic control, the signal and idler carrier frequencies are widely tuneable through dispersion engineering. We then demonstrate the application of these devices to quantum systems, by performing sub-Doppler spectroscopy of the hyperfine transitions of a Cs atomic vapor with our electro-optically-driven Kerr nonlinear light source. The generality, robustness, and agility of this approach as well as its compatibility with photonic integration are expected to lead to its widespread applications in areas such as quantum sensing, telecommunications, and atomic clocks., Comment: 17 pages, 5 figures

Efficient power coupling between on-chip guided and free-space optical modes requires precision spatial mode matching with apodized grating couplers. Yet, grating apodizations are often limited by the minimum feature size of the fabrication approach. This is especially challenging when small feature sizes are required to fabricate gratings at short wavelengths or to achieve weakly scattered light for large-area gratings. Here, we demonstrate a fish-bone grating coupler for precision beam shaping and the generation of millimeter-scale beams at 461 nm wavelength. Our design decouples the minimum feature size from the minimum achievable optical scattering strength, allowing smooth turn-on and continuous control of the emission. Our approach is compatible with commercial foundry photolithography and has reduced sensitivity to both the resolution and the variability of the fabrication approach compared to subwavelength meta-gratings, which often require electron beam lithography., Comment: 10 pages, 6 figures

Adaptation to dynamic conditions requires a certain degree of diversity. If all agents take the best current action, learning that the underlying state has changed and behavior should adapt will be slower. Diversity is harder to maintain when there is fast communication between agents, because they tend to find out and pursue the best action rapidly. We explore these issues using a model of (Bayesian) learning over a social network. Agents learn rapidly from and may also have incentives to coordinate with others to whom they are connected via strong links. We show, however, that when the underlying environment changes sufficiently rapidly, any network consisting of just strong links will do only a little better than random choice in the long run. In contrast, networks combining strong and weak links, whereby the latter type of links transmit information only slowly, can achieve much higher long-run average payoffs. The best social networks are those that combine a large fraction of agents into a strongly-connected component, while still maintaining a sufficient number of smaller communities that make diverse choices and communicate with this component via weak links.

Integrated nonlinear wavelength converters transfer optical energy from lasers or quantum emitters to other useful colors, but chromatic dispersion limits the range of achievable wavelength shifts. Moreover, because of geometric dispersion, fabrication tolerances reduce the accuracy with which devices produce specific target wavelengths. Here, we report nonlinear wavelength converters whose operation is not contingent on dispersion engineering; yet, the output wavelengths are controlled with high accuracy. In our scheme, coherent coupling between counter-propagating waves in a photonic crystal microresonator induces a photonic bandgap that isolates (in dispersion space) specific wavenumbers for nonlinear gain. We first demonstrate the wide applicability of this strategy to parametric nonlinear processes, by simulating its use in third harmonic generation, dispersive wave formation in Kerr microcombs, and four-wave mixing Bragg scattering. In experiments, we demonstrate Kerr optical parametric oscillators in which such wavenumber-selective coherent coupling designates the signal mode. As a result, differences between the targeted and realized signal wavelengths are <0.3 percent. Moreover, leveraging the bandgap-protected wavenumber selectivity, we continuously tune the output frequencies by nearly 300 GHz without compromising efficiency. Our results will bring about a paradigm shift in how microresonators are designed for nonlinear optics, and they make headway on the larger problem of building wavelength-accurate light sources using integrated photonics., Comment: 6 main figures, 1 supplemental figure

In 1990, one in five U.S. workers were aged over 50 years whereas today it is one in three. One possible explanation for this is that occupations have become more accommodating to the preferences of older workers. We explore this by constructing an "age-friendliness" index for occupations. We use Natural Language Processing to measure the degree of overlap between textual descriptions of occupations and characteristics which define age-friendliness. Our index provides an approximation to rankings produced by survey participants and has predictive power for the occupational share of older workers. We find that between 1990 and 2020 around three quarters of occupations have seen their age-friendliness increase and employment in above-average age-friendly occupations has risen by 49 million. However, older workers have not benefited disproportionately from this rise, with substantial gains going to younger females and college graduates and with male non-college educated workers losing out the most. These findings point to the need to frame the rise of age-friendly jobs in the context of other labour market trends and imperfections. Purely age-based policies are insufficient given both heterogeneity amongst older workers as well as similarities between groups of older and younger workers. The latter is especially apparent in the overlapping appeal of specific occupational characteristics.

Dispersion engineering of microring resonators is crucial for optical frequency comb applications, to achieve targeted bandwidths and powers of individual comb teeth. However, conventional microrings only present two geometric degrees of freedom -- width and thickness -- which limits the degree to which dispersion can be controlled. We present a technique where we tune individual resonance frequencies for arbitrary dispersion tailoring. Using a photonic crystal microring resonator that induces coupling to both directions of propagation within the ring, we investigate an intuitive design based on Fourier synthesis. Here, the desired photonic crystal spatial profile is obtained through a Fourier relationship with the targeted modal frequency shifts, where each modal shift is determined based on the corresponding effective index modulation of the ring. Experimentally, we demonstrate several distinct dispersion profiles over dozens of modes in transverse magnetic polarization. In contrast, we find that the transverse electric polarization requires a more advanced model that accounts for the discontinuity of the field at the modulated interface. Finally, we present simulations showing arbitrary frequency comb spectral envelope tailoring using our Frequency synthesis approach.

LCLS-II is currently in its commissioning phase at SLAC. It is an X-ray FEL driven by a CW superconducting LINAC. The beam injector plays a crucial role in the overall performance of the accelerator, and is critical to the final electron beam performance parameters. The LCLS-II injector comprises of a 185.7 MHz VHF copper gun cavity, and a 1.3 GHz two-cell L-band copper buncher cavity. The FPGA-based controller employs feedback and Self-Excited Loop logic in order to regulate the cavity fields. It also features several other functionalities, such as live detune computation, active frequency tracking, and waveform recording. The LLRF system drives the cavities via two 60 kW SSAs through two power couplers, and thus stabilizes the fields inside the plant. This paper provides an outline of the general functionalities of the system, alongside a description of its hardware, firmware and software architecture, before finalizing with the current status of the project and its future goals., Comment: Poster presented at LLRF Workshop 2022 (LLRF2022, arXiv:2208.13680)

Integrated heaters are a basic ingredient within the photonics toolbox, in particular for microresonator frequency tuning through the thermo-refractive effect. Resonators that are fully embedded in a solid cladding (typically SiO\textsubscript{2}) allow for straightforward lossless integration of heater elements. However, air-clad resonators, which are of great interest for short wavelength dispersion engineering and direct interfacing with atomic/molecular systems, do not usually have similarly low loss and efficient integrated heater integration through standard fabrication. Here, we develop a new approach in which the integrated heater is embedded in SiO$_2$ below the waveguiding layer, enabling more efficient heating and more arbitrary routing of the heater traces than possible in a lateral configuration. We incorporate these buried heaters within a stoichiometric Si$_3$N$_4$ process flow that includes high-temperature ($>$1000~$^\circ$C) annealing. Microring resonators with a 1~THz free spectral range and quality factors near 10$^6$ are demonstrated, and the resonant modes are tuned by nearly 1.5~THz, a 5$\times$ improvement compared to equivalent devices with lateral heaters\greg{.} Finally, we demonstrate broadband dissipative Kerr soliton generation in this platform, and show how the heaters can be utilized to aid in bringing relevant lock frequencies within a detectable range.

Contents INTRODUCTION 1 BEYOND THE TURING TEST: HARNESSING AI TO CREATE WIDELY 2 SHARED PROSPERITY Erik Brynjolfsson WE CAN'T PREDICT THE FUTURE OF WORK 4 Adam Thierer AUTOMATION, AI, AND [...]