International audienceEl Niño–Southern Oscillation (ENSO) is a naturally occurring mode of tropical Pacific variability, with global impacts on society and natural ecosystems. While it has long been known that El Niño events display a diverse range of amplitudes, triggers, spatial patterns, and life cycles, the realization that ENSO’s impacts can be highly sensitive to this event-to-event diversity is driving a renewed interest in the subject. This paper surveys our current state of knowledge of ENSO diversity, identifies key gaps in understanding, and outlines some promising future research directions

Summary The recent discovery that normal and neoplastic epithelial cells re-enter the stem-cell state raised an intriguing possibility in the context of cancer pathogenesis: the aggressiveness of carcinomas derives not from their existing content of cancer stem cells (CSCs), but from their proclivity to generate new CSCs from non-CSC populations. Here we demonstrate that non-CSCs of human basal breast cancers are plastic cell populations that readily switch from a non-CSC to CSC-state. The observed cell plasticity is dependent on ZEB1, a key regulator of the epithelial-mesenchymal transition. We find plastic non-CSCs maintain the ZEB1 promoter in a bivalent chromatin configuration enabling them to respond readily to microenvironmental signals, such as TGFbeta. In response, the ZEB1 promoter converts from a bivalent to active chromatin configuration, ZEB1 transcription increases and non-CSCs subsequently enter the CSC state. Our findings support a dynamic model where interconversions between low and high tumorigenic states occur frequently, thereby increasing tumorigenic and malignant potential.

AI is undergoing a paradigm shift with the rise of models (e.g., BERT, DALL-E, GPT-3) that are trained on broad data at scale and are adaptable to a wide range of downstream tasks. We call these models foundation models to underscore their critically central yet incomplete character. This report provides a thorough account of the opportunities and risks of foundation models, ranging from their capabilities (e.g., language, vision, robotics, reasoning, human interaction) and technical principles (e.g., model architectures, training procedures, data, systems, security, evaluation, theory) to their applications (e.g., law, healthcare, education) and societal impact (e.g., inequity, misuse, economic and environmental impact, legal and ethical considerations). Though foundation models are based on standard deep learning and transfer learning, their scale results in new emergent capabilities, and their effectiveness across so many tasks incentivizes homogenization. Homogenization provides powerful leverage but demands caution, as the defects of the foundation model are inherited by all the adapted models downstream. Despite the impending widespread deployment of foundation models, we currently lack a clear understanding of how they work, when they fail, and what they are even capable of due to their emergent properties. To tackle these questions, we believe much of the critical research on foundation models will require deep interdisciplinary collaboration commensurate with their fundamentally sociotechnical nature.

[1] Satellite observations suggest that the intensity of El Niño events in the central equatorial Pacific (CP) has almost doubled in the past three decades, with the strongest warming occurring in 2009-10. This is related to the increasing intensity as well as occurrence frequency of the so-called CP El Niño events since the 1990s. While sea surface temperature (SST) in the CP region during El Niño years has been increasing, those during neutral and La Niña years have not. Therefore, the well-documented warming trend of the warm pool in the CP region is primarily a result of more intense El Niño events rather than a general rise of background SST. Citation: Lee, T., and M.

Abstract. We first describe the principles and practical considerations behind the computer generation of the adjoint to the Massachusetts Institute of Technology ocean general circulation model (GCM) using R. Giering's software tool Tangent-Linear and Adjoint Model Compiler (TAMC). The TAMC's recipe for (FORTRAN-) line-by-line generation of adjoint code is explained by interpreting an adjoint model strictly as the operator that gives the sensitivity of the output of a model to its input. Then, the sensitivity of 1993 annual mean heat transport across 29øN in the Atlantic, to the hydrography on January 1, 1993, is calculated from a global solution of the GCM. The "kinematic sensitivity" to initial temperature variations is isolated, showing how the latter would influence heat transport if they did not affect the density and hence the flow. Over 1 year the heat transport at 29øN is influenced kinematically from regions up to 20 ø upstream in the western boundary current and up to 5 ø upstream in the interior. In contrast, the dynamical influences of initial temperature (and salinity) perturbations spread from as far as the rim of the Labrador Sea to the 29øN section along the western boundary. The sensitivities calculated with the adjoint compare excellently to those from a perturbation calculation with the dynamical model. Perturbations in initial interior salinity influence meridional overturning and heat transport when they have propagated to the western boundary and can thus influence the integrated east-west density difference. Our results support the notion that boundary monitoring of meridional mass and heat transports is feasible. IntroductionThe impending need to synthesize, basin-wide and globally, ocean data such as the entire World Ocean Circulation Experiment (WOCE) data set including altimetry and the surfaceforcing data obtained from weather centers and scatterometers makes imperative the use of sophisticated ocean general circulation models (GCMs) to (1) interpolate in space and time between the observations and (2) diagnose unobservable but important quantities such as vorticity and heat transports. Conversely (and, indeed, prior to all these interpretations), the data stream must be used to test and improve the GCMs (the stated goal 1 of International WOCE). A very powerful and general approach to synthesis is the use of optimization methods; we will concentrate on the particular flavor that has become known as the "adjoint approach" in meteorology and oceanography [e.g., Talagrand