Abstract
In everyday conversations, humans can take on different roles and adapt their vocabulary to their chosen roles. We explore whether LLMs can take on, that is impersonate, different roles when they generate text in-context. We ask LLMs to assume different personas before solving vision and language tasks. We do this by prefixing the prompt with a persona that is associated either with a social identity or domain expertise. In a multi-armed bandit task, we find that LLMs pretending to be children of different ages recover human-like developmental stages of exploration. In a language-based reasoning task, we find that LLMs impersonating domain experts perform better than LLMs impersonating non-domain experts. Finally, we test whether LLMs' impersonations are complementary to visual information when describing different categories. We find that impersonation can improve performance: an LLM prompted to be a bird expert describes birds better than one prompted to be a car expert. However, impersonation can also uncover LLMs' biases: an LLM prompted to be a man describes cars better than one prompted to be a woman. These findings demonstrate that LLMs are capable of taking on diverse roles and that this in-context impersonation can be used to uncover their strengths and hidden biases. Our code is available at https://github.com/ExplainableML/in-context-impersonation.

Abstract
METIS SCAO uses a wavefront control concept that deploys a 2-stage spatial reconstruction where the wavefront is first reconstructed on an intermediate space we call the virtual DM, and then projected onto the actual control space. This document addresses the projection of the wavefront estimation on the virtual deformable mirror (VDM) onto the control modes developed for METIS (Mid-infrared ELT Imager and Spectrograph). We present a new approach to project onto the control modes using an intermediate regularized projection on the M4 mirror and then convert to modes. This method enables us to utilise all modes for the projection and control in a stable manner, achieving high Strehl ratios for a wide range of conditions without the need for complex parameter tuning.

Abstract
The list of Adaptive Optics (AO) simulators in the community has constantly been growing, guided by different needs and purposes (Compass, HCIPY, OOMAO, SOAPY, YAO. . .). In this paper, we present OOPAO (Object Oriented Python Adaptive Optics), a simulation tool based on the Matlab distribution OOMAO to adapt its philosophy to the Python language. This code was initially intended for internal use but the choice was made to make it public as it can benefit the community since it is fully developed in Python. The OOPAO repository is available in free access on GitHub (https://github.com/cheritier/OOPAO) with several tutorials. The tool consists of a full end-to-end simulator designed for AO analysis purposes. The principle is that the light from a given light source can be propagated through multiple objects (Atmosphere, Telescope, Deformable Mirror, Wave-Front Sensors. . .) among which experimental features can be input, in the spirit of OOMAO. This paper provides an overview of the main capabilities of the code and can be used as a user manual for interested users.

Abstract
The segmented nature of the 10-m Keck telescopes combined with facility-class AO systems and science instruments, and a history of science-driven upgrades to these systems, offers a uniquely powerful pathfinder for future AO science facilities on the segmented ELTs. Keck’s 2035 Strategic Vision includes visible, high contrast and ground layer AO facilities all of which could support ELT AO pathfinding. Keck’s pathfinder strength is not just demonstrating new techniques or technologies but developing them into operational science capabilities. For example, since first Keck AO science in 1999, Keck has successfully implemented three generations of sodium-wavelength lasers and is currently implementing its third generation of real-time controller (this time GPU-based). Current pathfinder-related developments include laser tomography, near-infrared low order wavefront sensing and PSF-reconstruction for high Strehl ratio and high sky coverage on the Keck I AO system. Current AO-based primary mirror phasing techniques under development include the use of Zernike, pyramid and phase diversity techniques. High-contrast AO developments include near-infrared pyramid wavefront sensing, on-sky phase diversity, speckle nulling and predictive wavefront control. Another pathfinder development is the NASA Goddard-led ORCAS satellite to provide a bright artificial point source for AO-correction. A fast, visible science camera has been implemented in support of ORCAS, demonstrating 15 mas FWHM, and, in a further move toward the visible, ALPAO is developing a 2.5 mm spacing, 60x60 actuator deformable mirror for Keck. In addition, three new AO science instruments are planned: Liger as a prototype of TMT’s IRIS, HISPEC which is the same as TMT’s MODHIS (based on KPIC’s science success), and SCALES.

Abstract
The Mid-infrared ELT Imager and Spectrograph (METIS) is is one of three first-generation science instruments for the Extremely Large Telescope (ELT) and has recently completed its final design phase. Its Single Conjugate Adaptive Optics (SCAO) system will provide the performance of an extreme adaptive optics system which enables high contrast imaging observations in the thermal/mid-infrared wavelength domain (3 µm – 13.3 µm). The Real-Time Computer (RTC) is the central component of the SCAO real-time control system. It executes the time critical wavefront control loop as well as associated control tasks by processing the data from the pyramid wavefront sensor and controlling the set of ELT actuators dedicated to adaptive optics. A total of up to 4,866 commands to be computed at a loop rate of up to 1 kHz imposes a number of demanding constraints in terms of memory throughput and computing power on the Hard Real-Time Core (HRTC), which employs GPU acceleration for the bulk of computations. Several auxiliary functions need to be in place to establish and maintain the quality of the wavefront correction. Among them are the control of the pupil position, the compensation of misregistration and of non-common path aberration, and the adaptation of the temporal control parameters. The main wavefront control loop has been prototyped to verify timing requirements. A median RTC computation time of 382 µs was achieved for a 300k samples (5 minutes) run. The results are presented in this paper together with the foreseen RTC hardware and the software deployment within the SCAO Control System.

Abstract
METIS, the Mid-infrared ELT Imager and Spectrograph is among the first-generation instruments for ESO’s 39m Extremely Large Telescope (ELT). It will provide diffraction-limited spectroscopy and imaging, including coronagraphic capabilities, in the thermal/mid-infrared wavelength domain (3 µm – 13.3 µm). Its Single Conjugate Adaptive Optics (SCAO) system will be used for all observing modes, with High Contrast Imaging imposing the most demanding requirements on its performance. The final design review of METIS took place in the fall of 2022; the development of the instrument, including its SCAO system, has since entered the Manufacturing, Assembly, Integration and Testing (MAIT) phase. Numerous challenging aspects of an ELT AO system are addressed in the mature designs for the SCAO control system and the SCAO hardware module: the complex interaction with the telescope entities that participate in the AO control, wavefront reconstruction with a fragmented and moving pupil, secondary control tasks to deal with differential image motion, non-common path aberrations and mis-registration. A K-band pyramid wavefront sensor and a GPU-based RTC, tailored to needs of METIS at the ELT, are core components. The implementation of the METIS SCAO system includes thorough testing at several levels before the installation at the telescope. These tests require elaborate setups to mimic the conditions at the telescope. This paper provides an overview of the design of METIS SCAO as it will be implemented, the main results of the extensive analyses performed to support the final design, and the next steps on the path towards commissioning.