World models have garnered substantial interest in the AI community. These are internal representations that simulate aspects of the external world, track entities and states, capture causal relationships, and enable prediction of consequences. This contrasts with representations based solely on statistical correlations. A key motivation behind this research direction is that humans possess such mental world models, and finding evidence of similar representations in AI models might indicate that these models "understand" the world in a human-like way. In this paper, we use case studies from the philosophy of science literature to critically examine whether the world model framework adequately characterizes human-level understanding. We focus on specific philosophical analyses where the distinction between world model capabilities and human understanding is most pronounced. While these represent particular views of understanding rather than universal definitions, they help us explore the limits of world models.

Atomic data determined by analysis of observed atomic spectra are essential for plasma diagnostics. For each low-ionisation open d- and f-subshell atomic species, around $10^3$ fine structure level energies can be determined through years of analysis of $10^4$ observable spectral lines. We propose the automation of this task by casting the analysis procedure as a Markov decision process and solving it by graph reinforcement learning using reward functions learned on historical human decisions. In our evaluations on existing spectral line lists and theoretical calculations for Co II and Nd II-III, hundreds of level energies were computed within hours, agreeing with published values in 95% of cases for Co II and 54-87% for Nd II-III. As the current efficiency in atomic fine structure determination struggles to meet growing atomic data demands from astronomy and fusion science, our new artificial intelligence approach sets the stage for closing this gap.

We present a deep-learning based approach for measuring small planetary radial velocities in the presence of stellar variability. We use neural networks to reduce stellar RV jitter in three years of HARPS-N sun-as-a-star spectra. We develop and compare dimensionality-reduction and data splitting methods, as well as various neural network architectures including single line CNNs, an ensemble of single line CNNs, and a multi-line CNN. We inject planet-like RVs into the spectra and use the network to recover them. We find that the multi-line CNN is able to recover planets with 0.2 m/s semi-amplitude, 50 day period, with 8.8% error in the amplitude and 0.7% in the period. This approach shows promise for mitigating stellar RV variability and enabling the detection of small planetary RVs with unprecedented precision.

The accumulation of nonlinear effects in an optical fibre is often seen as a source of significant impairment for the propagating light signals, but the same effects, when properly managed, can provide a remarkable tool to tailor the temporal and spectral content of the signals. Indeed, depending on the regime of dispersion of the fibre and the frequency chirp, an initial pulse can be significantly expanded or compressed in the time or frequency domain, or it can be reshaped into advanced temporal waveforms such as parabolic, rectangular and triangular shapes [1]. Yet, due to the typically wide range of degrees of freedom involved, predicting the behaviour of nonlinear pulse shaping by numerical integration of the nonlinear Schrödinger equation (NLSE) or its extensions may be computationally demanding, especially when dealing with inverse-mapping problems. Recently, we have successfully introduced the use of the machine-learning (ML) method of artificial neural networks (NNs) as an efficient tool for complementing or substituting the NLSE in the modelling of nonlinear pulse shaping [2-5] or for predicting the generation of optical supercontinua [6, 7]. Fibre nonlinearity does not only affect the propagation of ultrashort pulses.

Deep learning techniques for gravitational-wave parameter estimation have emerged as a fast alternative to standard samplers $\unicode{x2013}$ producing results of comparable accuracy. These approaches (e.g., DINGO) enable amortized inference by training a normalizing flow to represent the Bayesian posterior conditional on observed data. By conditioning also on the noise power spectral density (PSD) they can even account for changing detector characteristics. However, training such networks requires knowing in advance the distribution of PSDs expected to be observed, and therefore can only take place once all data to be analyzed have been gathered. Here, we develop a probabilistic model to forecast future PSDs, greatly increasing the temporal scope of DINGO networks. Using PSDs from the second LIGO-Virgo observing run (O2) $\unicode{x2013}$ plus just a single PSD from the beginning of the third (O3) $\unicode{x2013}$ we show that we can train a DINGO network to perform accurate inference throughout O3 (on 37 real events). We therefore expect this approach to be a key component to enable the use of deep learning techniques for low-latency analyses of gravitational waves.

Recent methods in the machine learning literature have proposed a Gaussian noise-based exogenous signal to learn the parameters of a dynamic system. In this paper, we propose the use of a spectral lines-based deterministic exogenous signal to solve the same problem. Our theoretical analysis consists of a new toolkit which employs the theory of spectral lines, retains the stochastic setting, and leads to non-asymptotic bounds on the parameter estimation error. The results are shown to lead to a tunable parameter identification error. In particular, it is shown that the identification error can be minimized through an an optimal choice of the spectrum of the exogenous signal.

We propose a learning-based approach for estimating the spectrum of a multisinusoidal signal from a finite number of samples. A neural-network is trained to approximate the spectra of such signals on simulated data. The proposed methodology is very flexible: adapting to different signal and noise models only requires modifying the training data accordingly. Numerical experiments show that the approach performs competitively with classical methods designed for additive Gaussian noise at a range of noise levels, and is also effective in the presence of impulsive noise.

Doppler spectroscopy (also known as the radial-velocity method, or colloquially, the wobble method) is an indirect method for finding extrasolar planets and brown dwarfs from radial-velocity measurements via observation of Doppler shifts in the spectrum of the planet's parent star. Otto Struve proposed in 1952 the use of powerful spectrographs to detect distant planets. He described how a very large planet, as large as Jupiter, for example, would cause its parent star to wobble slightly as the two objects orbit around their center of mass.[2] He predicted that the small Doppler shifts to the light emitted by the star, caused by its continuously varying radial velocity, would be detectable by the most sensitive spectrographs as tiny red shifts and blue shifts in the star's emission. However, the technology of the time produced radial-velocity measurements with errors of 1,000 m/s or more, making them useless for the detection of orbiting planets.[3]

Space is a cold and barren place. Nothing can exist there, nothing!" Ludwig Von Drake, an obscure uncle of Donald Duck and a professor of astronomy, is sitting on a high stool in his observatory. When he sees that he is being filmed, he falls off and lands on the floor with a loud thump. "Now I can see stars I've never seen before!" he groans. He walks over to a table with a large pile of books on it. The thickest of them all is a guide to space travel that he wrote himself. In a 45 -minute- long monologue, he tells us in a thick German accent how mankind discovered the planets in our solar system and has fantasized about everything that might be crawling around on them. Every now and then, he picks up a book from the large pile and reads from it, and then throws it nonchalantly into a corner of the room. He tells us about Copernicus and Galileo, and about Kepler's dreams about Martians, Fontenelle's speculations about life on other planets, and even John Herschel's Great Moon Hoax. Science fiction comes to life in the colorful cartoon: Hairy space beings and flying saucers shoot across the screen. At the end, the professor has the last word. He finds all these fantasies poppycock; nothing can live in that empty, barren space! But, as he is speaking, Von Drake is kidnapped by a black Martian robot from one of his stories. The cartoon, Inside Outer Space, is part of Walt Disney's Wonderful World of Color, a television series from the 1960s. The absent minded duck professor hosts a number of episodes, each with their own topic: the history of flight, the color spectrum, space--all exciting stuff for American kids in the Space Age. Lou Allamandola spent his teenage years in the science- crazy 1960s. He grew up in a Catholic family in the state of New Jersey. His grandparents were immigrants from Italy, and he didn't learn to speak English until he went to school. He still clearly remembers the Disney cartoons with Ludwig Von Drake, which were broadcast on Saturday evenings. "Von Drake called the interstellar medium--the empty space between the stars and the planets--a barren place where nothing could exist," he tells me. "That was all we knew in the '60s.