arXiv subjects
Publications and source records attributed to Max Tegmark.
Success in the quest for artificial intelligence has the potential to bring unprecedented benefits to humanity, and it is therefore worthwhile to investigate how to maximize these benefits while avoiding potential pitfalls. This article gives numerous examples (which should by no means be construed as an exhaustive list) of such worthwhile research aimed at ensuring that AI remains robust and beneficial.
I examine the hypothesis that consciousness can be understood as a state of matter, "perceptronium", with distinctive information processing abilities. I explore five basic principles that may distinguish conscious matter from other physical systems such as solids, liquids and gases: the information, integration, independence, dynamics and utility principles. This approach generalizes Giulio Tononi's integrated information framework for neural-network-based consciousness to arbitrary quantum systems, and provides interesting links to error-correcting codes and condensed matter criticality, as well as an interesting connections between the emergence of consciousness and the emergence of time. (For more technical details, see arXiv:1401.1219).
The separation of the faint cosmological background signal from bright astrophysical foregrounds remains one of the most daunting challenges of mapping the high-redshift intergalactic medium with the redshifted 21 cm line of neutral hydrogen. Advances in mapping and modeling of diffuse and point source foregrounds have improved subtraction accuracy, but no subtraction scheme is perfect. Precisely quantifying the errors and error correlations due to missubtracted foregrounds allows for both the rigorous analysis of the 21 cm power spectrum and for the maximal isolation of the "EoR window" from foreground contamination. We present a method to infer the covariance of foreground residuals from the data itself in contrast to previous attempts at a priori modeling. We demonstrate our method by setting limits on the power spectrum using a 3 h integration from the 128-tile Murchison Widefield Array. Observing between 167 and 198 MHz, we find at 95% confidence a best limit of Delta^2(k) < 3.7 x 10^4 mK^2 at comoving scale k = 0.18 hMpc^-1 and at z = 6.8, consistent with existing limits.
I discuss the impact of computer progress on nuclear war policy, both by enabling more accurate nuclear winter simulations and by affecting the probability of war starting accidentally. I argue that from a cosmic perspective, humanity's track record of risk mitigation is inexcusably pathetic, jeopardizing the potential for life to flourish for billions of years.
We examine the hypothesis that consciousness can be understood as a state of matter, "perceptronium", with distinctive information processing abilities. We explore five basic principles that may distinguish conscious matter from other physical systems such as solids, liquids and gases: the information, integration, independence, dynamics and utility principles. If such principles can identify conscious entities, then they can help solve the quantum factorization problem: why do conscious observers like us perceive the particular Hilbert space factorization corresponding to classical space (rather than Fourier space, say), and more generally, why do we perceive the world around us as a dynamic hierarchy of objects that are strongly integrated and relatively independent? Tensor factorization of matrices is found to play a central role, and our technical results include a theorem about Hamiltonian separability (defined using Hilbert-Schmidt superoperators) being maximized in the energy eigenbasis. Our approach generalizes Giulio Tononi's integrated information framework for neural-network-based consciousness to arbitrary quantum systems, and we find interesting links to error-correcting codes, condensed matter criticality, and the Quantum Darwinism program, as well as an interesting connection between the emergence of consciousness and the emergence of time.
In order to study the "Cosmic Dawn" and the Epoch of Reionization with 21 cm tomography, we need to statistically separate the cosmological signal from foregrounds known to be orders of magnitude brighter. Over the last few years, we have learned much about the role our telescopes play in creating a putatively foreground-free region called the "EoR window." In this work, we examine how an interferometer's effects can be taken into account in a way that allows for the rigorous estimation of 21 cm power spectra from interferometric maps while mitigating foreground contamination and thus increasing sensitivity. This requires a precise understanding of the statistical relationship between the maps we make and the underlying true sky. While some of these calculations would be computationally infeasible if performed exactly, we explore several well-controlled approximations that make mapmaking and the calculation of map statistics much faster, especially for compact and highly redundant interferometers designed specifically for 21 cm cosmology. We demonstrate the utility of these methods and the parametrized trade-offs between accuracy and speed using one such telescope, the upcoming Hydrogen Epoch of Reionization Array, as a case study.
A number of experiments are currently working towards a measurement of the 21 cm signal from the Epoch of Reionization. Whether or not these experiments deliver a detection of cosmological emission, their limited sensitivity will prevent them from providing detailed information about the astrophysics of reionization. In this work, we consider what types of measurements will be enabled by a next-generation of larger 21 cm EoR telescopes. To calculate the type of constraints that will be possible with such arrays, we use simple models for the instrument, foreground emission, and the reionization history. We focus primarily on an instrument modeled after the $\sim 0.1~\rm{km}^2$ collecting area Hydrogen Epoch of Reionization Array (HERA) concept design, and parameterize the uncertainties with regard to foreground emission by considering different limits to the recently described "wedge" footprint in k-space. Uncertainties in the reionization history are accounted for using a series of simulations which vary the ionizing efficiency and minimum virial temperature of the galaxies responsible for reionization, as well as the mean free path of ionizing photons through the IGM. Given various combinations of models, we consider the significance of the possible power spectrum detections, the ability to trace the power spectrum evolution versus redshift, the detectability of salient power spectrum features, and the achievable level of quantitative constraints on astrophysical parameters. Ultimately, we find that $0.1~\rm{km}^2$ of collecting area is enough to ensure a very high significance ($\gtrsim30σ$) detection of the reionization power spectrum in even the most pessimistic scenarios. This sensitivity should allow for meaningful constraints on the reionization history and astrophysical parameters, especially if foreground subtraction techniques can be improved and successfully implemented.
Relentless progress in artificial intelligence (AI) is increasingly raising concerns that machines will replace humans on the job market, and perhaps altogether. Eliezer Yudkowski and others have explored the possibility that a promising future for humankind could be guaranteed by a superintelligent "Friendly AI", designed to safeguard humanity and its values. I argue that, from a physics perspective where everything is simply an arrangement of elementary particles, this might be even harder than it appears. Indeed, it may require thinking rigorously about the meaning of life: What is "meaning" in a particle arrangement? What is "life"? What is the ultimate ethical imperative, i.e., how should we strive to rearrange the particles of our Universe and shape its future? If we fail to answer the last question rigorously, this future is unlikely to contain humans.
We prove a generalization of the classic Groenewold-Lindblad entropy inequality, combining decoherence and the quantum Bayes theorem into a simple unified picture where decoherence increases entropy while observation decreases it. This provides a rigorous quantum-mechanical version of the second law of thermodynamics, governing how the entropy of a system (the entropy of its density matrix, partial-traced over the environment and conditioned on what is known) evolves under general decoherence and observation. The powerful tool of spectral majorization enables both simple alternative proofs of the classic Lindblad and Holevo inequalities without using strong subadditivity, and also novel inequalities for decoherence and observation that hold not only for von Neumann entropy, but also for arbitrary concave entropies.
We report on the MIT Epoch of Reionization (MITEoR) experiment, a pathfinder low-frequency radio interferometer whose goal is to test technologies that improve the calibration precision and reduce the cost of the high-sensitivity 3D mapping required for 21 cm cosmology. MITEoR accomplishes this by using massive baseline redundancy, which enables both automated precision calibration and correlator cost reduction. We demonstrate and quantify the power and robustness of redundancy for scalability and precision. We find that the calibration parameters precisely describe the effect of the instrument upon our measurements, allowing us to form a model that is consistent with $χ^2$ per degree of freedom < 1.2 for as much as 80% of the observations. We use these results to develop an optimal estimator of calibration parameters using Wiener filtering, and explore the question of how often and how finely in frequency visibilities must be reliably measured to solve for calibration coefficients. The success of MITEoR with its 64 dual-polarization elements bodes well for the more ambitious Hydrogen Epoch of Reionization Array (HERA) project and other next-generation instruments, which would incorporate many identical or similar technologies.
We present techniques for bridging the gap between idealized inverse covariance weighted quadratic estimation of 21 cm power spectra and the real-world challenges presented universally by interferometric observation. By carefully evaluating various estimators and adapting our techniques for large but incomplete data sets, we develop a robust power spectrum estimation framework that preserves the so-called "EoR window" and keeps track of estimator errors and covariances. We apply our method to observations from the 32-tile prototype of the Murchinson Widefield Array to demonstrate the importance of a judicious analysis technique. Lastly, we apply our method to investigate the dependence of the clean EoR window on frequency--especially the frequency dependence of the so-called "wedge" feature--and establish upper limits on the power spectrum from z = 6.2 to z = 11.7. Our lowest limit is Delta(k) < 0.3 Kelvin at 95% confidence at a comoving scale k = 0.046 Mpc^-1 and z = 9.5.
Mapping our universe in 3D by imaging the redshifted 21 cm line from neutral hydrogen has the potential to overtake the cosmic microwave background as our most powerful cosmological probe, because it can map a much larger volume of our Universe, shedding new light on the epoch of reionization, inflation, dark matter, dark energy, and neutrino masses. We report on MITEoR, a pathfinder low-frequency radio interferometer whose goal is to test technologies that greatly reduce the cost of such 3D mapping for a given sensitivity. MITEoR accomplishes this by using massive baseline redundancy both to enable automated precision calibration and to cut the correlator cost scaling from N^2 to NlogN, where N is the number of antennas. The success of MITEoR with its 64 dual-polarization elements bodes well for the more ambitious HERA project, which would incorporate many identical or similar technologies using an order of magnitude more antennas, each with dramatically larger collecting area.
[Abridged] The spatially averaged global spectrum of the redshifted 21cm line has generated much experimental interest, for it is potentially a direct probe of the Epoch of Reionization and the Dark Ages. Since the cosmological signal here has a purely spectral signature, most proposed experiments have little angular sensitivity. This is worrisome because with only spectra, the global 21cm signal can be difficult to distinguish from foregrounds such as Galactic synchrotron radiation, as both are spectrally smooth and the latter is orders of magnitude brighter. We establish a mathematical framework for global signal data analysis in a way that removes foregrounds optimally, complementing spectra with angular information. We explore various experimental design trade-offs, and find that 1) with spectral-only methods, it is impossible to mitigate errors that arise from uncertainties in foreground modeling; 2) foreground contamination can be significantly reduced for experiments with fine angular resolution; 3) most of the statistical significance in a positive detection during the Dark Ages comes from a characteristic high-redshift trough in the 21cm brightness temperature; and 4) Measurement errors decrease more rapidly with integration time for instruments with fine angular resolution. We show that if observations and algorithms are optimized based on these findings, an instrument with a 5 degree beam can achieve highly significant detections (greater than 5-sigma) of even extended (high Delta-z) reionization scenarios after integrating for 500 hrs. This is in contrast to instruments without angular resolution, which cannot detect gradual reionization. Abrupt ionization histories can be detected at the level of 10-100's of sigma. The expected errors are also low during the Dark Ages, with a 25-sigma detection of the expected cosmological signal after only 100 hrs of integration.
We develop and demonstrate an acceleration of the Liu & Tegmark quadratic estimator formalism for inverse variance foreground subtraction and power spectrum estimation in 21 cm tomography from O(N^3) to O(N log N), where N is the number of voxels of data. This technique makes feasible the megavoxel scale analysis necessary for current and upcoming radio interferometers by making only moderately restrictive assumptions about foreground models and survey geometry. We exploit iterative and Monte Carlo techniques and the symmetries of the foreground covariance matrices to quickly estimate the 21 cm brightness temperature power spectrum, P(k_parallel, k_perpendicular), the Fisher information matrix, the error bars, the window functions, and the bias. We also extend the Liu & Tegmark foreground model to include bright point sources with known positions in a way that scales as O[(N log N)(N point sources)] < O(N^5/3). As a first application of our method, we forecast error bars and window functions for the upcoming 128-tile deployment of the Murchinson Widefield Array, showing that 1000 hours of observation should prove sufficiently sensitive to detect the power spectrum signal from the Epoch of Reionization.
We develop an approach for linking the power spectra, bispectrum, and trispectrum to the geometric and kinematical features of multifield inflationary Lagrangians. Our geometric approach can also be useful in determining when a complicated multifield model can be well approximated by a model with one, two, or a handful of fields. To arrive at these results, we focus on the mode interactions in the kinematical basis, starting with the case of no sourcing and showing that there is a series of mode conservation laws analogous to the conservation law for the adiabatic mode in single-field inflation. We then treat the special case of a quadratic potential with canonical kinetic terms, showing that it produces a series of mode sourcing relations identical in form to that for the adiabatic mode. We build on this result to show that the mode sourcing relations for general multifield inflation are extension of this special case but contain higher-order covariant derivatives of the potential and corrections from the field metric. In parallel, we show how these interactions depend on the geometry of the inflationary Lagrangian and on the kinematics of the associated field trajectory. Finally, we consider how the mode interactions and effective number of fields active during inflation are reflected in the spectra and introduce a multifield consistency relation, as well as a multifield observable that can potentially distinguish two-field scenarios from scenarios involving three or more effective fields.
We analyze cosmology assuming unitary quantum mechanics, using a tripartite partition into system, observer and environment degrees of freedom. This generalizes the second law of thermodynamics to "The system's entropy can't decrease unless it interacts with the observer, and it can't increase unless it interacts with the environment." The former follows from the quantum Bayes Theorem we derive. We show that because of the long-range entanglement created by cosmological inflation, the cosmic entropy decreases exponentially rather than linearly with the number of bits of information observed, so that a given observer can reduce entropy by much more than the amount of information her brain can store. Indeed, we argue that as long as inflation has occurred in a non-negligible fraction of the volume, almost all sentient observers will find themselves in a post-inflationary low-entropy Hubble volume, and we humans have no reason to be surprised that we do so as well, which solves the so-called inflationary entropy problem. An arguably worse problem for unitary cosmology involves gamma-ray-burst constraints on the "Big Snap", a fourth cosmic doomsday scenario alongside the "Big Crunch", "Big Chill" and "Big Rip", where an increasingly granular nature of expanding space modifies our life-supporting laws of physics. Our tripartite framework also clarifies when it is valid to make the popular quantum gravity approximation that the Einstein tensor equals the quantum expectation value of the stress-energy tensor, and how problems with recent attempts to explain dark energy as gravitational backreaction from super-horizon scale fluctuations can be understood as a failure of this approximation.
We study the quantum measurement problem in the context of an infinite, statistically uniform space, as could be generated by eternal inflation. It has recently been argued that when identical copies of a quantum measurement system exist, the standard projection operators and Born rule method for calculating probabilities must be supplemented by estimates of relative frequencies of observers. We argue that an infinite space actually renders the Born rule redundant, by physically realizing all outcomes of a quantum measurement in different regions, with relative frequencies given by the square of the wave function amplitudes. Our formal argument hinges on properties of what we term the quantum confusion operator, which projects onto the Hilbert subspace where the Born rule fails, and we comment on its relation to the oft-discussed quantum frequency operator. This analysis unifies the classical and quantum levels of parallel universes that have been discussed in the literature, and has implications for several issues in quantum measurement theory. It also shows how, even for a single measurement, probabilities may be interpreted as relative frequencies in unitary (Everettian) quantum mechanics. We also argue that after discarding a zero-norm part of the wavefunction, the remainder consists of a superposition of indistinguishable terms, so that arguably "collapse" of the wavefunction is irrelevant, and the "many worlds" of Everett's interpretation are unified into one. Finally, the analysis suggests a "cosmological interpretation" of quantum theory in which the wave function describes the actual spatial collection of identical quantum systems, and quantum uncertainty is attributable to the observer's inability to self-locate in this collection.
Before it becomes a sensitive probe of the Epoch of Reionization, the Dark Ages, and fundamental physics, 21 cm tomography must successfully contend with the issue of foreground contamination. Broadband foreground sources are expected to be roughly four orders of magnitude larger than any cosmological signals, so precise foreground models will be necessary. Such foreground models often contain a large number of parameters, reflecting the complicated physics that governs foreground sources. In this paper, we concentrate on spectral modeling (neglecting, for instance, bright point source removal from spatial maps) and show that 21 cm tomography experiments will likely not be able to measure these parameters without large degeneracies, simply because the foreground spectra are so featureless and generic. However, we show that this is also an advantage, because it means that the foregrounds can be characterized to a high degree of accuracy once a small number of parameters (likely three or four, depending on one's instrumental specifications) are measured. This provides a simple understanding for why 21 cm foreground subtraction schemes are able to remove most of the contaminants by suppressing just a small handful of simple spectral forms. In addition, this suggests that the foreground modeling process should be relatively simple and will likely not be an impediment to the foreground subtraction schemes that are necessary for a successful 21 cm tomography experiment.