The progress of technological civilizations should enable their development of both biotechnology and molecular nanotechnology. Biotechnology engineering (replacement parts, genome program manipulation, etc.) should produce societies of individuals that do not age. Such civilizations will emphasize self-evolution toward information processing limits and the minimization of their long term hazard function. Molecular nanotechnology enables the rapid construction of nested Dyson shell supercomputers ("Matrioshka Brains") that fully utilize the energy and matter resources of solar systems and operate at the limits of intelligence and longevity given known physical laws. Over long time scales, galactic hazards eliminate civilizations that fail to follow these paths.Acceleration of technology trends allows civilizations to evolve from primitive stages, lacking an understanding of most laws of physics, to full Matrioshka Brains levels in ~200 years. Communicating civilizations, as we typically envision them, will thus have a short longevity (the Drake Equation "L" parameter). The intellectual capacity and observational capabilities of these integrated superintelligences are the reason for the historic failure of SETI. Their physical characteristics determines why they are difficult to observe and explain the Fermi Paradox. Current astronomical observations which include the missing mass of the galaxies, gravitational microlensing reports and galactic infrared observations (e.g. NGC 5907), suggest that many galaxies, including the Milky Way, may be Kardashev Type III civilizations composed of Matrioshka Brains.
Introduction
Most previous discussions of extraterrestrial intelligence have assumed such intelligences bear some similarity to ourselves in our current or near future. These discussions have failed to account for developments that can reasonably be expected in the fields of molecular nanotechnology[1] and biomedical technology[2, 3] in the next century. Intelligent species that decode their genomes should develop the ability to modify their genetic program to extend their lifespan to hundreds of years. The development of full scale molecular nanotechnology enables the uploading of minds into computers[4, 5] with lifespans limited by their hazard function of their local environment in space.
When civilizations have evolved indefinite lifespans there is likely to occur a shift of focus from such activities as simple survival and procreation or the accumulation of wealth to the maximization of intelligence and internal communications bandwidth. These civilizations will also focus on the minimization of their overall hazard function to allow the greatest longevity. Civilizations that are successful in these efforts should become the dominant species in a galaxy.
The fundamental problem facing an advanced technological civilization (ATC) is how to utilize the available matter and energy resources to provide the maximum computational capacity. Current estimates suggest nanotechnology enables mass doubling times for molecular scale machinery of an hour [1]. This estimate may be conservative given that bacteria constructed of lower strength materials and generally lacking directed delivery of molecular building blocks can double their mass in 20 minutes. Mass doubling times of this order allow the construction of objects with the mass of the galaxy M31 in 9 days.
As Dyson pointed out [6], if one has the power of a star available, one can dismantle planets like Jupiter and use it for construction materials. The problem for an ATC then becomes how to obtain such power. The use of self-replicating von Neumann machines allows rapid disassembly of low gravity bodies (e.g. asteroids or small planets). The material from these bodies is hurled into space where it is recast as solar power harvesting devices with an areal mass of ~1 kg/m2. These collectors beam power back to the source where it is used to exponentially expand the disassembly process. These methods allow planets such as Mercury to be turned into collectors harvesting the entire solar output in approximately 2 weeks [7]. Further applications of this method would make all of the nonstellar material in solar systems such as ours available for construction within 1000 years. So in considering the capacities for long lived ATC we should assume they minimally have the power resources of a star (~1026W) and the matter resources of a solar system (~1026 kg) at their disposal. As such they constitute Kardashev Type II civilizations [8].
Matrioshka Brain Architectures
Sandberg has discussed tradeoffs for megascale computational structures. They may be small and hot (requiring extensive error correction) or large and cold (being penalized with greater communications delays). Without concrete physical designs and computational algorithms it is difficult to determine which of the architectures proposed by Sandberg could be optimal. One possible architecture for a Dyson Shell [6] is a collection of orbiting satellites consisting of a power collector, a nanocomputer, a heat radiator and optical communications arrays (CCDs for receivers and VCSELs for transmitters). Aggregates of ~1012 nanocomputers have a small volume (1 cm3) but have a high power consumption (~105W) [1]. The heat dissipation requirements of such computational nodes prevent them from being closely packed without diverting excessive amounts of the available power into the circulation of cooling fluid. The inter-satellite distance is determined by the area requirements for power absorption and heat radiation and may vary from 0.5 m for high temperature computers to 100's of km for low temperature computers. This is due to the scaling of radiation efficiency with T4 from Stephan Boltzman Law. A shell of nanocomputer satellites harvesting the Sun's output would have the processing capacity of ~1042 Instructions per Second (IPS) as compared with a generous estimate of ~1017 for a human.
It can be shown however that single layer Dyson shells fail to use all of the available construction material and inefficiently utilize the available energy. For a shell with an orbital radius similar to that of the planet Mercury, the computer mass would be ~1016 kg and the radiator and coolant mass ~1018 kg. This is much less than the material available in the solar system. Since the mass utilization is dominated by the heat dissipation requirements, the full utilization of the matter resources of a solar system requires an architecture of nested orbiting arrays of computational satellites. This architecture is called a Matrioshka Brain (M-Brain) after the nesting Russian dolls. The optimal use of materials is to construct computers of varying architectures that best utilize the available elements at various temperatures. These materials could include TiC, Al2O3, Diamond, SiC, GaN, GaAs, Si, Organic, high-temperature superconductors and low-temperature superconductors as the distance of the shell from the star increases and the operating temperature decreases. Each shell could be powered by the waste heat of the next inner level. The innermost layers would consist of complete sun enveloping shells while the outermost layers would be incomplete due to the lack of material for radiator construction.
The high signaling rate and parallel architecture of VCSEL and CCD arrays enables very large bandwidths between satellites. Conservative estimates are in the range of ~104 Terabytes per node. The inter-node photons constitute a virtual memory capacity around the star of ~1015 Terabytes. Outer layers of the shell would have larger memory capacities but would require greater power for internodal communications due to the larger distances between the nodes.
From an architectural standpoint there are significant tradeoffs in power and material allocations in computers, memory storage and communication. These are presumably optimized for the specific computational requirements of individual M-Brains. Devices based on nanotechnology are extremely radiation sensitive [1], so a M-Brain must either migrate to low radiation environments (e.g. galactic halos), or devote non-trivial fractions of its computational capacity to running self-diagnostics and utilizing valuable energy resources to dismantle and replace components damaged by radiation or sacrifice precious mass resources to radiation shielding.
The Dismal Returns from Interstellar Colonization
Computer scientists have developed methods for evaluating the computational throughput of different parallel computing architectures. The LogP [10] and LogGP [11] methods may be adapted to compare human brains and various M-Brain configurations. These models have a number of variables, the most important of which for this analysis is the Latency to communicate a message from its source to its destination. Latency can be approximated simply by the inter-processor distance. In the human brain the Latency times are usually milliseconds (where the nodes are individual neurons). Within single satellite nanocomputer nodes these times are much less, on the order of 0.02-600 ns. Between nodes of a shell times vary from 0.5 ns to 0.5 ms. These ranges span 7 orders of magnitude. If M-Brains were to attempt the construction of a multi-star intelligence, their Latency penalty would be from 104 to 107, due to communications times from light-days to light-years. In addition a significant power penalty must be paid for inter- vs. intra-M-Brain communications. So while intelligence increases significantly as a solar system is developed, colonizing adjacent stars provides greatly reduced benefits and decreases still further as the colonization radius expands. While it is difficult to say interstellar colonization never occurs, it is clear that the benefits are small compared with the costs of leaving behind access to a huge memory base and computational capacity.
Observational Evidence
In the areas of astronomy and astrophysics, there are a number of unsolved problems or unexplained observations. These include (1) The Missing Mass problem; (2) The Gravitational Microlensing observations; (3) Missing stars in galactic halos; (4) Low Surface Brightness and Dwarf Galaxies; (5) The Gamma-Ray Halo of the Milky Way and infrared "glows" of other galaxies; (6) An excess of far-infrared light detected by COBE; (7) The possible age discrepancy between the universe and Globular Clusters; (8) Low temperature objects in the IRAS survey; (9) The arrangement of observable galaxies as "walls"; (10) Variations in the Cosmic Microwave Background Radiation; (11) Variations in the brightness of Type Ia supernovae.
It is doubtful that all of these phenomena can be attributed to M-Brains, but plausible arguments can be made in each instance [see reference 12 for details] that the observations could correspond to the astroengineering activities by long-lived ATC at the limits of known physics. Astronomers seem reluctant to consider that their observations could be explained by ATC and yet never discuss their implicit assumption. Namely, that all ATC must never evolve to the limits of intelligence and longevity allowed by physics. Expressed another way they are assuming that the probability of universal fatality for ATC is 1! If that assumption is invalid then astronomical observations must be interpreted in light of the fact that such civilizations may evolve, live indefinitely and eventually come to dominate galactic populations.
Conclusion
The "Fermi Paradox" [13] is often cited as evidence that extraterrestrial civilizations do not exist. However, nanotechnology allows the creation of bacteria sized agents that can function as remote probes or sensors for ATC. The fact that we can now envision such agents means that earlier efforts [14, 15, 16] to explain the apparent lack of ATC should be taken much more seriously. We cannot conclude that ATC are not here but may only conclude that they either have no interest in us (so there are no agents) or are primarily interested in observing us (free from contact). The lack of interest may be explained by the large difference between the tightly coupled intelligence of a M-Brain (~1042 IPS) and the loosely coupled intelligence of human civilization (~1026 IPS). Observation without contact may be explained by the problem of convergent evolution of ATC that share information. M-Brains may have strong incentives to allow the untainted development of civilizations such as ours since independent evolutionary histories may be the best method for creating novel computing architectures. So we may not be in a "Zoo" as proposed by Ball [16], but in a "nursery" designed to promote galactic diversity.
The material capacity of an M-Brain allows them to devote a relatively small fraction of their matter (< 1%) to construct ~1011 reflecting telescopes in with the diameter of the moon. This provides the capacity to continually observe objects of interest in their own galaxy as well as nearby galaxies. Presumably when they notice a civilization of interest to communicate with they could initiate communications using highly directional optical or microwave carriers that would minimize the energy requirements. The M-Brain capabilities discussed here point out the shortcomings of arguments such as those by Papagiannis [18] that argue that ATC should be visible around stars. ATC should not emit much of visible radiation and those SETI searches based on such assumptions seem likely to fail. Approaches that look at infrared survey data [19] seem much more likely to yield observations of ATC.
The M-Brain architecture outlined in this paper is of interest with regard to the evolution of extraterrestrial civilizations and a probable evolutionary path for humanity. If our rapid development of science and engineering is typical, civilizations go from relatively unscientific stages to M-Brain levels within a few hundred years. For these reasons we should explore the ideas presented here in greater detail.
References