1994 00 00 Q.J.R. Astronomical Society- V 35, Machine Intelligence, the Cost of Interstellar Travel and Fermi's Paradox - Louis K Scheffer

Magazine Overview

This document is an article from the "Q. J. R. astr. Soc." (Quarterly Journal of the Royal Astronomical Society), Volume 35, Issue No. 2, published in 1994. The article, titled "Machine Intelligence, the Cost of Interstellar Travel and Fermi's Paradox" by Louis K. Scheffer, explores the implications of machine intelligence for interstellar travel and its potential to resolve Fermi's Paradox.

Summary of the Article

The article posits that if machine intelligence is possible and based on computer architectures similar to today's, then interstellar travel can be achieved through data exchange rather than physical movement. This method is presented as many orders of magnitude cheaper than physical travel, providing a strong incentive for emerging societies to join existing galactic civilizations instead of physically colonizing the galaxy. Consequently, the author suggests that there might be at most one advanced civilization per galaxy, potentially with unified goals, which would remove the strongest underpinning of Fermi's paradox.

The paper includes a detailed analysis of the cost of interstellar communication via radio, considering both energy and fixed asset (antenna) costs. It concludes that deliberate communication is relatively cheap, while eavesdropping is likely futile.

I INTRODUCTION

The introduction reiterates Fermi's paradox: if Earth is typical, billions of civilizations should have evolved, and at least one should have colonized the galaxy. The author notes that previous studies on interstellar travel, such as those by von Hoerner, Sagan, and Crawford, have assumed physical travel via spacecraft, which is difficult and expensive due to high energy costs, even for slower-than-light methods.

Alternative to Physical Travel: Data Exchange

The article introduces the concept of 'teleportation' as a science fiction possibility, where a person's data is transmitted and reconstructed at a destination. While a theoretical objection related to quantum state collapse was addressed by Bennett et al. (1993), the practical problem of transmitting enormous amounts of information remains. A more feasible approach, drawing from current computer technology, is to stop a computer program, send its state as a stream of bits to a remote location, and resume its execution there. If conscious beings can be represented as computer programs, they could travel by shipping data, which is potentially much cheaper than physical travel.

The author estimates that a human being requires approximately 3x10¹⁴ bits to represent. Accelerating a 100 kg mass to 70% of the speed of light requires 10¹⁹ joules, whereas transmitting 3x10¹⁴ bits requires only 1.2x10⁻⁸ joules, representing a reduction of 8x10²⁶ in energy cost. For a 300 light-year trip, shipping data would cost approximately $4500 in energy, with total trip costs (including facilities) around $50,000. This is contrasted with physical interstellar travel, where even the lowest energy propulsion methods require energy costs of around $4x10¹³ for acceleration alone.

2 CAN A PROGRAM BE DEVELOPED WITH HUMAN CAPABILITIES?

This section addresses the question of whether a computer program can exhibit human-like intelligence and consciousness. While some philosophers like Searle and Penrose argue against it, most biologists and computer scientists believe it is possible in principle. The argument is made that if the state of every sub-atomic particle could be measured and its evolution simulated, human behavior could theoretically be predicted. Although quantum-level simulation is impractical due to the vast number of particles (10²³ atoms in the human body), a more practical approach focuses on simulating the nervous system, its neurones, and synapses. The article cites biological research suggesting that the nervous system controls behavior and that simulating neurones and synapses could replicate human attributes like intelligence and consciousness.

Arguments against this, such as Searle's 'Chinese Room' argument or Penrose's reliance on unknown quantum effects, are considered strained. The synapse emulation approach, while requiring significant memory and processing power, is conceptually simple. The article assumes that shorter conscious programs might also be possible, and even the 'brute force' simulation method would result in significantly cheaper interstellar travel than physical transfer.

3 WILL TOMORROW'S COMPUTERS ALLOW A STOP/RESTART CAPABILITY?

Current computers already possess the ability to save and restore program states, essential for multitasking and handling interrupts. This capability is expected to be maintained in future computers. The next step is to extend this to sending the program state to another computer and restarting it there. A fundamental theorem of computer science states that any digital computer can imitate any other digital computer, even across different instruction sets, through emulation programs. This process can be extended to emulate machines from other civilizations, requiring the transmission of the abstract machine's description and the program code.

3.1 Evolution of travel by information transfer

This concept is viewed as an extension of historical trends where information exchange replaces physical object exchange, such as from oral tradition to writing to digital files. Sending a person's representation rather than their physical body is a further step. It is also linked to virtual reality, where users can direct robots remotely, experiencing the environment as if they were physically present. With future technology, robots and computers could be located at various destinations, eliminating transit time limitations.

4 HOW MANY BITS ARE REQUIRED?

Estimating the number of bits needed to represent a human being is crucial for calculating the cost of data exchange travel. Two methods are proposed: simulating the physical brain and an abstract estimate starting from genetic information and accumulated experience.

Based on biological estimates, the human brain has about 10¹¹ neurones and 10¹⁴ synapses. Simulating these, with a margin of error, might require 10¹² neurones and 10¹⁵ synapses. Modeling a neurone and its synapses requires a certain number of bits to represent their state and connections. Considering various models, it is estimated that each synapse might require roughly 100 bits, leading to a total of approximately 10¹⁷ bits to describe the state of the brain.

An alternative, less constructive argument relies on Kolmogorov complexity, suggesting that the complexity of an object is the length of the shortest program that can recreate it. For deterministic systems, the initial state plus inputs can be smaller than the total state. This principle can be applied to analogue systems by sending an initial state and subsequent inputs and corrections.

For human beings, the number of bits is estimated based on genetic information (4x10⁹ bits) and accumulated experience, primarily visual. Assuming HDTV data rates, an old person's experience might be around 3x10¹⁶ bits, or more conservatively, 3x10¹⁵ bits. Other estimates for blind individuals (dominated by hearing) suggest around 6x10¹⁴ bits. The lower bound for proven memory feats is around 5x10⁷ bits, suggesting that representing a human mental state requires at least 5x10⁷ bits, and likely around 10¹¹ bits, with a more comprehensive estimate of 10¹⁵ bits.

5 THE COST OF SENDING INFORMATION ACROSS INTERSTELLAR DISTANCES

Studies by Oliver, Wolfe, and Drake et al. suggest that microwave transmission is the cheapest way to send information across interstellar distances in terms of energy. The article derives formulas for the energy cost per bit (E) and antenna cost (Cantenna) per trip, considering factors like temperature, distance, antenna radii, wavelength, and bandwidth. The optimal radius for minimizing cost is calculated, and the minimum cost is found to be sub-linear with respect to most parameters, except for the number of bits.

Reasonable values for parameters are discussed, including the cosmic background noise limit of 3 degrees K for temperature. Photon quantization noise imposes a minimum wavelength of about 5 mm (60 GHz). Antenna tolerance requirements are discussed, with examples like the Effelsberg radio telescope. The cost is further reduced by using the widest possible bandwidth.

Recurring Themes and Editorial Stance

The recurring themes in this article are the feasibility and cost-effectiveness of interstellar travel via data exchange, the potential for machine intelligence to enable such travel, and how this approach might resolve Fermi's Paradox. The editorial stance appears to be one of scientific inquiry, exploring theoretical possibilities and performing quantitative analyses based on current and projected technological capabilities. The article is rigorous in its approach, citing numerous scientific studies and providing detailed calculations to support its conclusions.

This issue of the Quarterly Journal of the Royal Astronomical Society, Volume 35, Number 2, published in 1994, features a significant article titled 'MACHINE INTELLIGENCE' by L.K. Scheffer. The content focuses on the feasibility and economics of interstellar communication, particularly in the context of the Fermi paradox and the potential for advanced civilizations.

Machine Intelligence and Interstellar Communication Costs

The article begins by discussing the engineering aspects of communication technology, noting that current technology yields a Ke (engineering margin) of about 10, with potential for improvement in efficiencies related to aperture efficiency, energy conversion, noise temperature, and coding overhead. A substantial portion of the analysis is dedicated to estimating the cost of large antennas required for interstellar communication. The cost of antennas is highly variable due to differing specifications and evolving technology. Historical data from NASA's Deep Space Network (DSN) suggests antenna costs scale with diameter raised to approximately the 2.78 power. Recasting historical costs into 1993 dollars, the article estimates the cost of a 1000-m antenna suitable for interstellar communication. For instance, the largest DSN antennas cost $50 imes 10^6$ for a 64-m diameter, which, using DSN scaling, would predict $100 imes 10^6$ for a 1000-m antenna. More recent estimates for a 100-m antenna at Greenbank were $55 imes 10^6$, implying a cost of $33 imes 10^6$ for a 1000-m antenna. However, factors like gravity and wind significantly impact ground-based antenna costs, which are less of a concern for space-based antennas. The Arecibo observatory's 305-m reflector, supported by the ground, was upgraded for a cost of $25 imes 10^6$ (in 1974 dollars), which, using DSN scaling, would suggest $700 imes 10^6$ for a 1000-m antenna.

Alternative approaches include antenna arrays, large space-based antennas, or antennas on the Moon. Studies suggest that for kilometer-scale apertures, space and ground-based antennas are of similar cost. An array with a 7 km² collecting area was estimated to cost roughly $30 imes 10^6$ (in 1975 dollars). Using DSN scaling, this implies a cost of $1.5 imes 10^9$ for a 1000-m antenna, though the scaling for arrays may differ.

A particularly interesting alternative is using the Sun's gravitational focusing for point-to-point communication. By positioning an antenna at a distance of at least 550 AU from the Sun, along the line passing through the Sun and the target star, the effective aperture can be greatly increased. This trades transportation costs for antenna costs, with calculations suggesting a factor of roughly $10^6$ gain in collecting area. Thus, a 1-m antenna at the gravitational focus could be equivalent to a conventional 1000-m antenna, with antenna costs being negligible but positioning costs potentially reaching $10^9 - 10^{10}$.

Cost Cases and Interstellar Travel

The article presents a Table I outlining low, medium, and high cost cases for interstellar communication. For a typical case, assuming a 300 light-year trip, a $10^9$ cost for a 1000-m antenna, $3 imes 10^{14}$ bits to be sent, and $0.08$ kWh energy cost, the total cost per trip is estimated at $52,500. This is considered orders of magnitude cheaper than physical interstellar travel.

Pessimistic assumptions, such as a 10,000 light-year trip, a $10^{10}$ antenna cost, $10^{17}$ bits, and high energy costs ($1 per kWh), result in a trip cost of $7 imes 10^8$. This is still significantly cheaper than the estimated cost of accelerating a 100 kg person to 70% of the speed of light, which is $1.2 imes 10^{12}$.

Optimistic assumptions, including 10 light-year trips, $10^{12}$ bits, a $10^9$ antenna cost, a 1000-year antenna life, and $0.01$ per kWh energy cost, yield a cost per trip of less than $2. When gravitational lensing is included, the cost per trip drops to $6 imes 10^{-5}$. The energy advantage of information transfer over physical travel is estimated to be on the order of $10^{17}$.

Accuracy of Cost Models and Wavelength Trade-offs

The analysis of cost models highlights that interstellar communication is relatively cheap. Errors in the cost estimates are unlikely to invalidate the main argument, as long as the estimates are not too low. Potential sources of error include the cube law scaling of antenna cost with size, the lack of dependence on wavelength, omission of labor costs, and the assumption of continuous link usage. However, even with variations in scaling exponents (1.5 to 3.5), costs remain within a factor of 2. The multiplicative constant A, representing the cost of antennas of this size, is less certain but has a limited impact on total cost due to its 4/7 power relationship.

Considering wavelength and cost trade-offs might lead to different optimum wavelengths, but this generally strengthens the argument for interstellar communication. The capacity of $10^{10}$ bits/sec is considered feasible. Labor costs are a small correction for short lifetimes but could become significant for long-lived antennas requiring maintenance.

Application to the Fermi Paradox

The Fermi paradox questions why, if intelligent life is common, we have not been colonized. The article suggests that travel by information transfer, being orders of magnitude cheaper and faster than physical travel, could offer an alternative explanation. If civilizations prioritize information transfer, they might spread across the galaxy without physical colonization. This could lead to a single, large galactic civilization. If this civilization has decided against colonizing Earth, the 'large number' argument of the Fermi paradox loses its force.

Furthermore, the capabilities enabled by information transfer technology could radically alter society, including personal immortality, duplication of persons, and enhanced intelligence. The merging of machine-based beings could accelerate evolution. These factors suggest that alien civilizations might have vastly different social structures and motivations than humans.

Conclusions and Observable Consequences

The two main conclusions are that information transfer is likely possible and significantly cheaper than physical travel. If intelligent life exists in the Galaxy, it probably forms one large civilization. Even if the Galaxy is saturated with radio traffic, cost minimization by civilizations would lead to highly directed signals, making them extremely difficult to detect from Earth without any conspiracy. The optimal frequency range for communication is likely 50-60 GHz, which is difficult to explore from the ground.

Evidence of a galactic civilization might include subtle signs like space-based structures of kilometer size or small ships at 'interesting' locations. Machine civilizations in Oort clouds using gravitational focusing for radio communication would be undetectable from our solar system.

Appendix A: Gravitational Focusing

Appendix A details the physics of gravitational focusing. Light rays passing near a massive object like the Sun are deflected by gravity. For a perfectly symmetrical Sun, grazing rays would focus at about 550 AU. A 1-m antenna at this focus could receive radiation from a ring around the Sun, potentially increasing collecting area by a factor of $2 imes 10^6$. The appendix also discusses the effect of the Sun's non-symmetry (oblateness) on focusing, concluding that gains of at least $3 imes 10^4$ and possibly up to $6 imes 10^6$ are achievable.

Recurring Themes and Editorial Stance

The recurring themes in this issue are the economic and technological feasibility of interstellar communication, the implications for the Fermi paradox, and the potential of advanced technologies like gravitational lensing. The editorial stance appears to be that interstellar communication is a plausible and potentially cost-effective endeavor, offering a compelling alternative to physical space travel and providing a framework for understanding the apparent absence of extraterrestrial civilizations. The journal, published by the Royal Astronomical Society, maintains a rigorous scientific approach, grounding its discussions in physics and engineering principles while exploring speculative but scientifically informed possibilities.