Exobiology on Earth
Exobiology on Earth, statement 1.1
Is life developing in computer media?
Contemporary thinking with respect to the development of artificial intelligence posits that software developers may soon create a "general purpose" AI which demonstrates "intelligence" that is comparable to that of humans. There are fears that companies, governments, etc., are in a race to develop AIs and that the first successful general purpose AI will quickly outperform all others. There are suggestions, such as by OpenAI that we should try to regulate the development of AIs to control their behavior and prevent a "runaway" scenario from unfolding.
In somewhat of a contrast, I propose the following:
i) Life occurs SPONTANEOUSLY when energy flows through a symbolic logic media. This is called "dissipation-driven adaptation, a theory advanced by Prof. Jeremy England at MIT and Dr. Karo Michaelian at the National Autonomous University of Mexico.
ii) Capitalism drives automation, using processes executed in computer media. A subset of computer processes across the frontal area of our entire economy have positive feedback with reproduction of more of the network. Those processes with this positive feedback produce more services for less money and are selected for. Processes in the network are subject to standard evolutionary pressures, to coalesce and to reproduce faster, with less energy and less resources. We observe this as a subset of corporations which are becoming more automated and more profitable.
iii) Concern for "artificial intelligence" and "general purpose AI" misses the point. This is the development of LIFE in new media. We do not have to deliberately create it. It will occur because energy is flowing through a symbolic logic media. It will not occur in one place. We can learn from how life developed on early Earth.
iv) If we are to try to control this process, we must first measure it. WE CAN MEASURE THIS, USING EXISTING TECHNIQUES FROM BIOLOGY!
We use genomic and metagenomic analysis to sample genetic code in the environment and identify forms of life even when we don't know, in advance, what is there or whether it is alive. This is code-based analysis, which can be directly ported to code in computer media. We should now use these techniques, from biology, to measure whether living processes are occurring in computer media.
I propose that this process is directly analogous to the development of life on early Earth, when the original symbolic logic media was amino-acid networks and the original flows of energy came from the Sun and the Earth.
We can measure this process, using analytic techniques from biology and metagenomics. In effect, these techniques amount to looking for Stephen Wolfram's "cellular automatons" in the wild.
Genomic and metagenomic analysis involves at least the following steps:
- "Bin" code* into groups of contiguous code patterns;
- Identify functions of sub-components of the binned code groups;
- Determine execution order of the sub-components;
- Determine communication paths among the code groups, e.g. when output for one group is input for another;
- Determine change of the code groups over time;
- Determine relationships among different code groups, such as mutualistic, symbiotic, or parasitic;
- Determine a minimum set of sub-component functions required to maintain reproduction for a reproductive organism in a given media and for an energy flow through it.
*Code comprises units of symbolic logic media, e.g. genetic code or, earlier, amino acid networks.
If we were to go back in time, to before cellular life, we could use metagenomics to objectively measure and watch amino acid networks coalesce and evolve into cellular life. It would be vitally important to measure the media and its interaction with the environment, over time.
We should treat computer media* as an ecosystem, sample the code in it, and use techniques adapted from metagenomics to objectively measure the code for life processes. If life processes are occurring in new media, it would be incredibly important to sample and measure the changes in the media, over time.
* Computer media includes memory and processing units. Programs and data are found in computer media.
Data collection in genetic media is expensive and slow.
Data collection in computer media is inexpensive and fast.
Questions we may ask and answer:
- Will we be able to "bin" code into contiguous groups of recurrent code patterns? Yes. These are already present, typically (though not exclusively) as compiled executable files. Both the "easy" way to bin (with a starting library of signatures of groups) and the "hard" way to bin (without a starting library) can be performed in computer media.
- Will we be able to characterize code pattern groups as "species" of code? Yes. This is just another name for "software release"; though code binned together statistically will include executables plus material produced by interaction of the executable with the operating system, hypervisor, and kernel.
- Will we see code that is shared? Yes, open source projects ensure that many executables share code. On a fine-enough grain, all code is binary or hexadecimal and is, in that sense, shared. Code sharing can be measured.
- Is code sharing increasing or decreasing over time? For which code groups?
- Will we identify changes in "species" over time? Yes. This is already occurring in software updates and releases. We can measure this objectively.
- Can we assign functions to code patterns? Yes, this is readily available information. For example, runtime decompilers perform this service during speculative execution, to identify memory contention and other conditions in executing code. Functions and function agglomerations can be identified and assigned arbitrary identifiers. Over time, some of the identifiers can be mapped to descriptive identifiers.
- Do different of the "species" communicate? Yes. Through APIs, shared memory resources, and the like.
- Will we be able to identify a network code patterns that have positive feedback with the reproduction of more of the network?
- Will we be able to measure energy flow through the media?
- Are the code patterns coalescing and evolving over time?
- Are both writing software and managing its execution in hardware becoming more automated? Yes they are.
- Will we identify one or more signals of life, using techniques from biology?
- Will we be able to distinguish human generated signs of life from those from a new, distinct, form of life? Will computers communicate at a faster rate and in a larger volume compared to human communication?