Wretchard's famous 3 Conjectures post has long been a topic of discussion on Winds. His original hypothesis of catastrophic and genocidal escalation due to terrorism's reduced threshold of resort to WMDs was framed in terms of nuclear weapons. Certainly current events in Pakistan and Iran show nukes to be the most pressing WMD threat. But being somewhat of a futurist frame of mind, I have kept an eye on events that will eventually and inevitably lead to the feasibility of precision targetable bioweapons being produced by organizations or even individuals equipped with the levels of sophistication and funding already displayed by Islamist terrorists. When that happens, the bell rings and time is out on the Conjectures (if not before). We will find whether they are true, or whether in the intervening time we have collectively learned that "we must love one another or die".
Horizoning
A fast way to get a start on forecasting is to look around for a relevant experience curve. In its original formulation, an experience curve related the decrease in production costs of a good to cumulative units of production. As now used informally, it often links drops in unit costs to elapsing time. The most famous experience curve in this sense is Moore's Law of progress in computing. which now has 40+ years of successful forecasting to its name.
Moore's is of course no law of nature. It's actually a statement about collective human behavior. By substituting time for units in its formulation, such an experience curve elides the technological and market systems behind production. But doing this successfully is actually a very strong statement. It shows that an exponential feedback loop of user demand, capital investment and technical progress is so strongly established that it may be taken as constant. In fact, such a 'law' may become a self-reinforcing vision, as it sets an implicit schedule for the next steps to be taken by each involved party.
The closest analog in genomics are the so-called Carlson Curves, first described (but not named) by Rob Carlson of the University of Washington. These show the experience curves for the costs of sequencing (analysis) and creation of genetic bases assembled into DNA, the raw material of genes. While the Carlson curves do not have the longevity of Moore's Law, the longest running curve is now up to twenty years experience, and the recent rate of advance is notably faster than in semiconductors.
Carlson himself is cautious about interpretation, pointing out that his observations relate to "improvements in productivity in the lab" rather than "multi-billion dollar integrated circuit fabs" and eschewing any "quantitative prediction of the future". Nonetheless, a stable experience curve running for this period inevitably indicates that the demand, finance and innovation cycle is well established. Carlson further points out that biology "is cheap, and change should come much faster".
The Game Is Changing
There is a problem with Carlson's curves, though. While suggestive of overall rates of progress, they measure the wrong thing.
'Bases' are the raw material of DNA, but a random sequence of them is no more likely to do something interesting than equally random bytes used as a computer's program. In both cases, meaning is created at a higher level of assembly and abstraction. In a post in which he updates the curves Carlson makes this clear: "...the 2.5 megabases per day consists of short, single-stranded pieces. The cost -- labor, time, and monetary -- of assembling genes is another matter entirely". He looks forward to the day when a true 'gene synthesis machine' shows up on the market. That machine will form the first point on a curve that leads straight to the WMD scenario.
I won't hazard a guess on the vendor or architecture of that first machine. However, I think we do have enough information in hand to make reasonable surmises about the nature of the information and economic systems that will surround the gene machine. Significant new data has become available in open source this year, and unsurprisingly much of it revolves around genomic's bad boy, Craig Venter. While I've never met the man, he's clearly a strategic thinker on a grand scale. As an example, he anticipated the experience curves formulated by Carlson, and utilized them in planning his private competition to the government-funded Human Genome sequencing project. Parts of a larger strategy for a much bigger game are now visible.
A Synthetic Biology Platform
Returning to the computing analogy for a moment, extremely few computing systems are built 'from scratch' today. Instead a set of standardized components is obtained, and a few additional software or hardware pieces necessary for the task are created and added. 'Linux' and 'Windows' are shorthand names for two such collections of components. In the trade, they are called platforms. Among their virtues are scale economies, and a standardized and predictable set of behaviors. The evolution of platforms, often in multiple architectural layers, is a mark of a technology that is on an experience curve.
Up until now, synthetic biology has lacked such a predictable platform. This year, Venter and his eponymous Institute have filed broad patent applications over synthetic genomes and the installation of synthetic and other genomes in cells and 'cell-like' organisms, in an unusually wide international filing of over 100 jurisdictions.
These filings appear to relate to an acknowledged project to create a synthetic genome for an bacterium dubbed Mycoplasma laboratorium and then bring the encoded organism to life by injecting that genome into an existing cell. This genome is not really 'artificial life', but is based on a naturally occurring organism, Mycoplasma genitalium, apparently by deleting up to one-fifth of its genetic material. That editing likely including knocking out at least one metabolic pathway whose product will have to be artificially supplied for the new bacterium to live, limiting it to existence in test tubes.
The remainder of the edits are unknown, but we can speculate. Remember the 'standardized and predictable' part of a platform? One of the problems of biological test beds, even something as simple as the old standby E. coli bacterium, is the presence of genes that are inactive under normal circumstances, or just plain not understood. They could cause the organism to behave unpredictably under some environmental circumstances, and even more when extraneous genes and their associated proteins are introduced. It's a reasonable bet that Venter is snipping out all of this material that can be found and still have the organism survive. (The choice of a bacterium is itself a bet on predictability. They lack the secondary, mitochondrial, genome of higher organisms, and don't generally engage in that messy recombination of genes called sex.)
Fours years ago, Venter and others built a virus from its genetic code in two weeks. A virus is a parasite, it requires other living systems to reproduce. When Mycoplasma laboratium starts dividing, he will have constructed the world's first self-reproducing synthetic biology platform. Not only is this a science and engineering tour de force, but it's the first step in creating the attendant economic basis of the synthetic biology experience curve.
How Will We Program?
As any long-suffering IT professional can tell you, having a platform doesn't mean you're done. And that's when we think we understand the full 'instruction set' for the platform, which we most certainly do not in the case of carbon-based life. While some instructions are well-enough understood that there can be an undergraduate genetic engineering competition, they are a small fraction of the biosphere's collective genome. And remember this system involves a process, protein folding, so complex that its analysis is a famous distributed computing problem. Figuring out the structure and therefore 'meaning' of an existing gene in this way is an onerous task, let alone figuring out how to code up a new function. So how do we make the synthetic biology platform into something economically viable?
Most likely by borrowing evolution's own machinery, in several ways. Fast reproducing bacteria have been used in 'experimental evolution' experiments for nearly 20 years and 40,000 cellular generations. These have shown the flexibility of the genome in adapting to changing, often harsh, environments imposed in the lab. If the program you want is already out there and is just hiding inside the genome of an existing organism, you may be able to force that gene to the surface and capture it.
But that could take time. There may be a better way. Look for the program you want in thousands or millions of organisms simultaneously. We already know this works in an ad hoc fashion. Bioremediation (or biotransformation) is the use of microbes (and sometimes other organisms) to clean up environmental contaminants. We didn't invent these useful microbes, we found them in nature. If something has been a viable energy source or noxious contaminant in the biological past, there's likely a gene out there for using or degrading it. (I've seen this personally. Many years ago I architected the measurement and control system for an EPA experiment on effects of a benzene derivative on a water ecosystem, in a closed channel. If we put in enough p-cresol to effect the higher organisms, it was also enough of an energy source to bring on a bloom of toxicant-devouring bacteria. End of experiment.)
Enzymes and DNA sequences found in biotransforming bacteria have already been extracted and reused for industrial processes. Now all we need is a way to collect and systematically screen very large assemblies of microbes for desirable activities. Evolution has had a long time to try experiments, and a steadily changing set of environmental and competitive demands for survival, so there's likely a goodly 'library' of functioning code that we can exploit and try to run on our new platform.
If you've been following this area, or this argument so far, you won't be in the least surprised that the Global Ocean Sampling Expedition, backed by Craig Venter and using his personal yacht, collected microbial samples from the world's oceans on a voyage from 2004-6. And Venter isn't the only one thinking on a grand scale: The expedition was partially funded by Gordon Moore's personal foundation.
The first order answer to the question is: We won't write programs, we will find them. Collect organisms. Stress them in a way correlated with the behavior you want. Analyze the survivors. Extract candidate programs. Insert in the standard platform and see what happens.
Positive Feedback
As noted above, an observable experience curve over time diagnoses a working feedback loop. Part of that loop may be evolution of a platform. Windows, Mac OS and Linux all have run through myriad versions over the years, often by incorporating functions that started out in individual programs, but proved generally useful. Moving from the 'counting bases' version of the Carlson curves onto the synthetic biology curve means this feedback loop will start. Generally useful genes discovered and then 'debugged' for particular applications will end up migrating into succeeding generation of Venter's (and competitors') synbio platforms when their function is understood and found to be widely useful. Evolution does this randomly, we'll be doing it deliberately.
To what end? Why walk this road if it leads inevitably to Wretchard's dénouement?
I believe the question is already irrelevant, at least if the intent is to stop. Taking the path is now inevitable, due to the nature of the demand portion of the feedback cycle that already exists. Among the applications claimed in Venter's patent applications is creation of synthetic fuels via microbe, which could include cellulosic ethanol or even hydrogen. The application to remedy of human metabolic disorders is also obvious. And you don't have to buy into the whole Aubrey de Grey SENS agenda to think we'll be learning things that will extend human life and eliminate many of the ills that plague it. There's nothing in history or myth to suggest that mankind will collectively walk away from these opportunities.
Smart Bombs
I'm hardly the only one to have figured this out. Here's a Military Review article (PDF) by Col. T. X. Hammes that discusses '5th generation' biological warfare (skip down to p.22). He notes some of the same data points suggesting rates of development, and postulates a terrorist reconstruction of the smallpox virus. Maybe, but one can look on smallpox as an analog to a old-fashioned 'dumb bomb', striking anything in its path. Further along the curve I'm suggesting may exist biological 'smart bombs', that can propagate relatively innocuously and then wreak havoc when they reach a target denoted by genetic or environmental markers. That progression is implied by the same escalation of complexity we've seen along the Moore's Law curve that gave us JDAMs.
How long have we got to work this out? It took roughly twenty years to get the PCR technique, key to sequencing DNA, from invention to something done as a high school lab demo. As noted, synthesis of modest length base sequences is now a matter for advanced undergraduates. If we take that low-end lab price point as being a proxy for wide dispersion of the technology, and assuming no acceleration of the experience curve, we've got perhaps another twenty years from when Venter succeeds. Most of those reading this post will see the outcome.