Everything is complex (awesome) / by Callum Lamont

Some things are just complex. Everywhere you look you can see more examples. This most recently came to the fore of my mind during a bit of background reading for my PhD. Here I was looking into the encapsulation and protection of integrated circuits. However, before I could investigate that, it would be useful to read about their fabrication techniques. Namely, how the components constructed on silicon wafers are ultimately made useful via the deposition of precise metal tracks (which are unfortunately quite susceptible to corrosion). But wait, how are those nanoscale structures (transistors, capacitors, diodes etc.) patterned into the silicon wafer in the first place? Ok so using using photolithography, a mask and a process to implant dopant impurities in the surface, the functional interfaces within the material are created which then control electron flow. What's the photolithography etching off though? Ah, there's a thin silicon dioxide layer formed in the previous step, to insulate and isolate the material, as well as precisely allow selective diffusion of dopants just mentioned. This is pretty easy to create. You just need a furnace and exacting controls over the flow of necessary constituent gases. But before we do any of this, we need an ultra flat perfect wafer of single crystal silicon. First, this requires electronic grade silicon of ultra high purity (too many steps to bother explaining). Then we need to turn this into a single crystal, requiring all crystallographic defects within the microstructure to be removed, including even minute atomic gaps within the lattice. The process essentially requires us to melt then, carefully and precisely, allowing the silicon to solidify in such a way that the thermodynamic and kinetic gods grant a single crystal to emerge. Production of a 3 meter ingot will typically take several hours. After which it can be sliced into wafers and then polished to produce an immaculately flat surface, using abrasive powders with diameter of 1 micron and then a chemical etch.

 

Then of course there is the degree of control and precision required for all the equipment used in patterning the 14 nm structures on the wafer. Not to mention the complexity and conceptual barriers to interface the simple operations of a billion transistors (either on or off, 1 or 0) with higher level software we can intuitively grasp. I point all this out so when your phone takes slightly longer than expected to load up your emails, you may be a little more appreciative of the countless man hours and ingenuity that allowed us to get to this point. As an aside, why is it so easy for us to quickly transform such complex advancements to an assumed convenience, then to unquestioned rights. How come when one is given an iPad, they generally take the time and effort to understand how to work it, but not how it works. Perhaps it's the incremental advances, whereby an iPad more or less operates in the same fashion as an iPhone, which is like a smaller computer/larger cell phone, which is the offspring of a sequential line of ever improving technologies. I am definitely guilty of this practice. Despite having driven a car almost everyday for the last 5 years, I have only the most broad understanding of how they actually work. I suspect there’s some evolutionary pressure to minimise our cognitive load where possible (aka being lazy). 

 

In spite of this rant, at least we (I mean someone else) knows how these things work. The matter gets more grim once you start venturing outside this bubble of technological order. Take the economy. Now think about how it really functions. It is the sum influence of millions of people buying, selling, saving and fretting (though not in an entirely egalitarian fashion) through an opaque and entangled web of interactions. Another layer of obfuscation is imposed when considering that not all players are acting in the most rational way. Some people have a better grasp (or at least think they do) of these interactions and can use it to their advantage. The end result being we now we have a few somewhat parasitic industries as a byproduct of this convoluted system of globalised trade. Other people take this assumed knowledge and try to apply it for good, helping guide policy decisions to improve employment and our quality of life . However, in the end, I view economists akin to weathermen, trying to forecast the dynamism of an intricate system where the whims of chaos theory have long since overthrown their initial presumptions. 

 

To surmise so far, in the scale of complexity, the haphazard meanderings of many seems to outweigh the directed and focused efforts of a few . However, the most intractable doesn’t involve even more people, just the one. You, or more specifically, you're body. In my opinion the most complex intertwined arrangement of information which, in its tortuous pathway of feedback loops and self-direction, miraculously results in you being here today. However, despite such a forbidding task, human ingenuity has once again allowed us to wrestle this concept with a certain degree of success. A great deal can already be spoken for the advancements in genetics and genomics. Considering Crick and Watson only proposed the DNA double helix a touch over 50 years ago, it's surprising the lengths we have come. The ever improving literacy of our genetic signatures may allow us to eradicate many human pathologies. Headway has also been applied to some other interesting avenues, such as creating crops which generate greater yields and are less prone to our increasingly volatile climate. Another particularly creative use of this biotech is as a tool for manipulating mosquito populations in the fight against malaria. As usual these technologies are double edged, with much caution needing to be paid towards the serious ethical implications involved. This is a level of complexity I'm not even bothering to touch, but if you want to use your imagination I’ll quickly point out that prenatal testing is getting to the stage where mothers could receive a blood test, from which we can reconstruct the genome of the foetus. What individuals do with this information...

 

Now, for all our mastery of the genome, we're not significantly closer to fully apprehending its secrets as we are that of the economy. Sure we have some broad, big picture ideas of how this machine operates, however, many of the notable tools we use for its manipulation (i.e. restriction enzymes, CRISPR) have come about more by chance than genius. Don’t be fooled, we are not creating these tools from scratch. They are derived from bacteria, in which they originally served as a line of defence from viral invaders, and have since been repurposed for our ends. It's like an alien civilisation leaving us cars and all the tools necessary for their construction, and from that we could develop..say.. trucks (or maybe only slightly different cars). That aside, the great challenge is not manipulating DNA, but understanding it. Each cell contains 3 billion base pairs, of 4 different distinctions, ordered (to varying degrees) into genes. Genes which can play multiple complex, yet seemingly unrelated, roles. Genes to express other genes. Sequences which promote or suppress expression of genes leagues down the genome from where they reside. And at this stage were largely taking about a system contained within itself. We can delve down the rabbit hole further and start considering the effects of environmental factors (e.g. stress and obesity) that can affect our epigenetics, which can then be passed on through generations.

 

A large degree of the aforementioned complexity arises from the feedback nature of the system, in which the end stage of this stream of information processing (enzymes and other proteins) can then act back onto the initial template. And despite my derision, at least we're making some semblance of a headway into this field. There is another another aspect of the human body that remains largely encrypted, cognition and consciousness. Similar to DNA, one can think of the brain as structured into different layers of information processing. In the genome we have the basepairs, these clustered into threes (called codons), which then code for amino acids in proteins. The structuring of information in the brain is somewhat more complex. Firstly we have single neurons and their inter-connections, then we have... something, then we have ideas thoughts feelings and desires. And surely there must be something in between, as neurons act in a very mechanistic sense, operating by simple physical rules and cues with regards to their inputs from other neurons and with an output of either fire or not fire. Obviously these neurons are not self aware, but how can they coordinate to create a system which is? We can make one step in the right direction by comparing this structure of information processing not to a biological counterpart, but to a digital one. As mentioned earlier, the base material foundation for electronics and computation is only one party in this tango of complexity. The innovation really lay in developing a method of translating it's very simple output (0 or 1) into higher order software. This is aided by compiler languages, which begin abstracting raw sequences into ideas or concepts more relatable to a human developer. Naturally, no one would expect it possible to extract information from the seemingly random noise of 1’s and 0’s output by a computer. Following this line of thought, it is supposed that if we step back a degree there may be an analogous higher order pattern of signalling between clusters of neurons, which carries this information in a more relatable form. It has also been proposed that such higher order signalling in the brain may feedback on itself to influence subsequent neuronal firing and thought patterns, in the same way that enzymes act back on the DNA which created them.  

 

So if thought, ingenuity and consciousness all arise from a complex representation of information, which at its core follows a very simple reproducible mechanistic procedure, then surely such processes are transferable into a nonbiological medium, say a computer chip. If possible, then I believe true artificial intelligence will eventually be achieved. Now you may have heard a lot about AI in recent times, particularly with advent of deep learning and Bayesian networks. However the more I read about these, the further I think we are from reaching true AI. These techniques are, at their core, pretty dumb. They are becoming more human than in recent years, however I would define intelligence as a global flexibility. Being able to go from absolutely nothing to eventually extending the topic past it’s original application. Think about how you use previous rules in language to create new words, maybe for comedic effect or maybe to help articulate your point. Additionally, we can also consider those who hear this use of language and, through drawing upon a vast network of discrete and distributed knowledge, piece together the implied meaning. This is so quintessentially human than not even all humans can do it all the time (there’s always one mate who sheepishly asks for clarification). Such flexibility, in adding, combining and reinterpreting previous concepts, in reasoning outside the existing framework of thought, is what I feel underpins true intelligence. Now, none of this is an argument against our inability to create AI, it’s just why we don’t really appear that close right now. Ultimately, it boils down to the fact that it’s just too bloody complex (but we’re still awesome)