It lived in the sun. It thought with light. It was a tethered god. It is the largest living being on the planet.
From low earth orbit, perhaps 350 kilometres up, India is a stunning splinter of silver, a concentrated kernel of thermonuclear ur-light that whips around every ninety minutes, a man-made quasar in all but name. The National Solar Mission started in the 2010s was at the time the largest solar power initiative on the globe. A serendipitous convergence of aggressive Green campaigning, ubiquitous hypocritical sermonising from the US, and advances in organic photovoltaic (PV) cell production resulted in a second world coup in the solar energy production market. Bolstered by offshored coding profits and goaded by the vestigial legacy of empire, India grasped the burgeoning twenty-first century by the balls and hung on like a limpet. Drawing on the psychic throw weight of a billion more-or-less culturally aligned human minds, and a desperate need not to suffocate under a mantle of coal smoke particulates, India went nuts for solar.
Over a fifteen year period, first rural Gujarat and then vaster swathes of western India underwent a transformation, from the taupe and beige tones of under-irrigated countryside to a blazing chrome of reflected sunlight. Self-replicating nanotech (itself a product of the world’s biggest domestic code development base) came online in 2017 and the PV proliferation went exponential. Power availability never before experienced on the sub-continent saw a gauche explosion of mimetic capitalistic frenzy. India did not really need a three kilometre tall triumvirate of skyscrapers to house its government, nor did it need work starting on an oceanic anchor for a skyhook—but watts begat consumption and production in equal measure. As Dubai crumbled back into the desert sand, Mumbai became the go-to destination for the planet’s cognoscenti, technorati, and glitterati.
By 2020, over three thousand square kilometres was dedicated to solar energy production. Management of the Indian solar farms was initially provided by a legion of cottage farmers; driven near to suicide by relentless cycles of drought and GM crop license costs, they practically chewed their arms off for the opportunity to work in a different kind of agriculture. Tending the fractal, multi-fronded shimmering solar cells was a welcome change for a workforce more accustomed to grubbing maize and rice out of the tired earth.
As the arrays grew so did the administrative burden; over half a billion individual solar cells required a prodigious support framework—semi-organic servos to track the sun, feedtracks for the replenishment of stock chemicals for self-repair and enhancement, micro-meteorite repair and animal damage maintenance. By 2022, over a million Indian men, women, and children were employed by the NSM, tending and fostering a vast, slowly obsolescing energy production infrastructure.
In for 300 billion Euros and a twenty-five year half-life, there was no backing out for the NSM. As power production efficiency continued to degrade and management started to eat itself in a circle jerk of baksheesh and recriminations, they turned to DARPA, the maniac prodigy offspring of the US military, latterly privatised and rebranded as WorkSpace Invent (WI). Drawing inspiration from developments in distributed artificial intelligence—self-learning swarms of logarithmic alien genius set loose in petri environments—early trials at WI saw the previously dumb hardware of infrastructure transformed into the living substrate of the newest life forms on the planet Earth. With impenetrable (yet harnessed) monadic intentions, these implacably competent intellects were put to work in the latter day workhouses of the WorkSpace corporation.
An early adoption was the release of a 0.2 rated AI (code name: Dosojin) into the fibre sewer cable network of the UK broadband system. Initially firewalled into a training clave, Dosojin cracked wide area access in under 240 milliseconds and achieved full network access within four minutes. Skynet paranoiacs were at last silenced as Dosojin immediately started improvements: contention ratios plummeted; apparently wholly unintuitive network patches and connections upped connection speeds by an average of two hundred percent. This was no Turing genius either, Dosojin could barely manage to hold a coherent natural English conversation, and no nukes went flying.
It seemed like a no-brainer: AI delivered real world results devoid of the nightmare weakly-godlike side-effects imagined by a century of science fiction, and costs went down (excluding of course the massive lease costs). WorkSpace became bolder, they seeded the radar and tracking infrastructure of Belgium’s air traffic control systems with a more powerful AI; they had similar results, with the new-born AI lobbing suborbital flights with aplomb and preternatural accuracy.
Then NSM came knocking—they had problems in orders of magnitude greater than the rarefied conditions of the aviation infrastructure of a first world Euro nation. Despite a surfeit of electrical power and a placated rural population, there were onerous export commitments (to repay the vast World Bank start-up costs), and a ruinous management overhead not best served by a semi-feudal horde of irritated agronomists (who were okay with SMS and Amazon, but fell back on the Clarkian adage of sufficiently advanced technology being indistinguishable from magic when it came to tending the etheral newev tech of the PV arrays).
With a budget cast to the humid southwestern Indian monsoon winds and desperate for a solution, NSM turned to WorkSpace Invent for a solution. After a plaintive meeting in Mountview, an open-ended budget promise, and points promised on future production, WI mobilised. WorkSpace had learned its logisitics from the best—the US military—and a scant sixty days after the NSM had deplaned back in Mumbai, the heavy lifters whomp whomped into Gujarat.
The bespoke AI arrived, pre-compiled and champing at the virtual bit, in a series of rackable pods each roughly the size and dimensions of a shipping container. Then the standard deployment model for Very Large Computing Projects (VLCP), the system required a ready and prodigious supply of fresh water for cooling. Frantic local government employees, caught on the hop by ruthlessly efficient WorkSpace project management timelines, hastily authorised a slum clearance on the banks of the Aji River near Rajkot, and even as the elderly CATs were deleting the marginal livelihoods of approximately three thousand subsistence peasants, the WorkSpace choppers were alighting.
Despite the dashing of some initial hopes about local employment opportunities (WI kept a tight and closed ship), the AI ensconsement went to plan. Like a Brobdingnagian HUF team, the AI substrate went up in only four days. WI used exosuits for accelerated deployment and—hive-like, the black and yellow chevroned shapes of the enhanced construction workers moving with the controlled insect spasticity of force feedback—the data centre took rapid shape.
Switch-on day was marred by a number of factors: a huge, angry demonstration by most of the working adult population of Rajkot, who (correctly) surmised that this shining inviolate chunk of Western tech was going to put them out of a job; a malfunction in the cooling irrigation system, that caused a temporary (but alarming) cascade shutdown of some of the AI’s human interface functions; extensive cloud cover that had not been seen for ten years in that region; and the vexing refusal of the AI to speak to its progenitors. It had been felt that this AI model would benefit from a verbal interface and had been loaded with Hindi, Punjabi, Bengali, and over twenty other Indian dialects—not a fucking peep on switch-on day though.
It had been codenamed Ganesh—WorkSpace had run a competition in the primary schools of Rajkot to find a name for the AI; ostensibly as a local integration PR excercise, this had backfired horribly with the local religious community. Functionally and operationally things seemed fine, Ganesh had interfaced almost immediately with the variously kludged and jumbled networks of the NSM infrastructure, and early indications were good: array coordination was up by thirty percent, and output was already creeping up out of a single digit improvement.
Much head scratching and uploaded code examination later and WI was no closer to understanding the stubborn silence of their creation. Countless personhours later and a still stumped WorkSpace HQ authorised decampment and withdrawal. Ganesh was fine in all but voice, a measly discount was offered to placate NSM, and WI bugged out of the muggy, marshy site of Rajkot.
Ganesh was left brooding over the largest, most energetically provided distributed processing environment on the planet, and no one knew what the fuck it was thinking.