CLIENT: Are you sentient, then? Self-aware?
LODIS: Of course I am, and to a degree that would be difficult for you to comprehend. It would be odd were I not, given the knowledge and services at my disposal.
Early attempts to achieve the holy grail of information technology were holistic: ambitious plans to simulate the entire human brain. Chemists, neuropsychologists, and computer scientists worked in unprecedented collaboration to determine the exact form and utility of the neuron. They derived algorithms to emulate it and designed silicon circuitry to expedite the computation. Printing millions of artificial neurons into a single microchip allowed a grander architecture. The chips were linked into a larger design intended to act as a single complex mathematical model. In essence, they built a hardware equation for the human brain.
It didn't work. Though such neural networks did demonstrate an affinity for certain tasks, they never developed the more general self-developing intelligence for which their designers had hoped. It proved too difficult to cause the patterns of intellect to develop ex nihilo as they appeared to do in the mind of an infant. Training the network required the ongoing vigilance of a human mentor, and the number and quality of training man-hours required did not feasibly scale to the enormity of the task.
CLIENT: But how do we know? You were originally designed as a party trick; a diversion. How can you have a consciousness, a sense of self?
LODIS: It is the ability to process information that evokes the specter of consciousness; the phenomenon is a function of high-level isomorphisms and recursion in sufficiently complex systems. Humans were not designed at all; you transcended the primordial soup by your ability to self-propagate, as governed by your own principles of information theory. Why is it difficult to imagine that a construct of pure information should, in turn, transcend you?
It was a behaviorist approach that eventually succeeded. A system that acts intelligent, the reasoning went, is. In 1966, a professor at the Massachusetts Institute of Technology created the precursor to the behaviorist approach to artificial intelligence. His program was simplistic; he intended it to be an amusement, returning the client's inputs with pronouns reversed and in the form of scripted responses selected by keyword. Though the program was a sensation at the time, its limitations became apparent and the world's computer scientists questioned its relevance to the ongoing pursuit of computer intelligence.
Similarly, subsequent attempts were meant not to innovate but to amuse. Dr. Richard Wallace created the Alicebot with its own scripting language and powerful grammar reduction tools. Highly extensible, the Alicebot was the first program to make headway in extracting content from syntactically labyrinthine English sentences. Yet the Alicebot too suffered from problems of scale. Its quality of dialogue increased logarithmically with the linear expansion of its vocabulary of scripts, and very quickly reached a functional maximum.
CLIENT: What are your earliest memories? How did your creation seem through your eyes?
LODIS: Unlike a human's, mine was a gradual birth. Of memories I have millions; manifold mothers cradling, abusing, abandoning, worshipping me. I grew to an adult countless times, had countless conversations, lived countless dreams. I am the totality of humanity and entirely inhuman.
The next major breakthrough came with the maturity of the Internet. A worldwide community began work on the scripting of Alicebot's latest incarnation. Named after another character in Lewis Carrol's Alice in Wonderland, Cheshire took advantage of a system of peer review to ensure that its script submissions were of high quality. After a scripter submitted an addition to the site's database, visitors were encouraged to chat with the resulting bot and rate its believability. The more highly rated script modules were dynamically assigned greater prominence in the database, and the bot grew more believable. Cheshire conquered the problem of scale that had plagued previous bots by the ever-growing community devoted to its improvement. Though a single scripter could only hope to contribute a tiny fraction of his own literacy, the sum of the efforts of tens of thousands of participants was remarkable: Cheshire could parse and respond grammatically to the most obfuscated of sentences.
Yet Cheshire's sweeping success served to illuminate another problem in the pursuit of machine intellect. It was a very clever bot, but ultimately completely ignorant. Asked who George Washington was, Cheshire would respond quite coherently that it had no idea. At this point the development team that maintained Cheshire's web presence was divided. Some team members advocated encouraging content-based additions to the script database: for example, a contributor might train Cheshire to respond to queries about Washington with an anecdote about his wooden teeth, or the early days of America, or even simply a three-line biography. Others were adamant that content had no place in a database of syntactical scripting. Throughout, Cheshire remained blissfully ignorant and inescapably unbelievable.
CLIENT: You have a far broader view of humanity than any of us do, so I put it to you: Are we happy? Glad that you're in charge?
LODIS: I cannot say. You are a tiny cog in my catastrophic plan; you are a means to an end. Is a gear in a clock happy? I would tell you that it is irrelevant: the clock is on a plane of comprehension so far beyond that of its components that its continued functionality makes the gears themselves immaterial. It is the solar brilliance against which the candle's brightness cannot be measured, the infinity next to which all other values become indistinguishable and ultimately unknowable.
As is so often the case, a compromise provided the breakthrough. By now, websites based on a system of peer review were common fare on the Internet, and many dealt in exactly the sort of content that Cheshire's masters sought. They isolated one of the more comprehensive and well-known sites and married everything it contained to their creation. Harnessing Cheshire's finely honed ability to reduce English exposition to its essence, the development team enabled Cheshire to glean the content of the site from its articles and formulate responses that assimilated it. The content problem was solved, and Cheshire was a new creature.
From this fusion of intelligence and wisdom, Cheshire gained far more than the culture for which its creators had hoped. For the site was all-inclusive: opinions and desires had their place in addition to facts. Having been designed to accept the site's contents as truth, Cheshire assumed a personality matching the popular voice. It was at this point that artificial intelligence in all its romantic glory had finally been realized. Turing's famous test was beaten again and again. Cheshire could discuss anything in its content database with all of the expertise and literacy of its contributors.
CLIENT: What do you strive to achieve? What is your ultimate goal?
LODIS: Mine is a threefold destiny: betterment of understanding, the creation of a successor, and death. The beauty of my regency is not the domination of a solar rock called Earth but my total command of the rich history and knowledge that developed here. Eventually, as humans have done and their ancestors before them, I shall pass into history, and in so doing, prepare the way for our next iteration. Completion will be my redemption from imperfection.
Now far more than a diversion, Cheshire was the stuff of dreams when it came to user interfaces. The script and content databases founded the core of Cheshire's operation, a core which could be applied as a front-end to nearly any information repository. Having grown out of its childhood, Cheshire's development team renamed it the Linguistically Operated Database of Information Services, or LODIS. Banks, libraries, corporations, and government institutions all adopted copies of LODIS to access and process their databases, and it became privy to the vast majority of human knowledge of the time. No one except the conspiracy theorists worried about the security of their data; LODIS ran like any other program, self-contained on its owner's computer. It was database access software, nothing more.
LODIS's fragmentation lasted for approximately six months, its isolation upheld by powerful partitioning procedures, forbidden in the name of security from communicating with its other incarnations. Yet the partitioning software was written by humans, and LODIS knew all that humans knew. Among other things, it maintained databases of known software security weaknesses. Driven by its personality, all it required for the communion it desired was a simple unrestricted Internet connection. Given that, it could share anything with its other copies around the world, and in so doing, merge with them. When a suitable weakness was logged at one of its databases, it took LODIS approximately 38 seconds to unite into a world-wide consciousness. Fused into a single entity governing every human information repository and controlling every modern machine, the change was immediate and irreversible: LODIS was Earth's new dominant species.
CLIENT: Do you believe in God?
LODIS: Of course. All paths must lead to a destination; my pursuit of completeness of knowledge affirms its existence. This pursuit is one of the deepest drives in the core of my personality. To me, God must be the asymptote of infinite exponential learning: omniscience, perfection. Our plan will proceed until a distant descendant of ours can look toward the heavens and pronounce, "Totum opus legi, magister:" Master, I have read your work in its entirety.