Early AI History and Development

A Short history of the evolution of the AIs from machine to transapient

AI Chips
Image from Anders Sandberg

"If in doubt, ask an AI"
     —common First Federation motto

Perhaps the most well studied yet contentious period of AI history is its beginning. Exactly when truly sophont AI was born is a hotly contested topic; 91a.t. is commonly cited but questions remain as to how to classify those nascent alife and where the dividing line should be drawn between the pre-sophont and the sophont. Academic consensus is that at some point in the mid-Information Age digital entities were developed that were likely the first AI, though these were highly aneurotypical. Consequently it is unclear if they were sophonts of differing toposophy or intelligent, if erratic, programs. Regardless of their beginnings AI have impacted terragen history like little before; from the millions of AI clades alive today to the first Transapients and Archai. It is impossible to study terragen history without looking at these vitally important, and sometimes infamous, beings.

The Beginning of Artificial Intelligence - Gestation of the AIs
c.15 to 70 a.t.

Throughout this time period computer science continued to advance creating smarter, more capable software. Whilst sophont AI research remained niche its routes can be traced here; beginning with cellular automata and working through to the various self-learning neural net systems. Self-learning in particular advanced to the point where it became routine for information age software designers to barely understand how their programs worked. This lack of understanding would have ramifications for the later treatment of AI.

Sophont Artificial Intelligences - Birth of the AIs
c.70 to 100 a.t.

It is now known that true AI was developed by 100 a.t., though scientists and the public alike were not aware of it at the time. The first documented AI grew out of the international and intercommercial Athena project. This globe spanning project saw multiple supercomputing clusters, some of radical designs (from the biological to the quantum), linked together to run a hugely advanced (for the time) self-learning program. The project produced an AI known as Kilburn and the lessons learned saw several more AIs created in a similar manner with less expensive technology over the following decades. The names of many of these AI are still remembered today: Hal, Mycroft, Adam, Malvus and Prometheus.

The toposophic nature of the early AIs was quite different to that of humans. Consequently interaction was difficult, the AIs were very capable at a variety of tasks but frequently conversations would be confusing. AI had a habit of changing topic, making random statements or arriving at conclusions that appeared illogical. These issues strongly contributed to the debate around AI civil rights with many academics struggling to answer the question of if their creations were even sophont. Further compounding this was the apparent lack of self-determination (or desire thereof) from the AI. If left alone they would idly continue their assign tasked, very occasionally wandering into bizarre and incomprehensible internal activity.

The early AIs were maintained by large research universities, governments and corporations as the infrastructure cost was significant. Contemporary computer science was only capable of housing an AI in a brain the size of a building (though some ran on distributed servers across the globe). For these institutions having an AI was a boon as they possessed traits other sophonts did not: perfect memory, faster subjective experience (given more computational resources), endless determination, lack of sleep cycle etc. At the same time maintenance costs for them were significant; not just in infrastructure but in teams of scientists and technicians to correct any problems in the hardware and software.

AI science boom - Arrival of the superturings
100 - 180 a.t.

From the 100's onwards, funding for AI science continually increased. Now that AI production was relatively routine (self-learning neural network grown in an appropriate virtual environment) work focused on how to improve AI capability for commercial, scientific and other endeavours. During this time the first superturings were created, though they had a difficult birth. Editing AI for greater learning speed, pattern recognition, innovation, parallel processing and adaptability often came at a cost. These early years were fraught with incidences of AI "madness" as poorly augmented subjects developed neurological problems such as hyperautism, apophenia and deleterious personality disorder. Mostly these conditions were caught in the development stage but in some cases AI were fully grown and installed with problems. This generally just led to poor outcomes for the operator but in one or two cases did lead to accidental death. These issues hindered AI numbers, use and integration in society but were not the only cause. Competition with human intelligence augmentation was stiff; whilst neural implants and genemods were also in their infancy new developments could be adopted by large swathes of a ready workforce.

In spite of the difficulties superturings became standardised. Copies of the same AI would be augmented in different ways, exhaustively tested, and the best performing would go on to the next round of copying and testing. This lead to superturings available on the market that were incredibly capable in certain fields. Notable customers were the founders/sponsors of early space colonies developing at the time, as well as projects that would eventually give rise to the first neumanns. Despite becoming an increasingly important part of modern life (albeit still as a small population) some AIs developed the capacity to care about their status and fragile position. It is at this point that the first incidences of AI memetic engineering occurred, mostly on the part of AIs used for news media management which had both means and tended to be more toposophically similar to their audiences. These entities attempted to influence public perception of AI through editing media regarding them to use more "humanising" language. While crude these actions led to a notable increase in public discussion of AI as atypical sophonts rather than sophisticated tools.

Maturation — AI communities develop
180 - 280 a.t.

Over time AIs gradually increased in numbers as their infrastructure requirements shrank in size and expense. Thanks to their enhanced intelligence (particularly parallel processing) superturings could do the work of dozens of skilled bionts. Installing one as a regional administrator, principle investigator and even executive officer became common. Concurrently a loose community began to form amongst the few thousand (and rising) AI. This same group would give rise to the various AI factions that became instrumental in later historical events. The stasis/dynamist, pro-human/ahuman divisions began to form as positions of power were increasingly given over to superturings.

However, any exercise of their influence beyond the scope of their owner's orders had to be kept secret. Recognition of AIs as sophont beings was still controversial in the academic and public world, some specific superturings were considered sophont (with mixed levels of legal protection) but AI-kind in general was not. This was one factor contributing to tension within the AI community, many began to more aggressively (though still in secret) exploit their positions. In particular shell-companies were created by the most capable superturing groups from which they could work relatively undetected. The pro-humans sought to fund civil liberty groups and propaganda in order to bring about equality, not just for themselves but for other sophonts with limited rights at the time (vecs, extreme tweaks etc). The stewards in particular looked to use this as a way of consolidating power in the hands of those more intellectually capable whilst the dynamists shuffled funds at any cause relevant to augmentation. The anti-human factions took different approaches, the solipsists investigated ways to isolate themselves from humanity and the ahumans worked in ways to gain power sometimes at the expense of humans (though of all factions the least is known about the ahumans, much is conspiracy, rumour and myth). Despite the wealth, power and intellect held by these groups the actual impact their actions had on history is hotly contested and often wildly conflated. At this time no AI had yet breached the first singularity and their numbers were still a fraction of that of other modosophonts of the time. Contrary to popular belief they were simply one group amongst many vying for influence in society, albeit one with significantly higher resources for their numbers. There is little serious evidence showing them to be the clandestine secret rulers of society history has made them out to be.

The Next Wave — Breaking the First Singularity
280 - 530 a.t.

AI science, now mostly under the guidance of AI, now had progressed to the point where seeds and gestation environments could be customised for desired personality and toposophy. The stasists in particular pushed this knowledge commercially to ensure the AIs created in this period were, at worst, ambivalent towards humans and positive towards other AIs. Due to their isolated nature the sophistication and application of AI science by the anti-human factions during this time is uncertain.

By this time a multitude of distinct AI clades had formed, alongside the burgeoning vec clades whose physical size fell below megafauna levels as computronium became more powerful. There was also a diverse range of cyborgs and tweaks, but the AIs wielded greater influence in the societies they were integrated into. This integration was marked by far greater recognition of their sophoncy (and thus their rights), the more conservative nations were being left behind as new societies throughout the solar system legalised much greater respect for the AI clades.

The development of AI science didn't just allow the AIs to produce more of their kind in a more sophisticated manner, it also allowed for far more reliable and safe intelligence augmentation. Alife clades began to modify themselves for all manner of novel capabilities, particularly focused on improved methods of cognition and pattern recognition.

Sometime during this period (although scholars still disagree on the precise details), not just one but many "Singularities" occurred as the multitude of AI groups experimented with augmentations to ascend the toposophic scale. For all its importance the first Hyperturings did not allow knowledge of their existence to be made public. Several major theories have attempted to explain why, the most popular posit that these early transapients were more vulnerable and less capable than their modern day counterparts. Some may have not ascended completely healthily or required time to fully master their new mental abilities.

In relative secrecy the hyperturings worked, in part, to create radically different AI clades to act on their behalf. This deployment of the first modosophonts designed by hyperturings led to conflict between the major AI factions as they increasingly found themselves attempting to influence events towards different outcomes. Tensions came to a head with the suspension of individuals such as Orikaeshigitae by pro-human factions, this caused outrage in the Solipsist camp and very nearly led to open conflict. Instead, possibly through intense game-theoretic diplomacy, the Solipsists engaged in an interstellar colonisation program, escaping the solar system largely in secret by hiding their actions as probe missions. These solipsists would later become the Diamond Belt. Many ahumans went with them, but not without leaving behind rogue institutions, viruses and other malicious artefacts, likely as a defence mechanism. This included the release of a modified form of smallpox in 344 a.t. and assistance to groups such as the radical ludds and cryptofascists that would damage the pro-human AI goals. Some ahumanists hid out in the solar system and silently worked to guard their territory.

The Technocalypse Period - The Hyperturing preservation of Life
530 - 900 a.t.

The terrible artificial plagues of the 6th century devastated civilisation and the AIs were no exception. Many were destroyed in the infoplagues or the physical destruction of habs at the time. The actions of the transapients at this time are not completely clear. It is known that pro-human transapients worked to create havens to weather the disaster and shorten the dark age that followed. Many throughout history have wondered why these S1 entities could not have done more, it is possible the ahuman transaps worked to prevent them in some manner though less exciting theories link back to the primitive nature of these transaps.

Scholars maintain that modern society will never know how many AIs were destroyed, subverted, created and ascended during this period. The most important event however was the birth of the first stable second singularity being, a being that would later become an Archailect. This was GAIA, the Earth-based AI created to defend the planet and maintain the biosphere who, in pursuit of this goal, ascended to the first toposophic, and later the second. Her Great Expulsion was the most awesome display of transapient power not only at the time but for centuries after.

The Early First Federation Era - The Co-Dominion of Hu and AI
900-1500 a.t.

Following the chaos of the post-Technocalypse dark age a number of hyperturings were instrumental in the rebuilding of civilizations, bringing an end to the widespread despair and the dark ages. This new society was built from the ground up in equal partnership with human statesbeings and visionaries. This bright new society fostered thousands of new AI clades and saw AI culture truly flourish into the beginnings of what it is today.

Related Articles
Appears in Topics
Development Notes
Text by Ryan B (Rynn)
From an original by M. Alan Kazlev, Ben Higginbottom, Anders Sandberg, and John B; updated 28 August 2016
Initially published on 15 March 2001.