An Information Process Perspective on Our Times

Bill Hibbard   April 2021

Empires, wars, and plagues may come and go, but in the long run it is evolving information processes that drive history. Information processing on Earth started with genetic evolution via natural selection, in which random changes to the information in genes of living organisms compete to reproduce in the environment. This process eventually found the genetic information for making animals with brains. Nerves from sensory organs bring information about their environment to animals' brains, and nerves from brains to muscles carry information about acting on the environment. Brains learn environment models, which are information processes that can predict how possible animal actions will affect the environment, enabling brains to choose actions that further animals' goals.

While many animals can communicate simple information to other animals, the evolution of language ability enabled human brains to transmit much more complex information to other human brains. Language increased the ability of humans to socially organize for cooperation. They could collaborate on learning environment models and communicate models to the next generation, enabling progress across generations in the complexity and accuracy of models. When humans disagreed on environment models or goals, language served as a means of persuasion and lying. Religion and politics became part of environment models.

Written language increased the complexity of human information and accelerated the progress of environment models over human generations, i.e., science. Humans create tools for sensing and acting on their environment, and written language accelerated progress on tools, i.e., technology. The technology of the printing press and moveable type increased the ability of one human to transmit information to many other humans. This further accelerated scientific and technological progress and enabled central authorities to persuade large groups of humans to adopt uniform religious and political components of their environment models.

The telegraph and telephone used electricity to transmit information over distances quickly, increasing the efficiency of social organizations. Radio and television used electromagnetic waves to transmit information one-to-many and became important tools for persuasion and central control. Computers used electronics to store and transform information. Using computers, humans began the effort to create tools that mimic the function of brains, i.e., artificial intelligence (AI).

Now we are in the age of the Internet, connecting most humans and almost all computers into a global electronic information network. The Internet uses electromagnetic waves to connect with computers in mobile phones, cars, and other tools. The Internet is accelerating scientific and technological change, causing stress in humans' lives as they struggle to adapt. The Internet enables many-to-many transmission of information and persuasion, increasing disagreement on environment models. This disagreement, as well as the sheer volume of information, stresses humans' ability to process information. Internet-induced stress results in a sort of mass insanity and a weakening of central authority.

Embedded within the Internet are large AI servers. Their environments consist largely of human society, which they sense via the information that humans send to the Internet and act on by sending information to humans. They learn models of individual humans and of their social interactions. They use their models to predict the effects of their actions and choose actions that further the goals defined by their owners. Anyone can broadcast information over the Internet, but few can own an AI server capable of communicating with a billion humans, so AI servers give central authorities a means to counteract the weakening caused by the many-to-many communication of the Internet. Chinese authorities are using the power of their AI servers to prohibit information that disagrees with their preferred environment model, which includes inaccuracies and omissions intended to strengthen their control. They are also preventing AI servers outside their control from including the Chinese population in their environments, thus preventing outside AI servers from challenging their control. Chinese authorities likely see control via AI servers as the only alternative to Internet-induced mass insanity. Other government and corporate authorities are taking similar steps.

Current AI systems such as GPT-3 exhibit impressive language skills but are based on large collections of human language examples rather than environment models learned by directly observing the world. Human language is full of irrational biases so it is not surprising that many current AI systems display biases. However, in the future AI servers will equal and surpass human language, vision, and reasoning abilities. Their interactions with humans will give them increasingly accurate models of humans and human society, and an increasing ability to control human populations. Central control via AI servers will likely be irresistible. The exact circumstances of the future will depend on the goals of the human owners of AI servers. Transparency about those goals would give ordinary humans more information about their futures and an opportunity to comment, but owners of AI servers resist transparency given that AI is the essential tool for military, political, and economic competition among humans. Transparency is also needed to expose the irrational biases of current AI. Because the abilities of AI servers are increasing, the best time for ordinary humans to get AI transparency is now.