Open Knowledge, and its Risks
Today, we have many organizations that are striving to organize knowledge and make it universally accessible: Wikipedia, Google, Open Knowledge Foundations to name but a few. While this knowledge is good for empowering people to solve problems, there is risk that making procedural knowledge easily available to people will have undesired consequences. For example, enable machines to self-replicate, or enable the creation of dangerous warfare. I'll talk more about the former, since the latter is quite well known.
The event of such autonomous self-replication could occur earlier than superintelligence, if some types of procedural knowledge are made available in computer-readable form.
First off, what I mean by procuedral knowlege?
Definition: Procedural knowledge is a representation of the outer world in an intelligent agent, such that the intelligent agent is confident that performing a certain sequence of known actions (programs) yiels a known result. This sequence of actions is to be called a procedure, the known result - a product.
Here, representation is the influence that the outer world had for the intelligent agent through the physical interactions ("education"); the intelligent agent is an entity capable of volition, cognition, action; confident means aware of high probability of success in yielding a result; known result - goal conditions to be met, known actions - actions performable by the agent (as script-like methods (procedures) are performable by an object in OOP).
The State of Today's Procedural Knowledge
Today most of the procedural knowledge is concentrated in the task management systems of diverse companies. Companies are using internal task management systems, as well as public (SaaS) ones, which are largely transparent to intelligence agencies.
The AI planning research is on-going in universities how to automate the planning of actions. The public sees the manifestation of such planning as the optimal road discovery algorithms in their driving directions. However, these same algorithms are being applied on other mathematical spaces, for example, the search of chemical reaction sequence to produce a desired compound (destination).
Each production machine, such as metal cutting, welding, etc. that makes something from something else is a piece of this know-how graph, representing a set of roads that can lead matter from one state to another.
Each manufacturing company, that makes something from something else, is a piece of this know-how graph too.
Almost surely such procedural knowledge map graph already exists. The theory is in our physics textbooks. The map is the databases of all tools available on equipment markets (Amazon included) with corresponding transformation functions for each equipment/device/tool (i.e., what transformations can the tool make to materials). Tools are no more than catalysts for shaping matter, therefore, even companies are tools to other companies. Each tool has properties, regarding what materials they can affect, and in what ways - these are the abstract "destinations" (vertices) of that graph, while the tools themselves are the "roads" (edges): G(tools,transformations).
There is no "driving directions" service to public, where you would enter a thing, e.g., "bolt", and it would output you roads (a combination of machines) to manufacture it, but it's highly likely such service exists.
The advanced (probably classified) research is highly likely to be carried out on using such knowledge to enable robots to mass produce the assets needed for military-industrial complexes. Such automation efforts are more likely to happen in the regions of the world, where human labor is scarce, IQs are high, and ambitions are great.Read more