Intelligence is the ability to learn, understand, or to deal with new or trying situations in the pursuit of an objective. Today most information systems show little intelligence. They are machines, and they are programmed to work the same way each time we use them. These systems work well when there is no change in the environment in which the machines operate, but they fail when the environment changes.
We try to think of all the possible ways the environment will change, and we try to build our information system machines to cope with environmental change. Unfortunately, we can’t possibly imagine all the ways the environment will change, but we try hard. When we come up against an unanticipated situation we change our machines to cope with the changed environment.
Information systems built as machines are expensive to build, operate and change. We see this every day with the work of people involved government as they try to work out policy and the rules of interaction. It turns out to be very very expensive to alter the way their information systems operate. Policy makers try to articulate what it is they wish the government to do and then specify the information machine to do it.
In a modern society where billions of people and organisations interact the effort of maintaining machine information systems is much higher than the cost of the goods and services these information systems support.
We can make our existing information systems smarter with techniques like artificial intelligence, but these are band-aid solutions. To make a real difference and to reduce the cost of information systems we need to build intelligent information systems that adapt and change themselves.
Examples of Intelligent Systems
If intelligence is the ability to adapt and change with changing circumstances all life-forms show intelligence. Life-forms adapt and change and learn to fit in with their environment for the life-form to survive with the least cost and effort. Individuals do not survive, but the species survive. We call the process evolution. Individuals change in small steps, and those best fitted to the environment survive longest and are more likely to reproduce.
Humans have evolved to communicate knowledge between individuals. The species is able to pass survival skills between members of the species through the information system we call language. We think the most advanced information system that exhibits intelligence is the human brain. If this is the case, then we would be wise to build our information systems to operate in a similar way to the brain.
People have long thought of computers as being electronic brains, and they have the capability of becoming more brain-like, but only if we structure the information systems to work more like a human brain. The good news is that we can make intelligent information systems from all existing information systems that use computers as a tool by making the information systems brain-like.
Heuristics and Information Systems
Our brains and life-forms information systems use heuristics to make decisions and to take action. Heuristics are shortcuts that reduce the cost of making a decision. It is too expensive to work out the ramifications of any decision and so we invent shortcuts or heuristics to reduce the cost.
Examples in the brain include using a rule of thumb, an educated guess, an intuitive judgment, guesstimate, stereotyping, profiling, or common sense. In life-forms, the rule of thumb is survival of the fittest. It is a way of communicating information between generations. Changes that reduce the chance of an individual surviving means those changes disappear from the life-form.
Today almost all information systems use life-form heuristics. Information systems work by repeating what worked last time. If it doesn’t work, we change the information system so that it will work as expected and hope that it doesn’t have secondary effects or emergent properties that give different outcomes in other situations. Anyone who has ever tried to change a large information system knows they are very, very difficult to change.
We need to find better ways for information systems to evolve. We need to go beyond machine heuristics.
Metaheuristics and Information Systems
A metaheuristic is a heuristic designed to find, generate, or select a heuristic to provide a sufficiently good solution to an optimisation problem, especially when we have incomplete or imperfect information or limited computation capacity. We use search metaheuristics to model information systems, and the techniques are widely known and used in forecasting. Engineering and Science use them to get answers to hard problems that have no known analytic solution. Metaheuristic solutions dynamically change the heuristics used over time.
Weather forecasting uses metaheuristics to predict the weather by modelling the atmosphere as a complex adaptive system. It works by dividing the atmosphere into cells and then treating each cell as an independent entity. Each cell interacts with neighbouring cells. The outcome of the interaction is probabilistic as it is uncertain how the cells will interact. The weather forecasting program runs many many simulations starting from known conditions and plot the results. Forecasters analyse the results and come up with a prediction and probabilities for temperature, rain, and wind.
Our information systems can operate the same way as weather models. In an information system for the economy, we have individual entities who interact with other entities. Each entity acts in a way dependent on what happened previously but how it acts depends on how other entities act. The rules it uses to determine how to act can be metaheuristics. The approach contrasts with machine information systems that use fixed heuristics to determine what action each entity will take with the passing of time.
We can build our information systems to operate with metaheuristics rather than fixed heuristics. When we do, information systems become low-cost and adaptable. Some examples of how to build intelligent information systems follow.
Information Systems to Fund Infrastructure
Typically infrastructure is constructed by a single organisation for use by a large number of people. The funding to build infrastructure comes from large third parties such as investors and financial organisations. We do this because of difficulties organising all the users to fund the infrastructure over many years. A large stable funding organisation accommodates changes to the environment by charging a fee for the use of money. To provide certainty the funding organisation uses machine information systems as it needs to remain stable and predictable.
The organisation that owns the infrastructure is often a local government organisation, and they pay investors for the use of their money to pay the builders. User fees on the infrastructure cover the cost of construction and the cost of operation. Again the information systems of the government are stable and machine like.
For a single infrastructure investment we can remove the need for the machine information systems of a financing body and the government. We can allow the people who use the infrastructure and who provide the funds to work together using an intelligent information system.
Metaheuristic intelligent information systems provide a way for a large number of people to coordinate their actions without the need for a third party funding organisation and outside the complexities of government. It means the people who use the infrastructure can fund the construction and the operation of the infrastructure and save the cost of renting money. For long-lasting infrastructure, it reduces the cost of providing infrastructure by at least a half and often more. The operation of metaheuristic information systems is low cost because they dynamically adapt and modify themselves with the passing of time and the changes to the government environment.
It is important that the governance structures of metaheuristic information systems are themselves metaheuristic information systems. Without this, the information systems soon degenerate to heuristic systems as some entities within the systems impose heuristics favourable to them at the expense of other entities.
A Metaheuristic for Identity
Identity is the ability of a person to distinguish another person from all other persons. A metaheuristic for people is the mutual identification of two people. If a person identifies a person and vice versa then if they meet again they mutually identify each other. A common heuristics is mutual facial recognition of past encounters. Other heuristics are common memories or common acquaintances. A metaheuristic identity system builds an identity on any combination of heuristics of mutual identification. It enables the success or failure of any heuristic to pass on identity information. Electronic communications adds many more heuristics for mutual identification
Mutual identification can apply to things or animals or people. So a person who uses a car identifies themselves to the car through the use of the heuristic of starting the car with a key and with an independent mutual identification of a location heuristic.
It is this combination of heuristics and choosing the most appropriate heuristic that makes metaheuristic identification secure, private and low-cost. Privacy and security come because only the parties mutually identifying each other have the information about their connection. Strength of identification comes from the number of times there has been mutual identifications and from the network of many pairs.
An identity system built this way is robust as there is no single point of failure. There is no central collection of identity. Identity happens afresh each time a connection or re-connection is made. The system adapts incrementally as the identity characteristics of people, such as appearance, change over time.
Summary
All machine like information systems have some part centralised. All client-server systems are machine like. We can turn any machine information system into an intelligent system by making all the clients independent and equal to the server. Each entity in the network is both a client and a server. By distributing control and with it the data associated with actions we can turn all our information systems into intelligent adaptive life like systems.
