This post is part of the A to Z Challenge, in which each day of April a post is made inspired by a letter of the alphabet. Each post will be related to the research done on upcoming trends for the near-future techothriller series Shadow Decade.
Machine Learning is a subfield of artificial intelligence that covers the construction and study of algorithms capable of learning from data without being programmed to do so. In general, they build a model of the real world from input and use that to make predictions.
There are three categories of the tasks performed by machine intelligence:
This is the sort of intelligence shown in my story, Simple Harmonic Motion. A computer is being taught to operate a spacecraft, and the data given is the different crew members training it. Generally, that’s the form it takes: A teacher trains the computer to map behavior to the desired output.
If you want a copy of Simple Harmonic Motion, you can download the half-hour audio drama free by signing up for Burning Brigid Media’s mailing list.
Unsupervised leaves the algorithm to find its own patterns in order to solve a problem, or as an end in itself.
The algorithm interacts with an environment, working towards a stated goal, learning which actions are optimal – for example, in playing a game.
Generally speaking, the goal isn’t to teach the algorithm a task, but you’re trying to teach it how to learn tasks.
Types of Machine Intelligence
- Genetic Algorithms – Heuristics that mimic natural selection, using mutation to find the best solution or method for a given problem or task. Iterations can be processed very quickly.
- Decision Tree – The AI creates a map of possibilities and outcomes to maximize the liklihood of a desired end state.
- Association Rule Learning – The AI discovers correlations between variables; certain behaviors are likely associated with other behaviors, which allows prediction.
- Artificial Neural Networks – Algorithms inspired by biological neural structures using artificial neurons. They find patterns and create statistical structures.
- Inductive Logical Programming – Paradigms based on formal logic. They build logic statements of examples, databases, and hypothesis, and create a program involving all of the positive examples, then suggest a theory to explain the facts present.
- Support Vehicle Machines – Allow the classification of new data. Provided with samples of data, they categorize new data based on shared traits.
- Cluster Analysis – Observations are categorized into sets called clusters, based on whatever criteria are considered important. This is a method of Unsupervised Learning.
- Belief Network – Focused on the probable relationship between elements; what causes what. This can be used to infer probabilities.
Real World Applications of Machine Intelligence
- Search Engines
- Information Security
- Data Mining