What is artificial intelligence?
We explain the most important terms to help you get on top of the hype.
Artificial Intelligence or AI is a term coined in computer science. While often used for any system that mimics any kind of cognitive functions, such as learning or problem solving, the term actually describes the study of intelligent agents,a system that is able to perceive and its environment and maximise the chance to successfully reach a goal.
Whilst often used interchangeably with the term "ai", machine learning actually is a subset of the discipline, describing algorithms and statistical models that perform a task without explicit instructions and instead rely on patterns and inference. Machine learning has been around for quite some time. A typical example would be a spam filter that learns the patterns that identify unwanted emails.
Deep learning is a method of machine learning that utilities a concept called artificial neural networks and are utilised in fields including computer vision, speech recognition, natural language processing, audio recognition, machine translation, bioinformatics, medical image analysis, material inspection and cleantech, where they have produced results comparable to or in some cases superior to human experts.
The three paradigms of machine learning
The field of machine learning is manly split into three paradigms: Supervised Learning, Unsupervised Learning and Reinforcement Learning. Most production deployments, however, are often comprised of multiple agents, working in conjunction and often featuring multiple paradigms.
Supervised learning is a term used to describe mapping input data to output data in training to create an inferred function that can be used to label new production data. Classic examples of supervised learning include machine recognition of objects in photographs that are labelled before being used to train a neural network.
In contrast to supervised learning, unsupervised learning does not require labeled training data. It primarily used clustering and principal component analysis to define and extrapolate algorithmic relationships between datasets. This can, for example, be used for anomaly detection in very large datasets.
In reinforcement learning, the focus is on creating agents that take or recommend actions, balancing exploration of new data and exploitation of existing data. The main feature of reinforcement learning is a reward function that rewards and agent for taking positive actions when controlling an environment. This is, for example, applied when using an agent to control a process such as a control system for a manufacturing tool or a data centre cooling system.
Artificial General Intelligence
AGI or Artificial General Intelligence describes an AI agent that is able to learn any intellectual task the same way a human could. For example, it could recognise a zebra and tell it apart from a horse when only being shown one picture of each.
Currently, though not yet in existence, artificial general intelligence is the focus of many AI research organisations and universities. It differentiates from the systems currently being used, which are also referred to as "narrow AI" that are build for a specific task while artificial general intelligence is describes an agent that has the full cognitive abilities of a human. We believe that artificial general intelligence can be an extremely powerful tool to solve many of the world's problems, but, once developed, it should be applied with utmost caution and high regards to ethical standards. We at neuralfinity decided to define a strong collection of ethical guidelines that we apply to all of our work.