An Attempt To Define Artificial Intelligence (or rather Machine Intelligence)

An Attempt To Define Artificial Intelligence (or rather Machine Intelligence) Version 0.8 (Beta)

Disclaimer: This is my personal view and does not reflect the position of any of the entities I am engaged with. The below represents the current state of my thinking and might be significantly revised at a later stage 

There is a lot of discussion around Artificial Intelligence, yet there is no generally accepted definition of the term. As a matter of fact, the term itself is probably ill-chosen and many in the field are rather referring to Machine Intelligence these days.

Some definitions are somewhat misleading, particularly when AI/ MI is either strictly related to Machine Learning or is described as attempt to replicate ‘Human’ Thinking or Intelligence. Whilst there is a branch that focuses on the replication of human thinking, it would omit a very large area in the field and many of today’s intelligent machines do ‘think differently’.

The following is an attempt to make a contribution to find a universal definition. It is by no means perfect or complete. It should be rather seen as 'food for thought' to spur the discussion. Any reactions/ thoughts are more than welcome

Machine Intelligence is a non-human, man-created integrated system that is capable of autonomous informed decision making with the ability to communicate, implement, adapt and evolve these cognitive constructs in the real world

There are broadly two parts to Machine Intelligence

• The Core Cognitive Technology (CCT) which includes digital reasoning and the judgement process, which ultimately creates both insights and an understanding how to generate value from these insights
• Supporting Machine Intelligence Infrastructure (SMII) includes all that is needed to feed the core cognitive technology with the necessary raw materials and capabilities to act ie. data collection hardware (IoT), data collection software (voice recognition, NLP, computer vision etc), data cleansing/ analysis, robotics etc.

The following deconstructs the definition into its components to explain why these are relevant:

Non-Human: While trivial, there is a need to exclude human intelligence

Man-Created: Again trivial, need to exclude intelligent systems created by nature, including animals, plants, aliens etc

Integrated: A coherent system that is well synced between the various sub-components (ie CCT and SMII) and seamlessly exchanges information

System: As per above intelligence cannot be created through a single component but is rather a collaboration across different components

Autonomous: System should not have to rely on other external systems (ie. humans) to function intelligently

Informed Decision Making (IDM): System should be able to independently construct and evaluate wide range of potential action paths and chose the best option to maximise a desired outcome

Communicate: Communication among intelligent machine systems and between machine and human intelligence. Machines will not only need rational intelligence but also emotional intelligence (IQ and EQ)

Implement: Likely to be executed by human in the earlier stages of the technological evolution ('augmentation' of human labour/ intelligence), however advanced robotics increasingly allow for direct real world interaction

Adapt and Evolve: (Self-) Adaption and Evolution is a core capability of any intelligent systems as it ultimately is required to maximise chances of survival in an ever-changing environment

Real World: The system needs to be able to operate under real world circumstances rather than in artificial/ constructed environments

There are a lot of question arising from that definition, for instance what is a desired outcome, what factors should be considered when, what does emotional machine intelligence really mean etc. I will explore these aspects in further thought pieces.