artificial_instinct

This is an update from a post I made many years ago suggesting that the acronym AI should stand for Artificial Instinct rather than Artificial Intelligence. Specifically after being disappointed by my fourth year paper where it became clear how simplistic the systems were in tools like expert systems.

Intelligence : The ability to acquire and apply knowledge and skills

Instinct : An innate pattern of behavior in animals in response to certain stimuli

Artificial intelligence is a lofty goal when applied in the general sense. The majority of AI systems deployed are extremely specialised and while they often combine learning or training it would be a serious problem if they changed their behaviour and started acquiring and applying ‘skills’.

Ahead of the The Dartmouth Summer Research Project in 1955 John McCarthy coined the term Artifical Intelligence and defined it as “The science and engineering of making intelligent machines”.

I personally feel that instinct is probably a more appropriate term given the limited scope. This allows for learning and behaviour modification, but implies a smaller and more rigid set of skills.

A lot of the results from AI systems success in mimicking humans say more about the human physce and our ability to paint personality onto anything.

This Paper by Hector J. Levesque talks about the kinds of additional consideration we should be giving to testing for more general purpose cognition to avoid being tricked by search engines and to actually determine if the system can perform it’s own mental simulations.

Artificial intelligence is a wonderfully complex combination of processes we are still trying to describe and experiment with – yet today we have many many useful, specialised, functioning learning systems which do not even come CLOSE to being intelligent. Of course that’s one of the bonuses of acronyms you can happily decode to your own preference (ah the manifold nature of understanding).

Cheers
David