Written by Whilst precursors to AI dating all the way back to references about mechanical men in Ancient Greece, modern concepts of AI really emerged in the 1950s.
Alan Turing devised the Turing test in 1950 whereby a person would be asked to determine whether they were conversing with an unseen machine or a human.
Marvin Minsky was involved in organising the Dartmouth Conference of 1956 with John McCarthy Claude Shannon and Nathan Rochester of IBM. The proposal for the conference included this assertion: “every aspect of learning or any other feature of intelligence can be so precisely described that a machine can be made to simulate it (see https://en.wikipedia.org/wiki/History_of_artificial_intelligence).
This period also saw the rise of symbolic reasoning and logic whereby machines could use symbols as a representation of the human thinking process and hence the result would be thinking machines. This was also the period when the concept of strong AI was proposed by John Searle as a means where machines could replicate the capabilities of then human mind (Russell & Norvig 2003, p. 947,952).
Weak or narrow AI – AI performs a specific (narrow) task but cannot generalise to other tasks.
Strong AI or Artificial General Intelligence (AGI)– AI can generalise across multiple tasks. So it can apply what it learned from one task and apply that learning to a completely different task.
Super Intelligent AI – transcends AGI and becomes smarter than humans.
A number of classical AI methods that developed such as search algorithms that used depth first search, breadth fist search uniform cost search, A* algorithm with heuristics, local beam search and other similar techniques.
None of these really scaled very well with larger search space. Computational cost became high with very large data sets. Also symbolic AI although still used today never really scaled across machine intelligence in the way that the earlier AI researchers had hoped.
Instead real world application of narrow AI that we have today focus on machine learning and deep learning that is still focussed on narrow AI.
Machine learning is a subset of AI whereby the machine learns to identify patterns from data.
Deep learning has been explained in a previous post but briefly it is a subset of both machine learning and intern AI and is the cutting edge of where we are today with application of artificial neural networks.
The exciting future developments relate to Deep Reinforcement Learning, GANs, capsules and Bayesian Neural Networks each of which will be visited in more detail in future blogs.