By Mark Davidson
For most people the term "Artificial Intelligence" conjures up images from movies. The menacing killer robots from The Terminator or perhaps something more subtle and sinister like HAL from 2001. Definitely entertaining and thought provoking but the concepts covered are more science fiction than science.
More recently, there appears to be interest in the field based on books written by experts mostly in a despondent tone. Several years ago, Stephen Hawking, Elon Musk and dozens of other AI experts were signatories on an open letter calling for research on the societal impacts of AI and raised concerns about the existential threat that AI has on humanity.
Why is there such interest in AI and how is it really impacting our life? In this series, I want to present a more optimistic view of AI that could benefit us rather than destroy us. But first I want to take a look at the fascinating history of AI which goes back a long way to the beginning of modern computing.
We start our history lesson with the English Mathematician Alan Turing. Turing is a pivotal character in the development in modern computer science. In the 1930s he laid the theoretical concepts of computer science through the formalization of algorithms and computation known as the Turing Machine. He helped with the British war effort during World War II as a cryptographer and helped to decode German communications. In the 1950s he was persecuted for his homosexuality but was issued a posthumous pardon by the British Government in 2013. Alan Turing was one of the first to postulate, “Can a machine think?”
He developed The Turing Test which is a simple test to determine if a human can detect if a computer is acting like a human in conversation. In modern times, we have achieved various levels of success in the development of chatbots or “socialbots” for technical support and other purposes. After interacting with one for a few minutes you can clearly ascertain that it’s not intelligent. Today, Amazon is continuing the Turing test legacy by sponsoring the $1,000,000 Alexa Prize - in which a socialbot must have a natural conversation for 20 minutes with the judges.
The formal beginning of AI was at the Dartmouth Conference of 1956, organized by Marvin Minsky and John McCarthy among others. The proposal for the conference was "every aspect of learning or any other feature of intelligence can be so precisely described that a machine can be made to simulate it”. McCarthy persuaded attendees to accept the term "Artificial Intelligence" as the name of this field. The conference was the moment that AI gained its name, its mission, its first success and its major players.
The field of AI continued to be popular in the 50s and 60s and we seemed to be on the cusp of machines that could think. In the 60s there was a theory that intelligence could be programmed. Smart people would be able to write the rules and instructions, triggering methods of search and recall in machines. This is known as "Symbolic AI" and the practitioners of these methods were known as the Symbolists. During this era, thinking machines were just 10 years away (the next technological breakthrough is always 10 years away!)
There was a lot of research into the brain functions and theories in the 50s and that influenced psychologist Frank Rosenblat to create the perceptron learning algorithm. Basically, it’s a series of algorithms that recognize relationships in a set of data - mimicking the connections between neurons in the human brain. These neural networks can adapt to changing input data by adjusting the weighted values of the connections so that the output data can remain consistent. The practitioners of these neural network-like calculations were known as the Connectionists.
It’s a difficult concept to fully grasp but using data to create a model as the Connectionists did is the complete opposite of the "rules and recall” approach of the Symbolists. Instead of writing rules and triggers for recall, the neural net approach learns from input data. Rosenblatt put his perceptron theory into practice with a machine that could recognize letters using a 20x20 array of photocells. It was a monster of a machine but it worked. Rosenblatt claimed that their system could be trained to see images, beat humans at chess and even reproduce. You’ll notice a common theme of the field of AI is a propensity towards hype and inflated expectations.
So the players in the field of AI in the 60s were the Symbolists with their “rules and recall” approach and the Connectionists with their more organic system based on learning from data. In Part 2 of this series we’ll learn how this environment, and the unfulfilled promises of both, led to the first “AI winter”.