What is Artificial Intelligence (AI)
Following the first article in the series, An Introduction to the Applications of Artificial Intelligence (AI), we now explore what the term Artificial Intelligence actually means and some common categories of AI. The purpose of the article isn’t to give the reader a scientific definition of what AI is or is not, but a simplified and useful way to think about it.
While origins of AI can be traced back to Alan Turing, when he developed the concept of what we think of as a general purpose computer, the actual term was created by a computer scientist called John McCarthy. During the summer of 1956, McCarthy hosted the inaugural conference on AI with a group of renowned researchers in related fields. This included John Nash (depicted in the movie A Beautiful Mind), Marvin Minsky (Co-founder of MIT’s AI Lab) and Claude Shannon (father of Information Theory). The story of how AI evolved from the mid 1950s to today reads like an epic adventure with phases of time described as “the golden age of AI” and consisting of multiple “AI Winters”. I may summarise the history in a future article, but for now, let’s just say that AI’s history is almost as interesting as the kinds of problems it can solve.
When thinking about AI, you first need to understand that AI is a field of study, in the same way that Physics or Computer science are fields of study. Actually, AI is a sub-field of computer science and specifically focusses on the study of Intelligent Agents. It’s a simplification, but think of an agent as a device consisting of software and hardware. This is the first important distinction:
AI is a field of study that allows us to develop intelligent agents, it’s not the agents themselves.
Now let’s consider the term intelligent. When we think of intelligence, at first we might think of it being something intrinsically human. Thinking about it a little more, we then realise that some animals could be described as intelligent too, but clearly their intelligence is different to ours, far less general in nature. Therefore, it’s helpful to think of simple or narrower forms of intelligence when thinking about what AI is. Unlike humans, animals don’t ponder and pontificate, they sense their environment and then act in a way that will help them best meet their current goal. For example, imagine an insect searching for food. If it senses food in a particular location it will alter its course and head towards the food. This provides a useful way to think about intelligence and is close to how it's defined on wikipedia.
AI is the study of Intelligent Agents: any device that perceives its environment and takes actions that maximise its chance of successfully achieving its goals.
The last thing we will cover in this article is the difference between the terms Super AI, General AI, Strong AI, Weak AI and Narrow AI. That’s quite the list! When most people talk about AI, well outside of sci-fi movies at least, they are referring to Weak or Narrow AI, that is AI that can solve a particular focussed set of problems. AlphaGo would be a good example of an incredibly effective Narrow AI intelligent agent, able to beat the best Go players in the world. As some of the problems described as Narrow or Weak AI can be incredibly valuable and complex, the terms Weak or Narrow doesn’t do justice for what’s actually being achieved. To this end, narrow and weak are typically dropped and it’s simply referred to as AI. The terms General and Strong AI are considered to be the level of intelligence of a typical human being, while Super AI is intelligence that is past that of a human. Super AI is often associated with the Singularity Point, a time in the future when technology will advance at an exponential rate, with potentially dangerous implications for humans. Today, experts believe that we are still a long way off from General and Super AI with highly speculative predictions being around 2045. The key take away here is:
The term AI generally refers to Weak or Narrow AI - AI that solves a particular focussed set of problems.