“I’m not going to say I told you so, but I did” – Nouriel Roubini.
Many people will be thinking that Artificial Intelligence is just the stuff of science fiction, that it doesn’t really matter to our day-to-day lives. Well, it’s a big risk to continue with this view and just as the end of the sub-prime economic bubble was predicted by Roubini, so too do an increasing number of experts predict a great shift in our economy and society, this time brought about by tech innovation with Artificial Intelligence.
In this three part piece, I explore the issues around AI and why we all need to be thinking about them, today.
Part I – What is AI?
A google search of the phrase “what is ai” yields the result “a three-toed sloth”. A three headed sloth might be a more suitable explanation as development of AI has been a slow process and breaks into three general areas:
- Specific Task intelligence
- General Purpose intelligence
- Social and Creative intelligence
A. Specific Task intelligence
You might not believe it, but this type of AI has been slowly permeating our lives for some while. If you’ve received a letter through the post with a hand-addressed envelope, or if you’ve used an online shopping service such as Amazon recently, or even used voice control software on your phone – then chances are there has been some form of AI assisting in the background.
Specific Task intelligence, as its name suggests is simply an algorithm or heuristic that has been programmed to perform a specific task. What makes it intelligent is its ability to ‘learn’ from its mistakes. How, exactly? Well, we’ll come on to that in a future article.
An example of Specific Task intelligence is the computer ‘Deep Blue’ which famously (and controversially) beat chess world champion Gary Kasparov in 1996. Useless at forecasting the weather, Deep Blue was the first AI to beat a reigning world champion at a chess match.
B. General Purpose intelligence
General Purpose intelligence or ‘strong’ AI as the industry refers to it is one of the eventual goal of researchers in this field. Rather than starting from a position of looking at a particular task and working out how to ‘teach’ intelligence to a machine, instead researchers are trying to understand how, for example, the human mind works, and how a machine might be built that can be ‘smart’ at general problem solving.
An example of general purpose intelligence is the computer ‘Watson’ named after IBM’s first CEO, Thomas Watson. While not strictly a general purpose intelligence, as it was designed to solve a particular problem (in this case how to win at the TV game show Jeopardy!), the Watson architecture has been designed to enable to it to be ‘re-trained’ for a multitude of other scenarios, and thus it might be considered an early form of strong AI.
C. Social and Creative intelligence
“I propose to consider the question, ‘Can machines think?’” – Alan Turing (1950).
The ‘Turing Test’ is a widely known but little understood milestone in our journey towards creating AI which seeks to qualify whether a machine could sufficiently imitate human responses that it might be indistinguishable from a human.
Not only does developing an AI to interact with people as easily as we do with each other help us to understand ourselves better, but also it’s a necessary step if we are in the future to design machines to perform roles for us where social interaction is an important component (such as in healthcare).
Other milestones in this field of AI are in the Arts such as attempts to ‘teach’ computers to draw or compose music. Again, mimicry has been a key milestone, but one that has already been easily surpassed. To the untrained human ear, the Bach style ‘Prelude 29’ by David Cope sounds indistinguishable from one of the original compositions by the German composer, but perhaps to an expert of the baroque the ‘soul-less’ melody is all too naked.
So, in summary, AI is technology that ‘learns’ from its mistakes often designed by imitating how humans perform tasks or functions. In the next two parts I’ll be exploring whether all the current fuss around AI is really warranted, and also why you should care. Part II will be looking at the economic impact of widespread AI, and Part III will be considering the societal impact of AI and whether it truly poses an existential risk to our way of life?