What Makes AI Intelligent? And When Will It Be?
Artificial intelligence or AI is a term that we hear pretty often these days. We have AI that beats us at chess and Go, we have AI that follows our shopping habits and gathers all of the best offers for us, we have AI that takes care of customer service, we have AI that drives us to work and home, and AI that cheats at online gambling (or if we don’t, we’ll have pretty soon). We can agree on what makes AI artificial – it’s a completely man-made collection of algorithms that learns and decides based on what it learns. But what makes it intelligent? And has it already become intelligent or is this something reserved for the future?
Is AI intelligent?
This is a question to which the best answer is “depends”. Depending on how you define intelligence, today’s so-called AI may be seen as pretty smart or incredibly dumb. And the definition of intelligence differs from one area to another – some see it as the ability to make decisions based on the information provided, others, as the ability to create original ideas. Emotion, creativity, and original thought are often included among the criteria used to define intelligence – and by this measure, today’s AI is nothing but a program that gathers information and reacts in pre-determined ways.
But there are AI systems that aid in translations, medical diagnosis, and autonomous driving – these are important, often vital areas in which pseudo-intelligence doesn’t make the cut.
When will AI be intelligent by all criteria?
What we call AI today is a very useful category of software. It can help in image analysis, it gathers and analyses large chunks of data, it can predict trends and invest automatically, it can predict malfunctions and help diagnose various medical conditions. At the same time, conversational AI systems like chatbots and conversational systems can be helpful in everyday life, and natural language processing applications can understand and categorize large texts like legal documents and research papers and categorize them by topics and associations.
One of the best-known criteria to test the ability of a conversational AI system is the so-called “Turing test” – it is a test of the machines’ ability to exhibit human-like behavior. If the responses of an AI are indistinguishable from a human’s in a natural language conversation. As of today, a single AI reportedly passed it – a chatbot called “Eugene Goostman” simulated a 13-year-old Ukrainian boy – and even this was contested by AI experts.
One thing is for sure: the artificial intelligence depicted in popular culture – from Commander Data from Star Trek to Skynet in the Terminator franchise or HAL 9000 – is still a long way away. It may be a long time before AI becomes truly intelligent… and we still don’t know whether we should praise or fear that moment.