In Douglas Adam's Hitchhiker's Guide to the Galaxy a computer called Deep Thought was asked to provide the answer to Life, the Universe and Everything. After the great computer program had run for "a very quick seven and a half million years" it answered: "42."
Although Deep Thought was fictional, its difficulty in solving a very human query shows us how challenging it can be to create true artificially intelligent computers - machines which possess human traits such as self-learning.
The possibilities of AI, such as driverless cars or robots that can 'feel', continue to fascinate. So much so that technology mavens including Google, Facebook and Microsoft are buying AI companies and advancing their own visions of machines with feelings.
To date, however, the quest for machines that have the characteristics of a human brain remains in its infancy. Today's AI devices tend to resemble single-cell organisms rather than full-fledged organic beings.
Rather than single organisms, some of today's smart machines can be like an 'exploded body' where the body parts send signals to a remote brain to analyze. For example, a power company with 1,000 smart wind turbines that send information to a centralized brain (computer algorithm). The turbines can tell the brain how they're running so if there's not much wind, it might spin up some diesel turbines. The wind turbines can also report on usage patterns for their sensitive internal mechanisms and the brain can predict likely impending failures and schedule a proactive maintenance call. The remote brain behaves in a manner that resembles instinct in nature - with an automatic response, rather than a thoughtful one.
Some increasingly smart and autonomous machines are also emerging, with smarter body parts integrated directly with the "brain". For example, the Google self-driving car has a vision system and other sensors and actuators in its "nervous system" communicating with the central 'brain', allowing it to steer, drive and avoid obstacles.
In business, the goal is to digitize information, analyze it and then make smart decisions based upon the analysis. Although existing technology can perform these tasks, and much faster than humans can, the challenge lies with adding the human qualities that can go beyond automated responses.
Science fiction can help us to understand these challenges. For example, in Star Trek Next Generation, Lieutenant Commander Data was a sentient, sapient android designed to self-learn the qualities and mannerisms of humans in order to actually evolve and become more human-like himself. Yet he remained a machine, even when he was give an 'emotion chip' to help him better understand human reactions.
Living beings must possess the following traits: movement, reproduction, specific organization, growth, respiration, irritability, excretion, feeding. Data had only two of these - movement and specific organization - or three (irritation) if you count his emotion chip experiment. Also, he was given simulated versions of feeding and respiration to fit in with humans. Robots such as Data are a long way away from the present capabilities.
Man is far from creating artificial life. And perhaps not certain what to create it for. The fear is always that robots will 'go rogue' if given enough sentience. When Data installed the emotion chip he was overwhelmed by conflicting emotions, and overloaded his positronic relays. He eventually learned to control his emotions, as humans must, but not before some terrifying experiences.
We have seen what havoc simple trading algorithms-gone-wild can wreak; the Flash Crash, the Hash Crash and Knight Capital's trading meltdown were all thanks to faulty algos. Imagine the consequences if truly intelligent, self-learning machines were programmed to take over financial markets. Or the world.
When creating intelligent machines, you really have to know what you are asking for.