AGI (Artificial General Intelligence) can be considered the ultimate goal of AI research. It encompasses intelligence that is not limited to predefined knowledge in limited areas but includes capabilities to learn, reason, plan, solve problems, think abstractly, comprehend ideas and use natural language.
After a period of optimism in the initial stages of AI, to date, AGI (sometimes called strong AI) has been considered too difficult to achieve. During the decades that followed (the so called AI-winter), investigation has mainly been focused on small sub-problems of AI, trying to solve specific issues, for instance in the field of speech recognition, computer vision and games.
Reflecting on the reduced AI approach one can come to the conclusion that these specific problems are very difficult to tackle exactly because they are taken out of context. For instance, speech recognition is much easier if you have information about the subject of conversation or if you can see the person who is talking.
Now, new insights and increasing processing power open the door to an integral approach. Tremendous progress has been made in the medical field of understanding what is happening in the brain. This medical research provides us with important information on how to attack various problems. Furthermore, the modern day computers and computer farms are more and more able to perform parallel tasks, are very fast, and can manage a tremendous amount of information.
In the oncoming 5 to 15 years we will see a tremendous leap forward in artificial intelligence. And what is just as important, as we will see below, the field of robotics will advance equally.
Intelligent machines (or robots) consist of two distinct parts, just like modern computers: hardware (the robot) and software (the AGI system). In my vision, the analogy between them goes almost all the way. We will have various hardware and software manufacturers. The hardware manufacturers make and/or assemble robot hardware much the same way as the current computer hardware manufacturers. Some compatibility standard will be agreed upon, although certain manufacturers will stick to their own designs. The software companies will create the operating systems (or more specific, the AGI systems) for the robots, much in the same way as the current software companies create operating systems for the current computer hardware.
Users will buy a robot just like a modern day computer: choose a brand, select a model, opt for desired extensions and look for upgrades. Both hardware and software will be extensible and upgradeable, buying some more graphical resolution for the eyes, a more flexible robot arm, more long-term memory, faster processing chips, etc. The parts may possibly be purchased from different manufacturers.
The possibilities of the AGI technology are beyond imagination. They include household and military robots, toys, computer games, man-machine interfaces, intelligent cars, intelligent houses, personal assistants, professional assistants (expert systems), consultancy systems, security systems, strategic planning, complex control systems, military applications and much more.
Especially interesting is the possible spin-off technology that can be applied in a wide variety of applications, without the need of a full-featured AGI system. Examples include computer vision (security, quality control), speech synthesis (man-machine interfaces, public transport announcements), speech recognition (security, man-machine interfaces, subtitling) and robot control.