Artificial intelligence has raced forward in the last few years, and these days, the electronic brains of the most advanced robots can surpass the average human intelligence. Though helpful in some cases, these systems are able to do things that might make you feel uncomfortable. Like what, you ask? Below are five of the creepiest advances in artificial intelligence, and there’s no sign of stopping it.
Schizophrenic supercomputer
Back in 2011, scientists at the University of Texas at Austin have simulated mental illness for a computer, testing schizophrenia on artificial intelligence units. The test subject, known as DISCERN, is a supercomputer that could function as a biological neural network and operate using the principles of the human brain. To recreate the mechanism behind schizophrenia, the team applied the concepts described in the theory of hyper-learning, which states that the schizophrenic brain processes and stores too much information by memorizing everything, including unnecessary details.
The scientists then emulated the schizophrenic brain in artificial intelligence by overloading the computer with many different stories. They found that at one point, the computer claimed responsibility for a terrorist act, telling researchers about setting off a bomb. It reported this incident because of confusion with a third party’s story about the explosion by terrorists, mixed with its own memory. In another study, the computer talked about itself as a third person because it couldn’t make out what it was at the moment.
Cheating robots
In 2013, Professor Roland Arkin from the School of Interactive Computing at the University of Georgia unveiled the results of an experiment in which scientists were capable of teaching a group of robots to cheat and deceive. The strategy of this behavior was based off that of birds and squirrels.
The study involved two robots. The first robot’s job was to find a place to hide, and the second robot’s mission was to locate where the first one was hiding. In this experiment, the bots had to make their way through an obstacle course with pre-installed physical objects that turned over as the robots moved along. The first robot paved the way, and the second one followed it by analyzing the tracks left along the path.
As time went on, the hiding robot began overturning obstacles to create a diversion and was hiding away from the mess it created. According to the researchers, this was not originally programmed, meaning the robot developed its own strategy through trial and error.
Ruthless robots
At the Laboratory of Intelligent Systems in Switzerland, scientists put a group of robots in the same room with predetermined sources of food and poison. The machines closest to the food earned points, while the ones who approached sources of poison lost points. All machines in the study were fitted with tiny blue lights that flashed erratically, as well as a sensor camera to help identify the light from the lamps of other robots. Each robot had the option to turn off its light if needed.
During the beginning of the experiment, the robots quickly learned that the largest congregation of blue lights was where the other robots congregated, next to the food. By blinking their lights, the robots showed competitors where the correct source was located.
It only took several phases of the experiment for the majority of the robots to turn off their lights, refusing to help each other.
Supercomputer with imagination
Among web-giant Google’s many projects, there is one that will always out from the crowd: a self-learning computer with a neural network simulation system.
In this study, the supercomputer was given free access to the Internet with the ability to examine the contents of the network. There were no restrictions or guidelines given, leaving the system to explore the entire human history and experience. What did it choose to look into? Images of kittens, just like some procrastinating people choose to do.
Later on, Google discovered that the computer developed its own concept of what a kitten should look like by generating the image with an analogue to our cerebral cortex based on a review of photographs seen earlier.
Future-predicting robot
Another self-learning computer known as “Nautilus” was fed millions of newspaper articles starting from 1945, by basing its search on two criteria: the nature of the publication and location. By using the information about past events, the computer was asked to come up with suggestions on what would happen in the future. Shockingly, the system had accurate guesses. It reportedly was able to locate Osama Bin Laden, for example. To be more specific, the Nautilus narrowed down the search area to a 200-km zone in northern Pakistan.
Source: learning-mind.com.
Advertisement
Learn more about Electronic Products Magazine
