An international team of researchers has developed the first "intelligent" computer that not only conducts experiments, but designs the research and analyzes the results. The development is considered a breakthrough in the scientific field of "artificial intelligence," in which computers match wits with their human creators.
The first time most people were exposed to the idea of intelligent computers capable thinking on their own was in the futuristic 1968 film, "2001: A Space Odyssey." The star of the show was an onboard computer named "HAL 9000" that talked with the the spacecraft's astronauts as it piloted a mission through space.
But HAL made a "human" mistake that threatened the mission, and one of the astronauts, named Dave, was forced to disable the rogue computer.
From the time computers were conceived more than a half century ago, scientists have entertained the idea of smart computers capable of learning. But true artificial intelligence, emanating from metal and circuit boards, has been an elusive goal.
But in what is being described as a watershed experiment, scientists in the United States, Great Britain, and Canada say they have developed a computer model that can think on its own.
In an article published in the journal Nature, investigators describe how they broke the barrier to artificial intelligence. They built and programmed a computer robot that is able to design simple genetics experiments involving yeast, carry them out, and interpret the data.
"Once we have given it its basic background knowledge, then it just goes," he said.
In the experiment, lead author Stephen Oliver, of the department of genomics at the University of Manchester in England, says computer scientists knew the results of the test run. That way they could tell whether the robot designed and analyzed the data correctly
"So, now we are engaged in trying to push this further; to try to get it to actually discover something new," he said.
Vincent Hanover, a professor of computer science at Iowa State University, says many computers are already programmed to collect huge amounts of data. "But to the best of my knowledge, it is the first time that someone has actually put together a complete system with a robot, which is actually (able to) run actual experiments in the lab," he said.
That, he says, potentially has a broader application. "When you have a computational model of scientific discovery you can start asking questions about how discovery happens," he said.
For now, the experimental computer will remain within the realm of research into artificial intelligence. Scientists want to see whether the computer makes a human mistake, like HAL.