It's been said that artificial intelligence is the future, and now, Australian scientists have succeeded in creating a computer program that was able to learn on its own, eventually performing a complex experiment with little to no supervision. The experiment in question involves the creation of a state of matter called Bose-Einstein condensate, which was only recently proven definitive and is considered one of the most difficult to pull off.
The AI assistant was able to learn on the job as scientists were in the process of recreating the experiment. Supposedly, the algorithm did not even need to be specially calibrated towards performing the experiment. It simply created its own process and was able to change certain parameters in its programming. Even more astonishing is the rate at which the algorithm was able to adapt to the task, as Australian National University in Canberra researcher and project co-lead Paul Wigley explains.
"It did things a person wouldn't guess, such as changing one laser's power up and down, and compensating with another," he said in a press release. "I didn't expect the machine could learn to do the experiment itself, from scratch, in under an hour."
The idea of using AI to perform many of the tasks that humans find cumbersome is not a new one. This particular case is actually one of the several recent attempts at developing actual learning computers which will then take over for their human counterparts. One of the most prominent German computer expert Jürgen Schmidhuber is actually attempting to create an AI that can replace him once he retires.
In terms of potential uses for the AI used in the Australian team's experiment, one of the most promising benefits they see for it is having a convenient research assistant. Instead of bringing an actual person wherever they go, scientists can simply bring the algorithm along, which will then help them with whatever experiments they are performing.