Just recently, researchers from the Bar-Ilan University in Israel have established a new type of ultrafast artificial intelligence algorithms with the combined use of advanced experiments on neuronal cultures and simulations based on brain dynamics. The new type of algorithm was said to outperform the learning rates of currently known learning algorithms.
The researchers state in an issue of Scientific Reports that they aim to bridge the gap between neurobiology and machine learning. At present, most people seem to think that the two are completely independent disciplines. Bar-Ilan University's Department of Physics and Gonda (Goldschmied) Multidisciplinary Brain Research Center professor and lead researcher of the study, Ido Kanter said that the absence of reciprocal influence between the two is puzzling.
Kanter likened the neurons of the brain with the bits in a modern computer. "The number of neurons in a brain is less than the number of bits in a typical disc size of modern personal computers, and the computational speed of the brain is like the second hand on a clock, even slower than the first computer invented over 70 years ago," he explained. "In addition, the brain's learning rules are very complicated and remote from the principles of learning steps in current artificial intelligence algorithms."
Traditionally, artificial intelligence algorithms are based on synchronous inputs, which means that well-defined inputs would predict the output. In contrast, brain dynamics do not work that way. It does not comply with synchronized inputs for its nerve cells. It is more complicated in a way that the biological scheme needs to adapt with simultaneous inputs, while physical reality develops. This part of brain dynamics is then a disadvantage.
To turn the table around, the researchers looked at this disadvantage as an advantage. This reflected in how their study showed that ultrafast learning rates are identical for both small and large networks. In addition to this, the researchers also found that learning can occur without the use of steps or modules, but through self-adaptation as well, based on asynchronous inputs. This is very similar with how the human brain works, particularly in the dendrites as observed in an experiment.
With their findings, Kanter and his team have definitely started rebuilding the bridge between neurobiology and machine learning, saying that the imperative center of artificial intelligence in the future should be the fundamental principles of how the human brain works.