Next generation of Artificial intelligence (AI) can be empowered from intensive studies of animal brains. Intensive research in the areas will further also help in solving difficult barriers in AI, explains a new theory.
Special efforts are drawn by neuroscientist Anthony Zador who has especially worked on understanding artificial neural networks (ANNs). These are known as computing systems, responsible for the success of the latest AI revolution, which is influenced by branching networks of neurons in human and animal brains.
Zador’s study however helps in fostering research in this revolution. According to Zador, learning algorithms are helping AI systems to gain superhuman performance with an increased number of complex problems such as poker and chess. However, if reports are to be believed, machines are still hindered by the simplest problems.
New research is now directing efforts in enabling robots to do organic activities such as building a nest or stalking a prey or even helping humans with mundane tasks such as doing dishes. The task of doing dishes is described by Google CEO Eric Schmidt as “literally the number one request… but an extraordinarily difficult problem” for a robot.
“The things that we find hard, like abstract thought or chess playing, are actually not the hard thing for machines. The things that we find easy, like interacting with the physical world, that’s what’s hard,” according to Zador. “The reason that we think it’s easy is that we had half a billion years of evolution that has wired up our circuits so that we do it effortlessly.”
This is one of the reasons why Zador believes that perfected general learning algorithm might not be the secret to quick learning. Zador observes biological neural networks created by evolution helps in providing scaffolding which enables quick and easy learning especially for tasks which are considered crucial for survival.
“You have squirrels that can jump from tree to tree within a few weeks after birth, but we don’t have mice learning the same thing. Why not?” Zador asks. “It’s because one is genetically predetermined to become a tree-dwelling creature.”
According to Zador genetic predisposition is innate circuitry which aids in determining an animal’s early learning. Furthermore, he adds the scaffolding networks are less generalized than perceived panacea of machine learning which is pursued by most AI experts. Zador believes if ANNs identified and adapted similar sets of circuitry, robots may be able to help in doing dishes in the future.
Zador’s study has been elaborately published in Nature Communications.