麻豆传媒

CHALLENGES don鈥檛 get much bigger than trying to create artificial consciousness. Some doubt if it can be done 鈥� or if it ever should. Bolder researchers are not put off, though. 鈥淲e have to consider machine consciousness as a grand challenge, like putting a man on the moon,鈥� says at the University of Palermo in Italy and editor of the . The journal was launched last year, a sign of the field鈥檚 growing momentum. Another landmark is the recently developed 鈥淐onscale鈥�, developed by Ra煤l Arrabales of the Carlos III University of Madrid in Spain to compare the intelligence of various software agents 鈥� and biological ones too (see diagram).

Perhaps the closest a , the Intelligent Distribution Agent built in 2003 by at the University of Memphis in Tennessee. IDA assigns sailors in the US navy to new jobs when they finish a tour of duty and has to juggle naval policies, job requirements, changing costs and sailors鈥� needs.

Like people, IDA has 鈥渃onscious鈥� and 鈥渦nconscious鈥� levels of processing. At the unconscious level she deploys software agents to gather data and process information. These agents compete to enter IDA鈥檚 鈥渃onscious鈥� workspace, where they interact with each other and decisions get made. The updated Learning IDA, or LIDA, was completed this year. She learns from what reaches her consciousness and uses this to guide future decisions. LIDA also has the benefit of 鈥渆motions鈥� 鈥� high-level goals that guide her decision-making.

Another advance emerged from designing robots able to maintain their function after being damaged. In 2006, at the University of Vermont in Burlington designed a walking robot with a continuously updated internal model of itself. If damaged, this self-knowledge allows it to devise an alternative gait using its remaining abilities. Having an internal 鈥渋magined鈥� model of ourselves is considered a key part of human sentience, taking the robot closer to self-awareness.

Along with an internal model, the robot developed by Owen Holland鈥檚 team at the University of Sussex, UK, is also anatomically human-like. 鈥淎 robot with a body that is very close to a human鈥檚 will develop cognition that is closer to the human variety,鈥� Owen claims.

None of these approaches solve what many consider to be the 鈥渉ard problem鈥� of consciousness: subjective awareness. No one yet knows how to design the software for that. But as machines grow in sophistication, the hard problem may simply evaporate 鈥� either because awareness emerges spontaneously or because we will simply assume it has emerged without knowing for sure. After all, when it comes to other humans, we can only assume they have subjective awareness too. We have no way of proving we are not the only self-aware individual in a world of unaware 鈥渮ombies鈥�.

鈥淵ou cannot prove that you鈥檙e not the only self-aware person in a world of unaware zombies鈥�

While we may never know for sure if a machine is experiencing consciousness or only appears to, building such a machine would revolutionise our understanding of the brain. 鈥淢y real goal is to figure out how minds work,鈥� says Franklin. 鈥淵ou really don鈥檛 know how something works until you can build it.鈥�

Return to the article鈥檚 main page: Picking our brains: Nine neural frontiers