What really differentiates us from computers? Not much, if a new paper from Google is anything to go by. Although there is no doubt that computers are getting smarter, they are still behind humans in certain tasks. But new research from Google seems to have bridged that gap to some extent. Published on Tuesday, the Google paper unveils how the digital giant taught its DeepMind artificial intelligence (AI) system to initiate continual learning, so that the computer could learn like humans from past experiences. While humans use systems consolidation and synaptic consolidation—the process of imprinting on slow-learning parts and memorising—Google took the same process to create an algorithm called Elastic Weight Consolidation (EWC).
DeepMind now, much like humans, can remember how it completed a certain task in the past and solve problems. But what it cannot do is solve any new problems that humans still can. So, Google has been able to develop selective cognition. But given how technology is progressing, even that may not be far away. Although developing a smart AI is still some time away, we are reaching a stage where it can move beyond mundane tasks and do something more constructive. Whether it will be able to displace the human connect or how many human jobs it would make redundant is anybody’s guess.