Promises and Limitations of Conscious Machines

Abstract

A conscious machine is a complex system with a large number of sensor data, with the capability of learning and emergent behaviour. It should have the ability to understand and act properly in real-world situations. In the context of design of a machine with emergent behaviour, a number of algorithms play a significant role; notable among them are fuzzy logic, machine learning algorithms, neural networks, genetic algorithms and a class of cognitive architectures. Humans have a large number of sensors such as auditory, taste and visual; the signals from these sensors pass through neural systems involving processing and learning. Conscious machines will have to be integrated with the real-world input-output capabilities and they should learn from experience. The ability to generalise is an important feature of consciousness which machines lack at present. An expert machine like Deep Blue cannot play games other than chess. While designing robots with such features, ethics and laws should be formulated to address questions like, what will ‘they’ think of ‘us’? What will a driverless car think and how it will act if confronted with a dog right in front while driving? Any cognitive system that seeks to model human cognition and intelligence will have to model human emotion as well. Engineering such a complex system is a challenging task. The chapter will attempt to survey the challenging issues from an AI perspective, promises to those challenges, and the limitations of such machines in the context of consciousness.

There is no pillow so soft as a clear conscience.

(French proverb)