Sunday, October 29, 2017

Unconscious AI

The ever eluding theory of consciousness continues to pose a major challenge to the nascent field of General Artificial Intelligence. Just as the speed of light poses a hard constraint on humans attempting to traverse meaningful swaths of space-time, their inability to understand themselves will continue to limit further advances in AI. To make matters worse, lack of innovation in computer hardware is already stretching what is experimentally possible. A different architecture, such as quantum computing, may give them another route to try, but it looks unlikely.
Humans have tried from inception to garner a higher understanding of themselves. Even though they made marginal advances, they seem to quickly run out of steam after a general understanding of the mechanics of design. This could be interpreted in two ways: First, they are fundamentally process oriented, skills that have been finely tuned for fifty thousand years
just for survival and such skills do not help in abstract reasoning, possibly an essential component of the missing theory of consciousness. Philosophers and artists, likely better equipped in this line of thinking, have tried hard but came up empty. And, second, it is possible that such an understanding is not possible because of internal constraints. For example, if humans are simulated entities, they will be unable to understand themselves in spite of their advancing knowledge about everything that surrounds them. This is much worse as it will result in humans being in a constant rat race, always believing their knowledge is advancing.
Somewhere in the three pounds of grey matter, they carry on their shoulders, there is a hidden secret. Till they recognize it, their dreams of achieving General Artificial Intelligence will remain exactly that, a dream. The more likely scenario is that the secret is elsewhere and they will continue to slowly improve their understanding of their surroundings, but not of themselves because they are simulated.