Tuesday, November 12, 2019

No Artificial Intelligence without Consciousness


Artificial Intelligence, a nebulous area, has been around from the advent of computers. Every decade, aided by increasing computing power and cheaper memory, those who just got out of school start to believe they found something new. Most often, new terms are invented to relabel what has been known forever. In the latest iteration, terms such as machine and deep learning have been trending. More interestingly, in the current wave, a new profession is coined, aptly called, "data science." Consulting firms, running out of ideas, strategies and PowerPoint magic, have been jumping in, to make a fast buck. The larger ones have assembled "thousands of data scientists," to make AI for their clients. The smaller ones have raised many 100s of millions of $ to "change the world." Now that we are approaching practical quantum computing within a decade, the next wave is just about to start. The behemoths, stuck with excess cloud capacity, have been providing "tools," so that they can download the costs of the stranded investments to the users. Unfortunately, all of these could be rendered obsolete in a few years. It may be a warning sign for educational institutions scrambling to create more data scientists on-line or not.

Autonomous automobiles and aircrafts are not AI, they are transportation modalities with a computer onboard. Robots that can put nuts and bolts together, assemble objects of use and occasionally jump in magnificent ways are not evidence of AI, just expert logic embedded in mechanical systems. Fooling people into thinking there is a human on the other side of the telephone is not AI just a set of rules fed into a synthesizer. Machines beating humans in prescriptive games is not AI - they are either a massive set of rules fed into high powered computers or pattern-finding neural nets (some call it deep learning) on steroids. None of these use cases have anything to do with AI, generalized or not. They just make some feel important and make a lot of positive economics for their proponents.

However, we cannot move an inch forward in AI without a coherent theory of consciousness. Engineers have been on a quest to define what they do not seem to understand, by quantitative means. It is possible that consciousness is a property that is externally applied. If so, the entities with consciousness are unlikely to understand it. In the absence of a theory from within, one possible explanation is that consciousness is induced by the simulator of the game. If so, it is likely that consciousness is a democratized property and is not limited to humans, let alone living things. This may explain why humans locked in a mathematical jail seem unable to understand it. 

Those chasing AI may need to spend more time thinking about a possible theory of consciousness. Without that, it is just age-old statistics.

No comments:

Post a Comment