Advances in Artificial Intelligence: Are you sure, we are on the right track?

TL;DR

This paper questions the current AI paradigm, proposing an integrated information framework that emphasizes physical and semantic information coupling, suggesting intelligence can emerge from simple information processing without mimicking the human brain.

Contribution

It introduces the concept of integrated information as a means to resolve duality issues in AI, advocating for a shift towards simpler, information-based AI systems independent of brain emulation.

Findings

Integrated information links physical and semantic data.

Intelligence can arise from simple biological entities.

AI development may not require complex brain-like models.

Abstract

Over the past decade, AI has made a remarkable progress. It is agreed that this is due to the recently revived Deep Learning technology. Deep Learning enables to process large amounts of data using simplified neuron networks that simulate the way in which the brain works. However, there is a different point of view, which posits that the brain is processing information, not data. This unresolved duality hampered AI progress for years. In this paper, I propose a notion of Integrated information that hopefully will resolve the problem. I consider integrated information as a coupling between two separate entities - physical information (that implies data processing) and semantic information (that provides physical information interpretation). In this regard, intelligence becomes a product of information processing. Extending further this line of thinking, it can be said that information processing does not require more a human brain for its implementation. Indeed, bacteria and amoebas exhibit intelligent behavior without any sign of a brain. That dramatically removes the need for AI systems to emulate the human brain complexity! The paper tries to explore this shift in AI systems design philosophy.